<?xml version="1.0" standalone="yes"?>
<Paper uid="J79-1049">
  <Title>Association for Computational Linguistics</Title>
  <Section position="1" start_page="0" end_page="0" type="metho">
    <SectionTitle>
KURZWEIL READINGMACHINE ................ 96
TRANSLATING MACHINE DEVELOPED ............ 97
AMERICAN JOURNAL OF COMPUTATIONAL LINGUISTICS is ~ublished
</SectionTitle>
    <Paragraph position="0"> by the Center for Applied Linguistics for the ~ssbciation  This paper describes a program which translates Engllsh into French, It is difficult to delineate the subset which a program can deal with, so sample sentences are given. The analyser is multiple path, single pass, akln to Woods' A.T.N. grammar. The syntax is dealt wi$h by matching. J with $emplates; the semantics by the application of semantic re$trictions to syntactically associated pairs of word meanings. To lirn~t the number of paths, all available guns, syntactic and semantic, are brought to bear at every stage. The output is a list of disambiguated word meanings, formed into a tree structure but with semantic rather than syntactio relationships betweep them.</Paragraph>
    <Paragraph position="1"> The French generation first makes appropriate tense changes, then finds the French word(s) and redistributes them if necessary. This may generate a French structure radically different from the English, Then the words are sequenced and put into the correct form.</Paragraph>
    <Paragraph position="2"> The program consists of about 8,500 Fortran irlstructions and the processing averages about 15 seconds per word on a 36.0-40.</Paragraph>
  </Section>
  <Section position="2" start_page="0" end_page="0" type="metho">
    <SectionTitle>
TABLE OF CONTENTS
</SectionTitle>
    <Paragraph position="0"> ............................... Sample Sentences ;..... 15 ................................ Description of Method 19 Flowchart ............................................ 2i ................ Reduction of the Passage to Base Form 22 ................. Generation of the French Translation 46 ........................................... Conclusion 48  My *banks are due to the Managers of IBM Near East and+IBM Lebanon and many members of the staff of IBM Lebanon, who have helped and encouraged me in every way. I am also indebted to Yorick Wilks, Margaret King and Walther Bischoff of the Fondazione dalle Molle, who have been very he3pful with their a advice and ideas.</Paragraph>
  </Section>
  <Section position="3" start_page="0" end_page="6" type="metho">
    <SectionTitle>
INTfRODUCTION
</SectionTitle>
    <Paragraph position="0"> There are, I hope three reasons why CLAM will be of interest to computational linguists.</Paragraph>
    <Paragraph position="1">  (1) Ir; 1s a working model. This is not a &amp;quot;paperf1  containing ltideas&amp;quot;. Lt is a description of a model which works, To be more specific, it is a description of a large program, written in FORTRAN, which runs on a 960-40. It accepts as input English $ext, carrieg out a syntactic and semantic analysis of! it, stores the result, and translates it into good French. (2) The subset of English which it is capable of analysing is, by present standards, extremely large, The vocabulary is about 1300 words, many of which have a variety of meanings. More important that the size of the vocabulary is, of course, the range of syntactic structure's and, perhaps most significantly, the degree of complexity of senkences which can be dealt with, Increasing the length and complexltg of sentences does not bring likelihood of combinat orial explosion. The amount of working store and computing time required to analyse a sentence is of the order of the number of words in a sentence, although of course it varies according to the number of meanings of the words and the types of syntactic structure involved.</Paragraph>
    <Paragraph position="2"> (3) The program is continuously extensible. This extensi'bility applies first t~ the subset of language which can be analysed, secondly to the target languages into which translations can be generated, and thirdly bo the uses to which the analysis of the rext can be put. Tn other words, I believe that the program embodies 3 sound method of syntactic and semantic analysis such as must be the basis of a computer lawage model. Extension of the subset of language which can be malysed is a matter of addition and refinement. It can be stated with confidence that such extension can be achieved because nothing fundamentally different from what has already been achieved is involved. New syntactic structures, well formed or otherwise, can be incorporated, by addition partly to the files and partly to Ohe program. Continual refinements can be made to the method of finding pronoun antecedents. This problem, which seems to be generally accepted as the most difficult single problem in analysis, will never be solved by one simple algorithm, and the fact that a particular program at any given stage of its development gives the wrong answer in a particular case, so far from invalidating the prograp, rather points the way to further refinements (cf. Wilks, June 1975). What is important is that the program should provide the tools which enable the refinement to be made, and CLAM does this.</Paragraph>
    <Paragraph position="3"> Extension of the target languages involves applying to other languages the same method which is used to generate French. This can be done, and indeed part of the actual program used for French would be generally applic3ble. It will be interesting to attack a language outside the Indo-European group, and Arabic is the first one I have in mind, although how soon this can be done is a question of time and priocities, The obvious use to which the analysis can be put other than  translation is a questionLanswer system, and work on this is at present in hanb. A question-answer system must be based on an effective malyser, and it is beljeved that CLAM can provide this. However, I do not maintain that the analyser Sb~uld be independent of9the memory and inferencihg part of the sy$te~n. Obviously it should not be independent of the memory, since an analyser nu st create and use its own memory, and although it would be theoretically possible for the analyser to have one type of memory and the latter part of the program to have another, this would be a ludicrous arrangement. The same argument applies tp inferencing, which again has to be performed. by an analyser. Therefore it seems that a question-answer system should be more integrated than rnany A.1. reseqrcher,~ appear to allow. On this score, I support the view of Wilks vis-a-vis Charniak.</Paragraph>
    <Paragraph position="4"> To create a que~tion~answer system, and, indeed, to improve the translation program, -t;he memory and the semantics of the present program have to be developed. I use the word &amp;quot;developed&amp;quot; advisedly because I believe that the ekisting memory and semantics form a sound basis upon which a more comprehensive system can be built.</Paragraph>
  </Section>
  <Section position="4" start_page="6" end_page="9" type="metho">
    <SectionTitle>
RESULTS
</SectionTitle>
    <Paragraph position="0"> Assessment of the Subset of Language which the Model can Analyse It is normal practice when describing a language model to leave d;iscussion of the achievedents of the model until the end.</Paragraph>
    <Paragraph position="1"> First comes the description of how the model doeslor ~ould operate, then, if it is actually in operation, an account of what &amp;t can do. En this description I am reversing 3he procedure, becauge I would not like to think of a reader ploughing through details of how something is done ii he subsequently comes to the conclusion that what was done was not worth doing anyway, Let him first sge what can be done, and then decide whether it is w~rth the trouble of reading on to discover how it is done.</Paragraph>
    <Paragraph position="2"> Having said this, I am immediately conYronted by the problem adrunbrated by Woods of how a reader can assess the range and scope of a particular model, and by implication, of how the programmer can honestly present it. There are two standard methods of prepentation. One is by rather sweeping general statements such as &amp;quot;the program can cope with noun clauses. adjectival clauses, conjunction, questions&amp;quot; etc., according to what claims are being made, Such generalisations are inevitably suspect and rightly so, since no reader will believe that he could not find, for example, adjectival clauses which the program could not cope with. The alternative method of presentation is to give sample sentences which the program has coped with, and hope that the reader will make for himself the type of generalisation which the programmer has scrupulously avoided. If the first method is adopted, the programmer may justifiably be branded as a charlatan. If the second, he runs the risk of having his sent-encesdismissed as &amp;quot;a few examples&amp;quot;, The problem is real, and the solution far from obvious: how to define a subset of language. Supposing that we were conCerned only with single sentences and not longer texts; and supposing that it were possible, which evidently it is not, $0 list all the sentences of the subset: the11 how can we find a definition which ,wou&amp;d include all the sentences which we have listed and exc1ud.e any which we have not listed? Two things are clear. The definition would be very long, ahd it would contain an agglomeration of embedded provisos, For example + the section on relative Clauses might include something like this: Relative clauses are admissible, provided that  1 . they do not contain more Zhan seven words ; 2. they do not contain a passive verb unless (a) it is a verb of Icookmgt ol- (b) the clause is a vsubject' clause; 3. there is no word between the noun and the relative  clause unless it is part of a supervening relative clause; 4, the noun 1s not part of a subsidiary clause unless the subsidiary clause is itself a relatlve clause provided that (a) the noun is not the object of the clause and (b) the noun is not a 'time1 noun.</Paragraph>
    <Paragraph position="3"> All of the above provisos are of a type which could well be applicable at any particular stage in the development of a program, although some may be more likely than others. The programmerls difficulty is that until he has tried an appropriate type of sentence, he probably will not realise the existence of a  particular limitation. The first indication of it is that the sentence doesn't work, and he then has to rack his brain to findrout why not, and alter the program to eliminate the limitation, thereby enlarging the subset in that particular direction (hoping that he is not at the same time being so stupid as to reduce it' in another). Therefore if a programmer asserts that his program can deal with e.g,, relative clauses of all types, he is probably not being dishonest but m~rely ignorant about the limitations of his om program. Whether such lirnitat50ns should rightly be described as bugs, which Wood,s implies, is dubious, because that is tantamount .Q;o execting a program which can deal with some relative clauses to be able to deal with all relative clauses. and asserting that insofar as it cannot, something has gone wrong. Rather might one think of a program in terms of a pool of water spreading slowly over an area and gradually covering more and mare of that area. The fact that the water covered a particular part of the area wottld carry no implication of covering any other part, although ther'e would be a reasonable expectation of its spreading to a contiguous area next. This analogy, though valuable in helping to destroy a misconception, is evidently incomplete in two respec%-s. It is two-dimensional, whereas.language is #multi-dimensional; and the program would advance not continuously, like a pool of water, but by fits and starts, in discrete steps. Each of these points is worth further examination, Lip service has long been paid to the multi-dimensional nature of language, and yet the importance of this aspect in  attempting to analyse language has rather slowly comb to be recognised, How many features are there whlch have to be talcen into account, and what are they? How many possible relationships can exist between t~hiclz of tlle~n? And, a rluostion raised with particular force by coniputer analysers, what combinations of features are relevant? As a simple esaztiple, consider two features, both as it happens syntactic although the argument applies to semantic as ~811 as syntac-tie features: ralative clauses, and the passive voice. If a program can analyse each of the features separately, does it follow that it can analyse them in combination? Suppose, for example, it can arlalyse both of these sentences:  1. The man who came to dinner stole the silver.</Paragraph>
    <Paragraph position="4"> 2. The man was hit by a bus.</Paragraph>
    <Paragraph position="5">  Does it follow that it can analyse this? The man who was hit by a bus stole the silver.</Paragraph>
    <Paragraph position="6"> Alas, it does not. It may iin fact be able to, but there is no logical rule from which it can be deduced that it must be able to. Is the absence of' suc a rule merely a computational quirk, or does it corres~ond to some linguLstic truth? In this case, but not necessarily in all such cases, I would say that it does so c~rrespond. We may asK ourselves if it is possible to imagine a language in which relative clauses e&amp;ist, and the passive voice exists, but in which the verb In a relative clause cannot be in the passive. Of course it is, and *here may for all I Know be such languages. It +is this absence of a logical rule of combination which makes the task of defining the bounds of a subset of language so appallingly difficult, especially when it is remembered that it is not merely combinations of two features, 9s in the above example, but combinations of many features which have to be taken into account. Multi-dimensiona3l.y is such a cardinal characteristic of language that analysgrs dealing with tiny subsets from which this characteristic lias been removed should probably be treated with reserve. They may give valuable insights, but they may also be nxisleading, I do not of course refer to the memory oP inferencing part of microworld models, which is usually their raison dletre, but: only to their interfade with natural language input, The designers of such models are inclined to regard the input analysis as little more than a tedious chore, and would be unllikely to take exception to what I am saying since they themselves normally make no great claims fox this part of their models. But ethers, commenting on the moaels, sometimes mae exaggerated claims on behalf of the analysers, and these claims should be guarded against. Perhaps Che relationship between a language and a t$ny subset of it with a stric*t;ly limited number of features should be thought of as akin to the relationship between a wall and a stone. They are recognisably composed of the same substance, but one has essential characteristics which the other totally.lacks.</Paragraph>
    <Paragraph position="7"> Before leaving the shbjsa~t of multi-dimensionality, 1 would like to touch briefly on the possibility, at some time in the future, of devising a standard method of determinine the extent of a subset of a language. The following idea could be considered, probably to be rejected, but at least it could provide a starting-point for discussion, A fiumber of features, say n, could be decided upon, the number varying according to the degree of subtlety of delineation required. An n-dimensional array with n+l columns in each dimension (fbr the n featul~es + 1 blank) would then contain an element corresponding to ewbry combination of these features. Some of these qlements would be irr~levant, since they would represent impossible combinations. The valid elements could be filled or not, according to whether tHe subset contained the combination of features which they represented.</Paragraph>
    <Paragraph position="8"> For a programmer building up the subset which his program was capable of analysing, such an array could provide both a measure of achievement and a guide to what was missing.</Paragraph>
    <Paragraph position="9"> To return now to the analogy of the puddle, the second respect in which it was incomplete Ftas that a program, as it develops, bes not advmce continuously, like water spreading, but by fits and starts, in discrebe steps. Suppose, for example, that a programmer is testing a particurar feature like, say, relative clauses, as that is the feature we have previously discussed. Sentences containing relative clauses Uave been entered repeatedly, and each time they have been rejected or analysed inco~rectly. Then at last comes the moment of triumph and relief when, for the first time, the program takes in such a sentence and analyses it c'orrectly. At that moment, the program has not merely edged forward, but it has leapt. In terms of the array postulated in the last paragraph, not just one but several elements will in all probability have been covered in one step. It will not be known for certain which elements, until more testing has been done and more sentences tried.</Paragraph>
    <Paragraph position="10"> BU~ just as it would be ludicrous to suppose that because a program can analyse one relative clause it can analyse all, i.t would be equally ludicrous not to edpect a program which can analyse one rela-t;ive clause to be able to analyse at least some others.</Paragraph>
    <Paragraph position="11"> There is a section of program common to all relative clauses, which has to work before any can be ana~ysed correctly, and once that section is wox3cing jn combination with ~anp feaeures, the liklghood is that it will work in combination with at least some others.</Paragraph>
    <Paragraph position="12"> What follows from this? Firstly, that no reliable method at present exists for the designer of a language model EU0 delineate the subset of language which his model can analyse,.</Paragraph>
    <Paragraph position="13"> Secondly, that desi&amp;quot;rab1e as it undoubtedly is, for the benefit of both the designers of models and those who seek to assess their scope, to devise such a method, it is going to be extremely difficult to do so. Thirdly, that in this unfortunate state of affairs a designer can but fall back on the established system of presenting a list of sentepces which his program has analysed correctly, and leave it to the reader to make his- own assessment of where the bounds of the subset analysable by the model lie.</Paragraph>
    <Paragraph position="14"> To discuss the sentences as simply a few examples would be unintelligent; equally unintelligent to see in them visions of universal.ity. Where, between these two extremes, the reader's judgement falls should depend upon the variety of the sent-ences, and upon their complexity. If the program can&amp;quot; deal with complexity in any area, it should be some indication of its power, perhaps yet unrealised, to do so in other areas, It would be a sign of its versatility, of its ability to disentangle elaborate pat* ern3 and resolite them into their elementary cornporls~lt s . After -khis rathw prolix introduction, 1 eorrle eventually to my own &amp;quot;Pist of ssntenc\@st1 ellat have been successftrlly annlyssd by CLAM. They fall into two categqries: those which have been tranglat~d into French, in which case the Frexic$l translation is given; and those which have sirtiply been analysed sylt actically, and semantically and reduced to a base form. Tliis 1s because durfng the last year 1 have not been working on the French generator but concentrating on certain asyects of the ana1yser;and so in order to save computer time, the French seneration has been omitted, Thus the sentenceswithout translation have been processed last.</Paragraph>
    <Paragraph position="15"> The question arises of what exactly is meant by &amp;quot;analwed syntactically and semantically&amp;quot; and &amp;quot;reduced to base form&amp;quot;. Ths will be more fully explained in the subsequent test. At this, stage a&amp;quot;t is sufficient to say that a syntactic tree has been formed and serntxntic ambiguities resolved, and that semantic relations betr3een words in the tree have been determined (e.g. a syntactic subject of a passive verb is recorded as the semantic object).</Paragraph>
    <Paragraph position="16"> Singre word meanings are retained as basic units. There is no Schankian-type resolutiun lnto semantic primitives, except insofar as this is implicit in the classification system. This is the base form from which the French has been generated. It has no% so far proved necessary to go any baser. Development, as will be explained later, is envisaged along the lines of extending the network rather than breaking down the units.</Paragraph>
  </Section>
  <Section position="5" start_page="9" end_page="19" type="metho">
    <SectionTitle>
SAMPLE SENTENCES
</SectionTitle>
    <Paragraph position="0"> The follow3.ng are samples of sentences which have been correctly anaW8ed by the program. They are given, together with the French translations where these have been produced by the program, and with comments on points .of interest in the s8rTtences.</Paragraph>
    <Paragraph position="1"> I The shirt which you sold is dirty.</Paragraph>
    <Paragraph position="2"> La chemise que vous avez vendue est sale.</Paragraph>
    <Paragraph position="3"> Relative clause.</Paragraph>
    <Paragraph position="4"> 2 , The man and woman doctors saw have eaten the bread, L'homme et la femme que les medecins on* vu ont mange le pain. Contact clause (relative clause with relative pronoun missing).</Paragraph>
    <Paragraph position="5"> Simple conjunative phrase, No article in English but article required in French.  Les instituteurs ecrivent des pieces en Mars dans des campagnes.</Paragraph>
    <Paragraph position="6"> Semantic resolution of &amp;quot;in&amp;quot;.</Paragraph>
    <Paragraph position="7"> lDes campagnest should be tcertnJrlrpays 14. He stood up to put the fire off.</Paragraph>
    <Paragraph position="8"> I1 stest leve pour eteindre le chauffage.</Paragraph>
    <Paragraph position="9"> Two-word verbs.</Paragraph>
    <Paragraph position="10"> 15. That waiter, fat and stupid, was breaking the plates. Ce serv'eur gros et stupide cassait les assiettes, Appositional adjectives between commas.</Paragraph>
    <Paragraph position="11"> Continuous tense.</Paragraph>
    <Paragraph position="12">  La femme avec qui vois avez nage est contente. Floating preposition at end of relative clause. 19. The woman looks depressed and bored.</Paragraph>
    <Paragraph position="13"> La femme a 11 air 0MUye apd drprime.</Paragraph>
    <Paragraph position="14"> Semantic resolution of &amp;quot;looks&amp;quot; lknuye' and ldeprirno should be f ominine. My ignorarlce. 20 .The Queens should have arrived.</Paragraph>
    <Paragraph position="15"> Les reines auraient dl1 arrivsr.</Paragraph>
    <Paragraph position="16"> tfShould have&amp;quot; -- difficult construction, 2 1 I. had to learn to shout.</Paragraph>
    <Paragraph position="17"> J1ai du apprendre a crier.</Paragraph>
    <Paragraph position="18"> Semantic resolution of &amp;quot;had&amp;quot;, 22. Yaur brother, you and found and your father bought her horses.</Paragraph>
    <Paragraph position="19"> Votre frere, vous et moi avons trouve et votre oere a achete ses chevaux, Mixed con ju-&amp;ion.</Paragraph>
    <Paragraph position="20"> 23, If you had come you would have met him.</Paragraph>
    <Paragraph position="21"> Vous llauriez recontre si vous etiez Venus. Conditional clause.</Paragraph>
    <Paragraph position="22"> Compound tenses.</Paragraph>
    <Paragraph position="23"> Pronoun object.</Paragraph>
    <Paragraph position="24"> Concord of past participle after &amp;quot;etre&amp;quot;. 24, Picking flowers is wrong, Cueillir des fleurs est mauvais, Gerund subject.</Paragraph>
    <Paragraph position="25"> 25. The king is as large as a cow.</Paragraph>
    <Paragraph position="26"> Le roi est aussi grand qu'une vache, &amp;quot;AsN comparative.</Paragraph>
    <Paragraph position="27">  I have never behaved rudely since you allowed me to stay. He prefers painting pictures to working. As many as six aeroplanes took off, Men bought the book and cloclc. They mended it. It often did work.</Paragraph>
    <Paragraph position="28"> Pronoun resolution.</Paragraph>
    <Paragraph position="29"> How good a game is tennis.</Paragraph>
    <Paragraph position="30"> I know which house the man w8S l$ving in, I know how easily embarrassed you are. How clean a brush did you sweep the room with. Men can understand which book is best.</Paragraph>
  </Section>
  <Section position="6" start_page="19" end_page="19" type="metho">
    <SectionTitle>
DESCRIPTION OF METHOD
</SectionTitle>
    <Paragraph position="0"> Before going into some detail about the method used to achieve these results, I would like to say something about the danger of over-sophistication on the part of the reader, There is a natural tendency for researchers, on reading something new, to look for points of broad similarity with something, anything, %*hat they have read, before; and, having found it, to sit back with relief and feel absolved -from reading any further. In a field in which vast amounts are being written, it is a proper self-defence on the part of the reader, but in A.1. in particular, it has its special dangers.</Paragraph>
    <Paragraph position="1"> When one passes from the realm of pure ideas to the- ,hard practicalities of writing a computer program, a subtle change of emphasis occurs, The ideas, all embracing they may have seemed at their inception, recede into the background, and what becoma vital are the details, the tiny mosaic pieces which determine whether the program succeeds. To Judge a computer program by a crude classification ofi its metl~od is like judging a picture by saying that it is impressionistic.</Paragraph>
    <Paragraph position="2"> Certainly it ;is impressionistic, but is it any good? In case the reader is not convinced by t'his argument, let me say immediately that this is a multiple-path, singlepass, left-to-right, word-by-word analyser, akin to the multiple-path analyser of Oettinger and the augmented transitiofi~networ grammar of Woods. In order to tackle the semantics, and indeed also the syntactics, the meanings of words have been coded according to a hierarchical taxonomy That they are coded has been largely dictated by the demands of FORTRAN, in which the program is written, although some system more overtly !Like a networ could have been used. That the classification should be essentially hierarchical, with certain necesgary refinements, has always seemed obvious.</Paragraph>
    <Paragraph position="3"> Some details of how the program works now follow. I will start wikh the syntactic and semantic analysis, and come later to the generation of the French.</Paragraph>
    <Paragraph position="4"> find codes</Paragraph>
  </Section>
  <Section position="7" start_page="19" end_page="19" type="metho">
    <SectionTitle>
FLOWCHART
</SectionTitle>
    <Paragraph position="0"> syntactic and semantic analysis French</Paragraph>
  </Section>
  <Section position="8" start_page="19" end_page="19" type="metho">
    <SectionTitle>
REDUCTION OF THE PASSAGE TO BASE FORM
</SectionTitle>
    <Paragraph position="0"> A glance at the flowchart on the preceding page shows that there are two main parts of the program: first, the single subroutine READ, and second, a,group of subroutines comprising th~ syntactic and semantic analysis.</Paragraph>
    <Paragraph position="1"> READ This subroutine first transgers the base form of  -the ppeceding sentence to semi-permanent store. Then ~t reads the next sentence. It looks up each word in the dictionary file (VOCAB). If it cannot find it at first, it tests for-certain endings such as -s, -ed, and -ing, subtracts them and tries again. When it finds the word it stores all the possible codes which are associated with the word in VOCAB. It also assembles compounds such as 'in front of1, 'in order tot, or infinitives, for which there is no single code, Proverbs or cliches can be 'similarly treated.</Paragraph>
    <Paragraph position="2"> Coding Every possible rheaning of a word has a code number containing a maximum of twelve digits. These code numbergs are stored with the word in VOCAB and extracted in READ, The coding is based on a straightforward classification. 2 1 1 1 For example the code of 'bull' is</Paragraph>
  </Section>
  <Section position="9" start_page="19" end_page="42" type="metho">
    <SectionTitle>
6 2 I
</SectionTitle>
    <Paragraph position="0"> noun concrete creature male Such classification is essential to reduce animal farm cow the number of syntactic and semantic patterns which have to be B stored. It may be noted in passing that the system of coding contains the elements of both syntactic and semantic classification. The distinction between the two is at tlmes tenuous. Further explanation of the coding is given in the appendix. Syntactic and Semantic Analysis This is Ohe most complicated part of the model, and comprises several subroutines. For ease of explanation, many of them are here treated as parts of the larger routines CON, UPDT and OACR. As the flowchart shows, these three routines operate in turn on each word of' a sentence, and when the end of the sentence is reached, a fourth routine, ENDR, is called on to operate on the complete sentence .</Paragraph>
    <Paragraph position="1"> Before giving some account of the functions of these routines, it is necessary to explain the term EP, and to describe JEP and JSP, the two principle files referred to in this part of the program.</Paragraph>
    <Paragraph position="2"> EP (~n~lish pattern). Take the sentence, The man with a  long nose always snores. The program breaks this down into four EPs, as follows: EP 1 EP 2 EP 3 EP 4 lead word snores man with nose subsidiary word man the nose a subsidiary word always with long An EP contains one lead word plus a number of subsidiaries, and is classified according to the nature of the lead word. Thus W 1 is a verb EP, EPs 2 and 4 are noun EPs, and EP 3 is a preposition EP. Man, the lead word of EP2, is a subsidiary of EP 1, so EP 2 is dependent on EP 1. Similarly, EP 3 is depenaent on EP 2, and EP 4 on EP 3.</Paragraph>
    <Paragraph position="3"> Splitting a sentence into EPs is simply forming it into a tree structure.</Paragraph>
    <Paragraph position="4"> snores /\ a long JEP This file contains the templates for all the different  -t;ypes of EP. FOT example the template for a noun EP contains various types of adjective in appropriate sequence, These are followed by the lead novn. This is then followed by adjectives, appositional nouns, prepositions and relative pronouns. In an EP certain positions, such as the lead, are necessary, while others are optional. In a noun EP, the only necessary position is the lead noun. In a preposition EP, besides the lead preposition, the following noun is necessary. By far the most compliCated EP is of course the verb EP. In this EP, later positions can be either closed, or opened, or made necessary, by a particular class of word in a particular posi'tibn. For example, a pre-verb subject closes a fiost-verb subject. A question verb makes a post-verb subject necessary. One class of verbs opens a subsequent gerund position and closes a subsequent infinitive, Once an EP has bean started, the program tests to see if the next word could occupy an open posi$ion on the template as far as the next necessary position (cf. below). Note that these templates are of syntacfic patterns and bear no relation to Wilkst semantic templates. JSP This file contains all the semantic patterns (SPS).  An example of an SP is 12119 21 21102. This means that all verbs whose codes start with the digits 12119 can have as subjects any nouns whose codes start with the digits 21102. More specifically, it means that human beings read1 or write. In this case the verb would be the lead of a verb EP, and the noun would be a subsidiary word in the subject position of the EP. The middle group of digits in the SP specify the relationship between subsidiary and the lead. In this case, 21 specifies subject of verb. Similarly, SPs govern the relationships between the lead verb and all other subsidiary positions in the verb EP, and between lea9 and subsidiaries of all the other EPs. For example, 621 2 226 means that time prepositions, whose code words start with the digits 621, can have as objects any time nouns whose codes start with digits 226. When deciding whether a word is acceptable in a subsidiary position of a particular EP, a,semantic match is made between that word and the lead word: JSP is searched to see if an SP exists permitting that word to be associated in that subsidiary position with that lead wofd (cf. below).</Paragraph>
    <Paragraph position="5"> Processing the Sentence The sentence is processed in a single pass word by word from left to right. After each word, a number of possible continuation paths are open. The next word is tested along each of these paths, and if no place can be found for it that path is closed. If no placescan be found, the path is reproduced n-1 times and the word added to each path Each path may then have one or more continuation^, Let us now return to th8 sentence, The man with a long nose always snores. The program goe's through the sentqnce word by word, starting from the first. At the beginning, a verb EP is llapentl, Thak is to say, the program loodcs for all posltiqns which could start a verb EP which the first word satisfies. In this case, 'the1 cannot be part of a verb El?, but only of a noun EP, so the program will skart a noun LP which is dependent on the subject position of a verb EP. The next word 11iust continua the noun EP. Therefore on going to the next word, only EP2 is &amp;quot;open&amp;quot;. ,'Many is then read, and EP2 and also EP1 &lt;are updated. At this point there are two alternative continuations. Either EP2 could be continued, as in fact happens, or EP2 could be closed&amp;quot; and EP1 continued. Therefore on going to the next word, EPs 1 aqd 2 are both open. So the process is continued through the sentence.</Paragraph>
    <Paragraph position="6"> As shown in the flowchart, there are three subroutines which operate on each word--CON, UPIX and OACR.</Paragraph>
    <Paragraph position="7"> CON takes each EP which is open, and tests each sense of the  word against each possible continuation of the EP. If the word could satisfy a position, it then looks to see whether a form match is necessary. In general, in English, a form test is only necessary between subject and verb, when the number and person must agree. If this hurdle is overcome, CON then proceeds to a semantic match. In general, the lead word of an EP must be matched semantically with every subsidiary word of that EP. For example a sub&amp;quot;jct must' be matched with a verb. So i% check is performed, to see if that particular noun taken in that particular sense could be the subject of that particular verb taken in that particular sense. Having found all the possible solutions, CON then gives way to UPDT.</Paragraph>
    <Paragraph position="8"> UPDT updates each EP according to the solutions found in CII_ CON. It reproduces EPs as necessary where more than one solution has Been found, and discards EPs which have becorne defunct because no solution has been found. It also determines which lqter positions of' an EP either can or must be filled as a result of the current word becoming a part of the EP. Tt Lllen hands over to OACR.</Paragraph>
    <Paragraph position="9"> OACR (open and Close ~outine) determines which EPs must  be kept for the next word. It also perforrns some juggling with EPs in certain rather tricky cases such as relative clauses. It then returns control to the root PTOgXaIn for the next word.</Paragraph>
    <Paragraph position="10"> When all the words of the sentence have been processed, ENDR is entered. This examines all existing solutions. It discards any that are incomplete, and perfornls sorne housekeeping on those which: are complete in order to separate them, In future, it will make a choice between alternative solutions, although this part of the program has not yet been written. After ENDR, the sentence has been reduced to one (or more) sets of connecked EPs. Within an EP, for each subsidiary word the relationship to the lead word ( eg. verb/ob ject , verb/timenoun, noun/article, etc.) is specified, as are the code(s) remaining as a result of the semantic matches which that word has undergone during the analysis.</Paragraph>
    <Paragraph position="11"> After this brief description of the functions of the various subroutilies, a Inore detailed explanation of the semmtic match follows. We thon show l~ow tho progrnrrl deals wit11 sams af the. more co~nplsx probllsr~ls ~~~'hiclr~ it silco~nt~t~rs.</Paragraph>
    <Paragraph position="12"> SerriariCic Mat ching 111 or*dsx* to illustrate tho rriotllcsd, ;I sixllplified exarnple is given, using the word i . Take the sexltence, She walked i.11 fields ixi Nay, Suppose after READ, tho following coclos aro i.11 store: The last digibs, 1 to 6, refer to the word number. 'Int is words 3 and 5 with codc nunibers 4-9, 1 1 - 16. Suppose the codes have the same meanings as shown ascribed to 1 1-16. Suppose further that &amp;quot;places11 start n with digits 2127, and that 'fieJdl is 21274, also that &amp;quot;time periods&amp;quot; start with digits 223, and that lfrnonths't start  17..22355.... 6, It may well be asked why the distinction has been made between the three place prepositions and betweer1 the two time prepositions. There could be two reasons: either that the concept of the preposition changes (which is probably not true here),.or that the tranaIation is different in spme target language. If it is only the second case, the distinctian could have been left for the ,program which generatds the target languwe to draw. However, it is more economical to deal with it @ring the semantic ntatching.</Paragraph>
    <Paragraph position="13"> Now let us see how the disambiguation procoas works.</Paragraph>
    <Paragraph position="14"> This example is simplified because it does not shaw the semantib matching acTOSS prepositions, between 'walked1 and 'fields1, and between 'walked1 and 'Mayt. Although sometimes necessary for complete disambiguation, it is not so in this example, and as i.4; complicates the explanation, I will omlt it here for the sake of simplicity, After the second word, there is only one EP open.</Paragraph>
    <Paragraph position="15"> Cbde range (??he meaning of &amp;quot;code EP 1 lead wdk 2-2 range1* will appear subject she 1-1 prgsently. ) The thipd word, in1, has two syntactic classes, ad'verb or preposition. Both are acceptable at this point in the verb EP. So a semantic match is performed between each class of 'in' and the lea4 word 'walk'.</Paragraph>
    <Paragraph position="16"> Suppose that one SP gives 114 5 52, another gives 1 1 6 62, and another gives 11 6 63.</Paragraph>
    <Paragraph position="17"> All the codes of 'in' are accepted-- code 4 by the first SP, codes 5, 6 and 7 by the second SP, and codes 8 and 9 by the third The EP has to be reproduced because there are two syntactic classes of 'in'. We therefore have the following: code range code ranges EP 1 lead walk 2-2 EP2 lead walk 2-2 2-2 subject she 1-1 subject she 1-1 1 - 1 adverb in 4- 4 preposition in 5-7 819 in EP2 there are two code ranges, one for r;he place preposition and one for the tinis prepositiong. EP3, a preposition EP attackled to EP2, is now opened, and for the next word this preposition EP and EP1 are open, but EP2 is closed, The next word, 'fields', is a place noun. It is not accepted in EP1, which is therefore disca-rded. It is accewted in EP3 as the object of the preposition, so a sernantlc match IS performed between 'in1 and 'fields', Suppose there is an SP, 62 2 2127, Codes 5-7 are then accepted by this SP, and EP3 then looks like this: code range EP 3 lead in 5-7 object&lt; field 10-10 A reconciliation is now carried out between the codes of 'inf in EP3 and EP2, As a result, the second code range in EP2 1s elirninat ed.</Paragraph>
    <Paragraph position="18"> The next word, 'in1 again, IS now read, and the process is repeated, EP2 now looks like this: lead walk 2-2 2-2 subject she 1-1 1-1 preposition in 5- 7 5-7 preposition in 12-14 15-16 This time, on 'May', the relevant SP is 6311 2 2235.</Paragraph>
    <Paragraph position="19"> There would also be an SP like this: 63 2 223.</Paragraph>
    <Paragraph position="20"> But the first SP gives a narrower code range (16-16 instead of 15-16), and so it is preferred. This time, on recgnciliation, the first code range in EP2 is eliminated and the second is reduced. So at the end, the three EPs are thus: EP2 lead walk 2-2 EP3 lead in 5-7 EP~ lead in 16-16 subject she 1-1 object fields 10-10 object May 17-17 preposition in 5-7 preposition in 16- 16 We are now left with a code range for the first 'in1 containing three codes, In such eases, it is the first code of the range which 1s selected. So 'in' has been dlsambfguated to 621 in the first case, and tg 6311 in the second.</Paragraph>
    <Paragraph position="21"> Syntactic Complexities Of course, it is all very well for a program to be able to digest, She walked in fields in May. But can it also cope with this? The farmers we were talking about grew, and the greengrocers, thieves and liars, sold those apples.</Paragraph>
    <Paragraph position="22"> In other words, the program must be capable of being expanded to deal with the myriad complexities and exceptions of natural language. However sound the principles underlying a program may be, such expansion involves a deal of intricate and detaiJ.ed work. At every stage,, flexibility and rigidity liave to be balmced. The program must be flexible enough ta e'nvisage possibilities, but rigid enough to exclude impossibilities and to latch onto the right solution when it appears. The programmer's task resembles a tailor's. Let out an inch or two mare, taka in a couple thore. It .c~.ould be satisfactory inaeed if an algorithm could be Pound both concise and comprehensive which would encompass all the requirements, but language is such a barnacled growth that this seems on the face of it improbable, It would be surpri.sing if excrescences in the program were not necessary to deal with excrescences in the language. In the development of this program, when the treatment of a new structure has been added, whenever possible the original framework has been adapted to incorporate it, thereby avoid~ng the necessity of adding large sections of program. This is only commonsensical. Nevertheless, the program has grown considerably with its c~pacity to handle largerareas of language.</Paragraph>
    <Paragraph position="23"> Here is perhaps a suitable point to emphasise that, since this is a multiple-path analyser', at sach point all the available information, syntact'ic and semantic, has been deployed to eliminate incorrect paths. This has been done not only to avoid unnecessary computatian, but also because the storage limits have made i$ essential. There are only 25 EPs. Frequently during testing this store overflowed, but interestingly enough ~t has always been possible to bring the demand on it back within bounds by finding some restriction which had been overlooked and which cut out one of the paths. It had originally been feared that 25 EPs would not be nearly enough. One of the satisfying discoveries of the program is that it'i.3. Of course the deployment of all available information is nod &amp;he only approach. Most of the earlier program concentrated on the syntax and paid little heed to the semantics. Wilks, on the other hand, is relying primarily on the semantics suzd is tbaking from the syntax only what is absolutely necessary. It will be fascinating if his research is able to deternine exackly how much of the syntax is unnecessary. There are obvious redundancies in the form of unnecessary safe-guards in language, No one who has struggled with German case endings is ignorant of this, In English, we have the concord between subject and verb in the third perSbn of the present, patently unnecessary since it exists only in this one instance, There are many sentences in which the semantics alone are clearly sufficient. In the sentence, &amp;quot;The man ate the steak with a fork,&amp;quot;, the words could appear in any sequence and the meaning would be decipherable, although it might take longer to decipher. The interesting question is what features of the syntax can be consistently ignored, without occasional sentences cropping up which can only be deciphered with the help of these features.</Paragraph>
    <Paragraph position="24"> There now follows a description of the treatment of three notoriotisly awkward problems-- relative clauses, pronouns, and  conjunction.</Paragraph>
    <Paragraph position="25"> Relative Clauses Six cases are distinguished: 1. The man who met you. 5. The man you met.</Paragraph>
    <Paragraph position="26"> . The man who(m) you met. 6. The man you gave it to.</Paragraph>
    <Paragraph position="27"> 3. The rnan who (m) you gave it to.</Paragraph>
    <Paragraph position="28"> 4, The man to whom you gave it.</Paragraph>
    <Paragraph position="29">  After the lead of a noun EP, a relative pronoun (94), a preposition (6), and a contact noun (2~) are all possible continuations.</Paragraph>
    <Paragraph position="30"> lWhol has three relative pronoun codes, starting with, 941, subject of relative clause, 942, object of relative clause, 943, object of prepositicm in relative clause,'Whom' ~bviously only has the last two.</Paragraph>
    <Paragraph position="31"> EP1 man EP2 man EP3 man EP~ --the the the (man) (subject) who 9-41 who 942 who 943</Paragraph>
    <Paragraph position="33"> When a relative pronoun is recognised, the noun EP, EP1, is reproduced to EPs 2 and 3, and the codes 941, 942, and 943 are to added to separate noun EPs. Then in OACR, new EPs 4 6 are opened dependent uppn the noun EPs. In the case of 941 and 942, the lead of the noun EP, *mant, is Bntered in the neQ EPs,as subject and objec-k respectively. They are marked so as to avoid translation, but they are necessary for semantic matching in the relative clause. In the case of 943, an additional new preposition EP, EP7, is opened dependent upon the relative clause EP, and the lead of the noun EP, 'man', is entered as the object of this prepositiatn EP. The relative clause EP is marked as waiting for a fldating preposition, although when a preposition comes this EP is reproduced, and in one EP the preposition is taken as the floating preposition, while in the other EP it is talcen as another preposition. This is necessary to allow for such clauses as, the man wliom you gave tile book in the end to.</Paragraph>
    <Paragraph position="34"> In the cases of 941, 942 and 943, the only EP which is open for the next word is the relative clause EP. For 941 the next necessary word in the EP is the lead verb, while for 942 and 943 the next necessary word is the subject. In practice, one or more of these EPs is usually eliminated on the next word.</Paragraph>
    <Paragraph position="35"> When a contact noun is recognised, it is marked in the noun EP as being in reality a relative pronoun. Then the procedure for 942 and 943 above is followed; but in addition, the contact noun is entered as the subject of the relative clause EP.</Paragraph>
    <Paragraph position="36"> When a prepusition is recognised, the noun EP is reproduced once, because the preposition might be in the noun EP, like dog in a manger, or it might be in a relative clause. For the relative clause path, a preposition EP and a relative clause EP are opened. Only the preposition EP is left open for the next word, which must be a relative pronoun, For indirect questions, I don't know which house he bought.</Paragraph>
    <Paragraph position="37"> I don't know what he lived in. etc the treatment is somewhat similar to that for relative clauses. Pronouns For either a translation or a questio~l-a~isweri~lg program, the noun which the pronoun replaces, called here the replacement noun, has to be identified. In a questionallawering program, the reasons are obvious enough. In a translation progrml, it is necsssary for sertiantic ~natchlng and also because in many target languages the gender of the prohoun varies with that of the replacement noun.</Paragraph>
    <Paragraph position="38"> The replacement noun might; be in the same sentence as the pronoun, or in a previous sentence. Therefore, in dealing with pronouns, the program must be able to refer to preceding sentences. So after ENDR, the essential information for the sentence just processed is extracted from the first chain of EPs and stored. At present, th$s is only done for one chain of EPs, i,e. one solution. This essential information consists of a tree, containing one code for each word and the relatlon of each word to the code to which it is attached. Reverting to, The mail with a long nose always snores., the information is as follows, snores....,...l.....e.1175 tense.........Z.......tense, mood, code.,.T 1 m~.. ...~..t..3e.~~~~m211021~e~~~o~~m~e1 1 ........... the 4.....~..4032..................3 with .*C.......5e.e....6~ 7e...e.e*.0***.0....3 .......... nqse 6e...m~~212ee~eem~e~~~e~~~~~~e5 a......~ .m...m7mee~~~~4~331~m~~me~~~m~~om~~~ 6 .......... long 8....... 4176.a.~...e.~e.e~*..m 6 r alwgYs ........9.......33 6e..-...........e.m.1 The fasr; co~umn polrl~s b~ tile code to which the word is attached. The previous column contains any relationship information not implicit in the code itself or, in the case of a pronoun, a pointer to the code of the replacement noun, 1% is important to notice that the code itself usually does provide the relationship information. For example 61, the first two digits of 'witht, specify with some precision the relationship of 'witht to 'man', With the preceding sentences available in this form, the processing of a pronoun works as fo1;lows. When the pronoun is first encountered for a semqntic match, all the possime replacement nouns are found; that is to say, all those nouns which agree in number and person with the pronoun and which are either before the pronoun in the same sentende but not in the same clause, or in a preceding.sentence. The program only goes back through the preceding sentences until suitable noun has been found, If for example there werB one or more suitable nouns in the second sentence before the current one,  it would not examine the third sentence besore the current one. Consider the following sentences.</Paragraph>
    <Paragraph position="39"> The man went into the shop where he had seen the raincoat.</Paragraph>
    <Paragraph position="40"> He bought a hat an4 took it away.</Paragraph>
    <Paragraph position="41"> For 'het, the only possible replacement noun is 'man' because it is the only noun which agrees in person. For it, the program firids 'hat1, 'raincoatt, and 'shopt as possible replacement nouns, If there were a preceding sentence, it would not bother to search it. Semantic matches are then carried out between 'take' and each of the three nouns and all three nouns are accepted, so they are all entered into the EP after it. But the code ranges for 'shop' are more restricted than for 'hat1 and 'raincoat1, because the physical-movement meaning of 'taker is excluded with 'shopt because 'shopt is imrnoveable. When 'away' is read and matched with 'taket, all meanings of 'take' except the physical-movement meaning are eliminated. tShopt is now left dangling, so to speak, and is eliminated as a possible replacement noun. So when the end of the sentence is reached, there are two possible swviving replacement nouns, 'hat' and 'raincoat'. There is no semantic reason for preferring one of these to the other, becaase the number of digits matched in the semantic match with 'take' is the same in both cases. Therefore in ENDR a choice is made according to a formula of priorities and 'hatt is selected, as; a rmre recent verb object.</Paragraph>
    <Paragraph position="42"> This &amp;quot;formula of prioritiesv, which is only applied if there is no semantic preference for one noun, is probably at  the moment .a rather blunt instrumeat. It is concerned with two factors -- which noun occurred in a 1-ater clause, and which noun has the same function as the pronoun; subject, objept, preposition object, or object of the same preposition. In the majority of cases it produces the correct answer, but it is possible to think up examples in which it doesn't.</Paragraph>
    <Paragraph position="43"> With experience of use, the formula will be refined.</Paragraph>
    <Paragraph position="44"> A complication is addea ~y rhe possibility that, when a subdect, 'itt may be impersonal. This sense is treated essentially as one possible replacement noun.</Paragraph>
    <Paragraph position="45"> There is still work to be done in developing the for,mula of priorities. CLAM extracts the information required to solve the pronoun problem. The question is, how to use it.</Paragraph>
    <Paragraph position="46"> Conjunction No mt of the program is more complex than that dealing with conjunct'fon. The principles are clear,Qeven simple, enough; but applying them has demanded a considerable amount of care. Consider the fragment, He cleaned the carpets in the bedroom and......</Paragraph>
    <Paragraph position="47"> When 'and1 is read, the EPs are as follows: EPI cleaned EP2 carpets EP3 in EP~ bedroom he the bedroom the carpets in All four of these EPs are alive, which is to say that the next word might be a continuation of any of them. On recognising a conjunction, the program looks for nnssible continuations in all alive EPs, from the beginning of the EP up to the point which has been reached. It carries out the necessary semantic matches, it opens a new &amp;quot;conjeptt or conjunctive EP for each solution, and it enters dummy words in both the conjep and the EPs above it in the chain where necessary. To clarify this prooedure, we will consider two possible continuations. (a) ..... and I...... '1' can only be the subject of a verb EP, so the conjep., EP5, must be joined to EP1. The program adds a 5(: entry, and opens EP5 thus: EP1 cleaned EP5----carpets EP5 is dependent on EPI at the subject position  (b) ..... and curtains. 'Curtainsf could be joined to  EP2 as the lead, or JSY~ as the object. The conjep IS attached to the lower EP, EP2, but a dummy word is entered in EP1 and the semantic match is carried out between the dummy word, 'curtains', and the lead of the EP, 'cleanedt,</Paragraph>
    <Paragraph position="49"> curtains K5 lead position, 'The1 ris entered as a dummy word in EP5 because it comes before the point at which EP5 is dependent on EP2. A semantic match is carried out be%w;een 'thev and 'curtainst.</Paragraph>
    <Paragraph position="50"> 'CurtainsT might also be the subject of a verb EP, so EPI is reproduced and another conjep started, attached to the reproduced EP at -%he subject position, as for.....and I...... aboqe. This path is unlikely to be correct, and will probably soon b~ eliminated, An attempt is also made to attach curtainst to EP~ in the lead position, but it fails because a dummy word curtainsi is then put into EP3, and the semantic match between 'int and 'curtains1 is tried +and fails, Now let us see what the EPs 1ookd.ike at the end of a more complex conjunctive sentence: I, you and Nellie saw, watched and greeted the men, women and tired children.</Paragraph>
    <Paragraph position="52"> It will be seen that control passes from the conjeps 5 and 6 up to EP1 before 'men', so that 'men' is entered as a word in EPI. But it is also entered as a dummy word in EPs 5 and 6, and semantic matches are carried out with 'watched1 and 'greeted1. Also 'woment and 'children', although only dummy words in EPI, are entered as dummy words in EPs 5 and  remains closed until the last necessary word up to the branch has been filled. If the sentence had read, 1 you and Nellie saw, and 'he1 watched and greeted.,etc. EP5 would have opened with 'he'. I, 'you1 and 'Nellie1 would not have been entered in it as durnrny WQ~~S. El? would flclvo remained open, and EP1 closed until arter tho lead w~rd twa-l;chedlm A coma is treated as a possible conjunction or as FI possible braclcst. Bocauso of the dual role of c.i corntrln, tllc programming associated with it is rather awlc~qard.</Paragraph>
    <Paragraph position="53"> To sum up the treatment of conjunction, the possible continuations from a conjunction, particularly if'there have been previous conj~mctions in the sentence, can be numerous. But by the strict use of dumrriy entries and their associated semantic matches, false continuat~ons are usually quickly nosed out and eliminated. Also, for the recordirlg of the full meaning of a corij~uictive senterice for- tho purpose of later interrogation, the dummy entry system is of course essential. And in the special case of comparative sente~~ces, it is only by such a system that it can be clearly established exactly what is being compared.</Paragraph>
    <Paragraph position="54"> Summary. I conclude this section with an assessment of what the analysis can and cannot achieve. The purpose of analysis might be described as follows: to select, from among all the possible meanings of each word in the passage, its correct meaning in the context, and to determine what semantic relationships exist between '~qhich words, CLAN can do thzs with considerable efficiency within the confines of 9 single sentence. It is just beginning to enlarge its horizens to deal with longer texts.</Paragraph>
    <Paragraph position="55"> To clarify this statement let us consider the aids which enable us to select one meaning of a word raeher than another, and see which of them CLAM applies.</Paragraph>
    <Paragraph position="56">  1. Syntactic class. Example: &amp;quot;The car will work when the mechhic finishes his wo'rlc. Here tho word 'worlc' is evidehtly a verb on the first occasion and 3 noun on the second. CLAM can usually deal easily enough with this type of ambiguity.</Paragraph>
    <Paragraph position="57"> 2. Rules for pronoun antecedents, This has already been discussed at some length.. The rules are both semantic and syntactic. When the rules are determined, CLAM will be in a position to apply th6m.</Paragraph>
    <Paragraph position="58"> 3. Semantic restrictions on syntactically associated pairs  of'words which exclude one meaning. Example: &amp;quot;He took off his g~andmother.~ Here the two word verb 'take offt mugt me'mimicl. The personal subject and the existence of an object excludes the sense of a plane taking off, 'Grandmother1 as object excludes the sense of taking off clothes, Such restrictions are the basis of CLAM'S semantic match, and ambiguities of this sort are resolved as a matter of course. 4. Semantic restrictions on syntactically associated pairs of words khich give preference to one meaning. Example: &amp;quot;I killed ths man with a gun.&amp;quot; Here, there is a synta~tic as well as a semantic ambiguity. It is less straightforward than the previous example because the ambiguous word is 'with', whlch might be ~JI instrument preposition attached to the verb kill1, or a possessicon preposition attached to the noun Vrnanl. The semantic relationships which determine the choice, however, ~nly involve 'witht indl ectly. They are between 'kill1 and 'gun1 in one case, and between 'man1 and 'gun1 in the other, NormalLy tlXle preference would be for the instrunlent; interpretation because lg~pl' is more strongly associated with 'killt as an instrument than with 'man1 as a possession. CLISrcI chooses the stronger association by taking the 'deepert semadtic match, ar 2n other words the match involving the Larger number of digits. It does this correctly, but as we shall see in a moment, it is not always correct to do so, 5, Remoter contextual environment. Sometimes the factors enabling a choice to be made are more remote from the word in question than In the examples given above, In order to find these factors, a longer journey has to be made into the environment of the word, Examples: (i) &amp;quot;The mayor hit the alderman so hard that he fell down.&amp;quot; The normal rules for selection of pronotin antecedents would prefer 'mayor' as the antecedent of 'he' because it is the subjec-tt, but in the environment of hitting, it is much more likely to be the person hit who falls down rather than tb hitter, so 'alderman' must be preferred.</Paragraph>
    <Paragraph position="59"> (ii) &amp;quot;Two men came in. One had a gun and tho other had a knife. I killed the man with a gun.&amp;quot; Here 'withT is obviously not an instrument preposition attached to lkilll, but a possession preposition attached to 'menv. This is so because the definite article 'thet attached to vmpn' implies that 'man1 has already been defined, But in fact two men have already been defined, and more information is needed to determine which of them is referred to. The only possible additional information which could satisfy this requirement is 'with a gunr, whioh does suffice to distinguish one of the previously determined men. Therefore this phrase must be attached to 'manv, At present, CLAM could not resolve either sf these ambiguities. In order to do so it would need, in the fix&amp;quot;st case, more information about the environment of 'hitt than is contained in the semantic restrictions now at its disposal, and in the second case, both a better memory and a routine for dealing with definition of nouns. Work is in progress on these vital additions. They will involve adding to the type and range of the semantic relationships between pairs of words referred to in the definition of the purpose of analysis given at the beginning of this summary. At present, CLAM only holds semantic relationships be-tween words which are syntactically related. This is not enough.</Paragraph>
    <Paragraph position="60"> Adding to the types of relationships held, and extending them to pairs of words which are syntactically remote, will greatly increase the scope of the model.</Paragraph>
  </Section>
  <Section position="10" start_page="42" end_page="42" type="metho">
    <SectionTitle>
GENERATION OF TFIE FIXENCFI TRANSLATION
</SectionTitle>
    <Paragraph position="0"> As shown in the flowchart, the sentence is operated on sequentially by four subroutines--TWEN, ITRN, FRORM and PRIN. Briefly the function of each of these subroutines is as follows.</Paragraph>
    <Paragraph position="1"> TWEN examines all the verbs. It welds them (joins auxiliaries to main verbs), and determines their tense in French. Thls is not of course necessarily the sane as in English. Other features of the sentence often have to be examined. Thus, &amp;quot;When he arrives we will meet him&amp;quot;, becomes in French, &amp;quot;When he will arrive we will meet him&amp;quot;. And &amp;quot;I have been here for five years&amp;quot; becomes &amp;quot;I am here since five years.!! Gerunds, infinitives and participles are also dealt with by TWEN, It may well be asked why the weld part of this routine is thus left until the French generation.</Paragraph>
    <Paragraph position="2"> Should it not be done avring the reduction of the English sentence to base fo'rm? The answer is that logically it should, and it will sooner or later be transferred, probably to ENDR. But at present it doesn't matter. The part of the program described in the section on pronouns which stores the base form of the last sentences is in fact performed after the French translation has been generated, and therefore, after the verbs have been welded.</Paragraph>
    <Paragraph position="3"> ITRN takes each word in the sentence in turn. I;t finds the code number in FRILE, the French dictionary file, and extracts the French word(s). Sometimes of course there is more than one, Sometimes there is zero because the English word does not have to be translated. Any particular French word may not have the same function in the sentence as the English word.</Paragraph>
    <Paragraph position="4"> In such cases, the French word entry in FRILE is followed by a code which specifies the word's function in x'glation to the English word being translated. For example, if 2.32237 isl the code for 'potato', -t;hen'FR1LE entry will be</Paragraph>
  </Section>
  <Section position="11" start_page="42" end_page="42" type="metho">
    <SectionTitle>
212237 POMME F DE 6 TERRE 6x2. The F after POMME shows that
</SectionTitle>
    <Paragraph position="0"> it is feminine. The 6 after DE shows that its x'unc$ion is as a plcepqsition in the EP of which POMME is the lead. The 6x2 after TEW shows that it is the object in the EP of which DE is the lead, Sometimes it isaecessary to go up Che tree. Fop example Y1x5 means an adverb (5) in the verb EP (1) of which the English word is a subsidiary (Y). It is thus possible to generate a French sentdnce of a radically different shape from English.</Paragraph>
    <Paragraph position="1"> ITRY also finds a French sequence code for each word.</Paragraph>
    <Paragraph position="2"> This is a code which provides the ordering of words within an EP. All lead words have the code 200. A pre-noun adjective may have a code 140, and a post-noun adjective 350. So these codes do not determine what is the actual sequence of words in the sentence, but they do provide the basic information from which the sequence is derived in FRORM.</Paragraph>
    <Paragraph position="3"> FRORM first derives the actual sequence of words in %he ,sentence. It then takes each French word and puts it into the correct rorrn. Obviously the most arduous part of this tqsk is finding the forms of the verbs. FRORM refers to tables which contain the verb endrings for both irregular bd regular verbs, and the irregular feminine and plural endings for  nouns and adjectives.</Paragraph>
    <Paragraph position="4"> PRIN prints the French translation, haying made any  necessary elisions. If there is more than one solution, it prints alterna-bive translations of particular words on subsequen.1; linss or, if appropriate, it will print complete alternative sentences.</Paragraph>
  </Section>
  <Section position="12" start_page="42" end_page="42" type="metho">
    <SectionTitle>
CONCLUSION
</SectionTitle>
    <Paragraph position="0"> Programming Details and Future Developments Programmersmay be interested in some details. The program runs on a 360-40 using 146~ of core store. The program is mitten in FORTRAN IV, not an ideal choice but the best available in the circumstances. The reduction of the English to base form requires about 6,000 instructions, and the French generation about 2,500. At present all the files are 1cep-k in core store except for the two large dictionary files VOCAB and FRILE, which are accessed on disk. It will eventually be necessary to Keep JSP also on disk.</Paragraph>
    <Paragraph position="1"> At present the processing takes about 15 seconds per word on average, of which READ takes 4076, the semantic and syntactic analysis about 20%, and the French generation 40%.</Paragraph>
    <Paragraph position="2"> No serims attempt has yet been made to optimise the program and this time could certainly be peduced. But the reduction would be offset by the eventual need to keep JSP on disk. So as a practical proposition for translating texts, it would be necessary for the processing time to be reduced by a factor of about 10. Presumably this will come sooner or later with improvement in hardware.</Paragraph>
    <Paragraph position="3"> There are certain improvements which would have -bo be made to the pogram before it could be used, apart from the extension of the vocabulary. Most obvious:  (a) there are some syntactic structures such as inversion after negatives which the program does not at present recognise; (b) a, selection routine must be incorporated in ENDR to choose between alternative solutions if more than one emerges; (c) if no solution emerges the program should try  again, selectively suppressing semantic matching, allowing words to be used outside their normal sense; (d) the sizes of some of the temporary stores would have to be increased.</Paragraph>
    <Paragraph position="4"> No particular dirficulty is anticipated with any of these developments, in that they involve no methodology fundamentally different from what has already been applied. It is primarily a matter of time and priorities. However with a fifth development, namely the extension of the memory as outlined at the end of the section oh analysis, new ground must be covered, and work on this is at present in progress.</Paragraph>
  </Section>
  <Section position="13" start_page="42" end_page="42" type="metho">
    <SectionTitle>
WASHINGTON OFFICE
STATE UNIVERSITY OF NEW YORK
SEPTEMBER 1974
AJCL EDITOR'S NOTE
</SectionTitle>
    <Paragraph position="0"> The document reproduced on these frames came to the Editor's attention recently, in spite of the publication date. AJCL thanks the State University of New ~orE, copyright holder, and Mr. Mort Grant, Director of the Research Foundation of SUNY; Ms. Winifred R. Widmer, Assistant Director for Administration; Mr. William F1. Claire, Director of the SUNY Washington Office; Mr. Jim Kalas and Ms. Susan G.,Sorrels of that Office; for their several contributions to the appearance here of RFP.</Paragraph>
    <Paragraph position="1"> Some of the names and telephone numbers must no longer be correct; but ACL and AJCL do not have the means to bring them up to date. The general understanding of the-system which this document provides is valuable to any ACL member who chooses to make use of it. Broader Federal support for the usepof computers in linguistics is probably the only way to bring the field to a much higher level of activity.</Paragraph>
    <Paragraph position="2"> Individual copies are available to all autlorized State University of New York personnel free of charge. Bulk orders, or other requests for copies can be arranged at a cost-of-print~ng charge of $1.50 per copy. All requests regarding this publication should be directed to the Washington</Paragraph>
  </Section>
  <Section position="14" start_page="42" end_page="42" type="metho">
    <SectionTitle>
INTRODUCTION
</SectionTitle>
    <Paragraph position="0"> In view of an incredsing interest in government agencies in pinpointing thclr rrscach needs, the Washington Office of State University of New York feels that it is appropn.ltc for its campuses to familiarize themselves withQ what is commonly c;dled the RFP plocrss. Thr &amp;quot;request for proposal,&amp;quot; or the governrnetlt way of saylrlg th~t it is putr~r~g out bids to fulfill a specific need, is bccoming something that all cnnlpuses sllould bc awnrc of in thri efforts to secure all possible funding for research xnd devdopment from outside sources. Fcder'd contracting out, as it were, can be a complicated process, but we hope that potcnt1.J 1nvcstig.rtors dl rcgard this information as preliminary. Aftcr that, there isn't ~riy substitute for dic .unount of work involved.</Paragraph>
    <Paragraph position="1"> Tlus booklct is dcsylcd to givc carnyur rcp~esent~~ti\lcs L~rl ovr.nr1eiv of tlic ngerlclm involved, as wen as ill1 undc~stnndin~ of '~nd cspla~l~~tlon of the RFP contract mrch;wnm. We II~VC listed 15 sepalate agncics c~irrerltly invo1vt.d m this contrdct rncchnnisrn. thelr irltcrcst ~nd appropriate addresses for more inforrliation. We have also included a glossav of tcrnls .IS Ll guide and all index of the v,xious offices and bureaus within each of the 15 agencies.</Paragraph>
    <Paragraph position="2"> The time span from inception of an idea to actnnl publication IS always a lengthy one and many people have been involved with thls report. Willi.lrn Hedbcrq. L Susan Sorrcls. Jim Kalas, ~nd Carole Combs have all participated in its dcvelopmcnt. Bryan Sw.lrtz. \vlio lvined our office in the summer of 1974 as an lntcln from the Stony Brook campus. dcscrvcs sFcclal pralsc for his coordination of many activities leading to the publlcatlo~i of the book. Witllout Ms dedication to the project, it might have bcen delc~)red considcrnbl~. We are grateful to .dl of these people for thcir assistance. Wc hope tfi.~t the book w~ll prow GI v.duable \r!nrLing glide for anyone interested in thc contrLIct system. Thc Wbwhlngton Office would bc plc~scd to provlde additional inEUorm.~tion about thc c~ge~~clcs lnvolvcd.</Paragraph>
  </Section>
  <Section position="15" start_page="42" end_page="42" type="metho">
    <SectionTitle>
I. FEDERAL PRQCUREMENT FOR RESEARCH AND DEVBLOPMENT: AN
EXPLANATION OF THE RFP CONTRACT MECHANISM
</SectionTitle>
    <Paragraph position="0"> Each year the federal government contracts for billions of dollars of work to support efforts deemed to be in the national interest. A significant percentage of the contract services are in the form of Research and ~evelopment (R &amp; D) or programmatic work which colleges and universities are particularly wcll-suited to perform.</Paragraph>
    <Paragraph position="1"> The government commits these funds in eithet of two WAYS: grants or contracts.</Paragraph>
    <Paragraph position="2"> University researchers are generally more familiar with the grant procedure than with the contract procedure. Under a grant program, a given federal agency is authorized to grant funds to non-profit institutions, frequently educational institutions, for the purpose of supporting research or a in a given general area. A body of general conditions are established by the Congress and refined by the applicable agency to set parameters for the pro$am as a whole. A specific grant for a program can be made so long as it fits within the gevral stpndards (the Guidelines) of the program and meets whatever qualitative standards for review that have been established.</Paragraph>
    <Paragraph position="3"> Since grant funds, particularly in research areas, have been inc~asingly competitive in recent years, the university community ' is turning its attention to potential support through the contract mechanismi. A contract, unlike a grant, is generally awarded for the purpose of meeting a specific reqwrement that a federal agency has determined to be important to the achievement of .the agency:s responsibilities. The contract will tend, therefore, to be specific in terms of the agency's needs, and in the case of research, wdl usually be for the purpose of &amp;quot;applied&amp;quot; rather than &amp;quot;basic&amp;quot; research. The contract, also unlike the grant, can be negotiated with any organization that is legally constituted to do business w~th the government, whether non-profit or profit-making.</Paragraph>
    <Paragraph position="4"> Government procurement m its present massive scope is a relatively new phenomenon, having emerged fin as prpscnt form in the past quarter of a century. Each of the many federal agencies involved in procurement - and most are - developed a set of standards'and procedures to administer the contracting process. While the procedures used by any gven agency tend to be relativel) orderly in terms of that agency's requirements, they will not necessarily conform to the practices of anoJler agency. Looking at the federal government as a whole, however, the set of. practices is varied, highly technical and often confusing. The general descriptions offered in 'this statement, therefore, are subject to the differing qualifications and refinements established by individd government agencies.</Paragraph>
    <Paragraph position="5"> In recent years there has been an attempt to bring mter-agency consistency to the  procurement' process. A speclal government task force has recommended legislation -which would bring uniformity in procurement policy. Congress is considering such legislation, which will be discussed later.</Paragraph>
    <Paragraph position="6"> Basic contractual agreements with the federal government are developed in one of four ways: A. Sole Sotrrce I'roclrrement. The government negotiates with one and only one provider on the grounds that the product or service is uniqucly provided by the organization in question.</Paragraph>
    <Paragraph position="7"> B. Sitlqlc Solrrco Procir r(wlolt Thc governmcot ncgotlatcs with onc and onl).</Paragraph>
    <Paragraph position="8"> onc profidel in ir~stances whcrc tllcrc could bc cnmpctlttve Lids but becnusc of f'ictors such .IS locnnon, co~lvenienc~ or spoc1.d urgency competitive bidding 16 unwarranted.</Paragraph>
    <Paragraph position="9"> C. Cutrlpc&gt;titi~~e Nrgothltiotr In instances where providers are few and known to the government, cornpetltion is litnited to tllc known promders and publicity reg,uding thr availability of .I contract is not rcqulrcd.</Paragraph>
    <Paragraph position="10"> 1). O~CVP CIotrip~tlttoil. CorltrI~cts .ue lct by open bid, utili~in~ the mmdatory  provisions for ndvcrtts~ng publlcly the availability of the contract, and ope,, solicitation of proposals AS dcscrlbcd Intcr.</Paragraph>
    <Paragraph position="11"> In dddition to the four types of contractual processes, there are special types of contractual arrangements for special purposes. For university representatives, the Ope11 and No Cost Contracts 'Ire of particular interest. These contracts permit the exchange of equipment, p~~rticularly scientific equipment, and documents for specified purposes wlthout actual fiscal trmiaction.</Paragraph>
    <Paragraph position="12"> of the various contractual patterns, the Open Competition is becoming the most common contract mcchanhm. Since the government purchases goods and servlces on a dady bass and in amounts totaling blllions of dollars annually, the problem for potential investigators is how to d~scover th~t a contract of interest wdl be let and whether thew organizations have sufficiently reasonable chances of winning the contract to justify the extensive and uncompensatablc cffort required to apply.</Paragraph>
    <Paragraph position="13"> Each competitive contract let by the government must be announced as open to cornpetltion. The announcement i$ in the form of a Request for Proposal (RFP). The announcement must be published in the journal that has begome the pnrnary source for all contract information the Conirnerce B~cslness Dally. The Commerce Business Daily is pb-Lshed In Chicago as a senrice of the Department of Commerce. (It may be ordered from the Government Pnntlng Office, Washingtpn, D.C.; the subscription rate is currently 563.50 per year.) The Commerce ~~rs~~zoss Dally has a small staff a~~d consequently must publish RFP announcements. as they arc written by the contributing.agency. As a result, information that is published regarmng a gven competition is often scant and late, with proposal deadlines somctlrnes only a matter of days away from the dare of *jpblication. ~raquently a notice d indicate the need to send for more detailed information promded by the agency offering the contract. Writing for the necessary informat~on cuts Into the short tiine available fir preparation of a proposal to meet a deadl~ne date.</Paragraph>
    <Paragraph position="14"> Alternately, most agencies provldc an additional advantage to prospective bldders in the form of &amp;quot;Bidders Lists.&amp;quot; All current prospective bidders on a &amp;quot;Bldders List&amp;quot; are notified of a pen&amp;ng RFP, generally at the same time the Conzmerce Business Daily is notified, thus gvlng the bidder on the list the advantage of several days.</Paragraph>
    <Paragraph position="15"> In order to be included on a &amp;quot;bidders list,&amp;quot; an investigator must contact the agency wth which he intends to do business and request an %pplicatim.</Paragraph>
    <Paragraph position="16"> Most agencies now use Standard Form 129*for placegent on bidders lists.</Paragraph>
    <Paragraph position="17"> While there is some 'inter-agency crossllstlng of bldders, for the most part appl~cants must expect to' submit Standard Form 129 separately to each agency for placement on a biddcrs list.</Paragraph>
    <Paragraph position="18"> There are some difficult problems in the mechanics of the contract process o$ which interested college and university representatives should be aware. For example, government agencles,have established the rule that each organizot~on on a &amp;quot;bidders list&amp;quot; must respond either positively 01 negatively to every RFP it rccelvcs. 1f an organization falls to respond to an RFP with at least a notice of- intent, it is removcd from the &amp;quot;bidders list.&amp;quot; Maintaining status on a &amp;quot;bidders list&amp;quot; therefore can become a tedious adrnln~strative task justrfiablc only to those organizations that lare geared to a heavy volilmc of government contracting. The &amp;quot;bidders list&amp;quot; does permlt the applicant organization to indicate the types of proposals in which it 1s interested and its capacities to perform tasks, however, so that RFP's are sent to the Grganlzation only when the RFP 1s relevant to the organizations' capabilities. Some of the administrative burden on prospective contractors 1s eliminated by this procedure. Even hi the best of cacumstances,'however, bidders frequently have as llttle as two weeks to prepare a proposal. Gcenerally the competltlon has the same tlme constraints,, but the process does require staff who are available and who can &amp;quot;drop everything&amp;quot; in order' to meet a proposal deadline. Pnvate research corporations and consulting firms can administer their wok so that they may respond to proposal preparation on short notice; respondmg quickly to an RFP 1s frequently 'more difficult for university investigators with heavy comkitrnents to teaching and other university-relyed activities.</Paragraph>
    <Paragraph position="19"> Except for -the time constrants under wlilch one must work, ihe preparation of contract ptoposals differs little ,from the preparation of grant proposals. In one sense, the preparation of a contract proposal IS easier\,because specific terms and objectives are generally stated in the WP. The proposal is frequently a statement of the methodology and procedures for acbiewy! prescnbed objectives. On the other hand, the contract leaves little latitude for the creative investigator to detennme'the character of the rtisearcb to his own standards. The agency sponsoring the RFP sometimes employs qualified persollb who can state a valid set of object&amp;ves and specifications for the contract; sometimes, however, it does not. It might be .necessary for he response to ~IJ RFP to contain arguments for dtenng objectives and specifications that diplo~atically demonstrate how the governments' Interests d be served.</Paragraph>
    <Paragraph position="20"> Contracting also can take place by mean6 of an &amp;quot;Unsolicited Proposal.&amp;quot; The investigator sul~mits a proposal which hk has initiated to an appropnate federal agency and the agency can respond at its discretion it the proposal meets agency needs.</Paragraph>
    <Paragraph position="21"> Generally the unsolicited proposal rises as a result of prim informal discussions between the investigator and the government agent.</Paragraph>
    <Paragraph position="22"> When a contract has been awarded by a gart~culnr agency, thc investigator's fiscal agent must meet with thc agency's contract officer to negotiate the final terms of tbc contract. Thc type of negotiatiotl wll depend upon whether the contlact is i;trr~l-l'riccb or Cuct~ei?rrbursable. A Fixed-Pflce contrLct is one in wlllcll thc olganizdtion and the agency agree upon contract terms for a re-determined fixed cost. Once the Fixed-Price contract is signed, thc organization must rncct thc terms of the contract for the prlce regardless of circurnstdnccs. The fixed-price con tract is generally more appro priiltk fbr products than for services. The Cost-Reimb~mable contract is ncgoti~tcd on the basis of cost ertim~tcs of monthly operating levels. The tnontlily bperdting level will bc the sum of .dlbwn6le costs on a monthly basls during thc tirne of the contract.</Paragraph>
    <Paragraph position="23"> ~t is understood that thc Rnal prlce of .l Cost-~cimbursablr contract can go up (Cost Ovcrrun) or down from tfic original' cstimatc as a function of cdlowablc costs actucdly incurred.</Paragraph>
    <Paragraph position="24"> Thc ~ixcd-price contract is easier to administer and simplifies record-keepmg.</Paragraph>
    <Paragraph position="25"> The reason for this is that once the pncc IS set the government is relgtlvely unconcerned about how contract expenditures break down by category or line items. With the Cost-Relmbursable contract, completc documcntatlon by cost category must be strictly maintained sincc the documcntaaon is the sole basis for relrnbursement for expehditures under the coktract The adrnlnistration of the Cost-Reimbursable contract is much more tedious, therefore, than is the case !or the Fixed-Price contract, but doos have the advantage, given adequate book-keeping, of entallins no risk to d~e contractor.</Paragraph>
    <Paragraph position="26"> Bccause the tot~l cost of a cost-~cimbursablc contr'1c.t is an estirndte rather than a firm figure, the &lt;~w.irdln~ of n colnpctltlvc bid can be based on f'lctors othcr tllan the &amp;quot;botto~n line&amp;quot; cost. Clearly the level of operational cost wlll be a significa~\t factor in determination of an award, but in tlie absence of strict FixcdCost. the way is opened for numcrou~ judgment factors to affect the decision of c~ward. while the government has tried to establish safeguards agrunst4 arbitrariness and favoritism, the latitude in decision-making is still wde in open competition contracting and even wlder in the other contractural forms.</Paragraph>
  </Section>
  <Section position="16" start_page="42" end_page="42" type="metho">
    <SectionTitle>
WHAT CONGRESS IS DOING ABOUT THE PROCUREMENT PROCESS
</SectionTitle>
    <Paragraph position="0"> What has happened over the years is that federal procurement has become a big business. Contracts for goods, services (which include research and development), and facilities have reached $60 billion annually. Although no statistics are available to describe the amount of federal contracts awaded colleges and universities, we do know that college and university participation in federal research and development efforts represents approximately 14% of the total federal Investment in R &amp; D, and that an increasing amount of that money is awarded by contract.</Paragraph>
    <Paragraph position="1"> What is dist urblng is that despite thc magnitude of federal cxpcnditures contract, the govern~lieilt has no angle office responsible for policies to govern the procurement process, and no single source to provlde guidance and inforrnatlon about federal procurement laws and regulations. In fact, the fcderal government has no clear-cut definition for the terms grant and contract.</Paragraph>
    <Paragraph position="2"> As a result, we have been faced wlth confusing and complicated rules, excessive paperwork, favoritism and abuses in procurement practices, cost overruns, and many other factors that led to a thorough study of procurement practices by the Commission on Govcrnment Procurement.</Paragraph>
    <Paragraph position="3"> The results of the Commission's work is now demonstrated in the Congress. Using the Commission's report and recommendations as a bluepnnt, Congrcss is considering 1cgisl;ltion to address procurement problems, including two major bllls of interest to State University of New York.</Paragraph>
    <Paragraph position="4"> One measure, certain to brlng about fundamental changcs a necessary to irnprovc the system, would establish a central procurerncnt policy office in the Erecutivc office, which would provide overall guidance and direction for federal procurcment policy Thc office would establish a government-wde system of uniform procurcment regulations, coordinating policies* and regulations of federal agencies, and exercising leadership in monitoring and revising future policies and procedures. The Adml~listrator would be appointed by the President and approved by the Senate. dl major decisions by thc office would be approved by Congress.</Paragraph>
    <Paragraph position="5"> At this writing, a House-Senate conference committee is considering the measure.</Paragraph>
    <Paragraph position="6"> Sources inkcate that the blll should be agreed to in the near future and wJ1 bc scnt to thc President for his signature.</Paragraph>
    <Paragraph position="7"> The other major procurement blll (S 3514, HR 9060) now under consideration by Congressional committees hopefully would eliminate the proscnt confusion among fcdcral age~~cic in deterrninillg whether the legal Instrument used to support Lhc desired goods or servlces should be a @wt or contract, or cooperative agreement, simply by defining the terms clearly. Undcr both House and Senate versions of the bdl, the contracts would be used whenever the principal purpose of the agreement would be to purchase goods or sewces &amp;quot;for the direct benefit or use of the federal government&amp;quot; for &amp;quot;ultimatc ~ublic use.&amp;quot; Grants would be used as the legal instrument when the objective would be &amp;quot;to accomplish a public purpose authorized by federal statute&amp;quot; rather than for dlrect use by the federal government.</Paragraph>
    <Paragraph position="8"> Cooperative agreement would be used to apply whenever the acnvlty would be authorized by federal statute but &amp;quot;substantial involvement&amp;quot; of the federal agency is also necessary.</Paragraph>
    <Paragraph position="9"> At this writing, the bill is under consideration by House and Scnate Government Operations Committees.</Paragraph>
    <Paragraph position="10"> Since the legslation would directly affect agreements for research and development by State University of New York and federal agellcies, a is recommended that interested persons remain aware of the measure's status.</Paragraph>
  </Section>
  <Section position="17" start_page="42" end_page="63" type="metho">
    <SectionTitle>
11. A&amp;PHABETICAL LISTING OF FEDERAL AGENCIES THAT CONTRACT FOR
RESEARCH AND DEVELOPMENT
DEPARTMENT OF AGIRICULTURE
~gricultural Research Service
</SectionTitle>
    <Paragraph position="0"> Principal Interests. Physical, biological, chemical and engineering research; includes the 21,000 acre Agricultural ~esearch Center, ~cltsville, ~ar~lbnd. Clothing and housing research, household economics, and human nutrition research. Ncw and improved industrial and food products and processing mcthods for agricultural commodities. Soil and water, crop, animal husbandry, entornology and agricultural cngineering research.</Paragraph>
  </Section>
  <Section position="18" start_page="63" end_page="63" type="metho">
    <SectionTitle>
11. ALPHABETICAL LISTING OF FEDERAlL AGENCIES THAT CONTRACT FOR
RESEARCH AND DEVELOPMENT
DEPARTMENT OF AGRICULTURE
~~ricultural ~esearch Service
</SectionTitle>
    <Paragraph position="0"> hncipal Interests: Physical, biologcal, chemical and engineering research; includes the 11,000 acre Agricultural ~csearch Center, ~eltsvdle, Maryland. clothing and housing research, household economics, and human nutrition research. New and improved industrial and food products and  @procpssing methods for agrlculturd commodities. Soil and water, crop, animal husbandry, entomology and agricultural cngineerlng research.</Paragraph>
    <Paragraph position="1">  vides support for institutions engaged in comprehensive mnrinc research, education, and ad visory service programs, supports mdividual projects in marine research and development, and sponsors education of ocean scientists and ehgineers, marine technicians, and other specialists at selected colleges and universitics.  anization. Four major areas: life, mathematical, enginrenng, and physical sciences. Address: Exec. Dir: W. J. Price</Paragraph>
  </Section>
  <Section position="19" start_page="63" end_page="63" type="metho">
    <SectionTitle>
DEPARTMENT OF THE NAVY
OFFICE OF NAVAL RESEARCH
</SectionTitle>
    <Paragraph position="0"> ~avd Research Navd Analysis Interests: Acor .ticsi geography, geophysics, metallurgy, chemistry, including physical, organic, inorganic and colloid and propulsion chemistry, solid state physics, nuclear physic's, electronics, mathematics, logistics and mathematical statistics, fluid dynamics, physiology, biochemistry, microbiology, medicine, and dentistry, biology, group psychology, phydologcal psychology, enpneering psychology and personncl training.</Paragraph>
    <Paragraph position="1"> Naval ~pplications Interests: Research and exploratory development in the field of &amp;craft and missile systems, air warfare, surface vessel problems, amphibious warfare systems, underwater wea-Interests: Preliminary investigations of systems and warfare problems; directs and monitors systems and warfare analysis studies undertaken by the Naval warfare Research Center and selectively sponsored under contract with ifidustry, research institutes, and universitics.</Paragraph>
    <Paragraph position="2">  pons, underwater communications and navigation, noise reduction in naval vessels, and advanced submarine problems.</Paragraph>
  </Section>
  <Section position="20" start_page="63" end_page="63" type="metho">
    <SectionTitle>
ENVIRONMENTAL PROTECTION AGENCY
</SectionTitle>
    <Paragraph position="0"> Itlterests: The Agency generates iesearch and development needs in the ,areas of air and water pollution control and abatement, pesticldes research and' control, solid wastc management, noise pollution, and radiation monitoring and control. In all srcas of EPA concern, there IS need for fLlther research into the effects of pollutants on the environment and</Paragraph>
  </Section>
  <Section position="21" start_page="63" end_page="63" type="metho">
    <SectionTitle>
DEPARTNIENT OF HEALTH, EDUCATION, AND WELFARE
NATIONAL INSTITUTES OF HEALTH
</SectionTitle>
    <Paragraph position="0"> Interests. Conducts laboratory and clinical tesearch on the prevention,, diagnosis, and treatment of diseases and disabil~ty. Supports memcal research in the nation's nonprofit biomedical research institutions. Generallythe Institutes conduct and support research and the Divlslons provide administration and service. They are listed below.</Paragraph>
    <Paragraph position="1"> Address.</Paragraph>
    <Section position="1" start_page="63" end_page="63" type="sub_section">
      <SectionTitle>
Diseases
</SectionTitle>
      <Paragraph position="0"> Interests: Research and development on infectious diseases and immunological phenomena of man with special emphasis on vaccines and other immunizing agents, immunosuppressive drugs, drugs for prevention and treatment of virus diseases, test systems and reagents for identifyylng tissue an tigens and diseaseproducing agents. Current emphasis is on the development of vaccines for vlral and bacterial diseases, antiviral substances and tissue typing for human organ transplanta*n. null</Paragraph>
    </Section>
    <Section position="2" start_page="63" end_page="63" type="sub_section">
      <SectionTitle>
Human Development
</SectionTitle>
      <Paragraph position="0"> Interests. Research and development progams in areas relating to maternal health, chdd health, human growth and development, maturation and the aging process, ~eproductive and perinatal biology, mental retardation and population problems. null Address: Illrector, National Institute of Child Health</Paragraph>
    </Section>
    <Section position="3" start_page="63" end_page="63" type="sub_section">
      <SectionTitle>
Sciences
</SectionTitle>
      <Paragraph position="0"> In tercsts. 1. Research, development, and evaluation of rap~d, rcliablc automated systems and instrliments, for potential application in all aspects of clinical chcrnistry, toxicology, hematology, microbiology, virology, blood banking, etc. Subject axeas of interest include sample collection and labelling techniques, new or improved analytical methods, data handllng and reduction techniques for compact computers, miniaturized and portable test sfsterds for emergency use, all intended to increase reliability, throughout, and clinical signlficance. null 2. ~esearch, development, and evaluation in all aspects of therapeutic drug use, includng synthesis, testing, assays in body fluids, and surveillance for effectiveness, side effects, and drug mteractions. The principal alm is to promote safer and more effectlve use of drugs.</Paragraph>
      <Paragraph position="1"> Related problcrns include dose-respo;se patterns, kinetics of uptake, distribution, and ehmination, metaboiic transformations of administered drugs, and quantitative analytical methods and instruments for identificatiori and assay.</Paragraph>
      <Paragraph position="2"> 3. Research, development, and production in areas where technological constrants impede progress in genetics research. Representative problem areas include isolation, synthesis, separayions, purification, and production procedures for material and synthetic gen,e,tic material: such as nuclear acids, synthetic genetic materials such as nuclear acids, related enzymes, tlssue culture cells, gene tically determined animals, etc.</Paragraph>
      <Paragraph position="3"> Address.</Paragraph>
      <Paragraph position="4"> Ihrector, National Institute df General ~edical</Paragraph>
    </Section>
    <Section position="4" start_page="63" end_page="63" type="sub_section">
      <SectionTitle>
and Stroke
</SectionTitle>
      <Paragraph position="0"> Interests: 1. Research on diseases of the nervous and sensory system with emphasis upon: cerebrovascular disorders, epllepsy and related convulsive disorders, head and spinal cord trauma, infections of the nervous system, debilitating diseases of viral etiology, braln death, and cerebral palsy and other neurological disorders of Infancy and early life.  2. Development and experimental application of new or advancing technologies fnr the study of the nemous and sensory system ana related disorders, 3. utilization of unique or unusual services to support research programs, such as.  computer programming and data processing; animal housing 'and #maintenance; preparation of tissue cultures, ma1 antigens, and other research products; and the provision of support for highlL. complex field research.</Paragraph>
      <Paragraph position="1">  Interests. The Institute has three' principal scientific program areas, which accomplish a significant part of their research programs through collaborativc research, under contract with profit-n1akmg, nonprofit, .md educational institutions as follows:</Paragraph>
    </Section>
  </Section>
  <Section position="22" start_page="63" end_page="63" type="metho">
    <SectionTitle>
1. Etiology Progr(s~ Atea - Plans
</SectionTitle>
    <Paragraph position="0"> and directs a progran~ of laboratory, field and demographic research on the etiology and natural llistory of cancer; evaluates enwonmental carcinogenic hazards, mechanisms of cancer induction, and the natura history of neoplasms; and serves as the focal point for the Federal Government on the synthesis of clinical, epidemiologccd and environmental data relating to etiology.</Paragraph>
  </Section>
  <Section position="23" start_page="63" end_page="63" type="metho">
    <SectionTitle>
2. Chemotherapy ProgmlG Area
</SectionTitle>
    <Paragraph position="0"> Plans, directs and coordinates the Institute's Integrated cancer chemotherapy activltles, including intramural laboratory and clinical stuhes, contractor research, and research conducted in cooperation wth other Federal agencies; participates in evaluation of grant applications in the field of cancer chemotherapy, and plans and directs the research aspects of the ~dtimore Cancer  Research Center, A collaborative effort be-tween the Division of Hospitds and the NCI .</Paragraph>
    <Paragraph position="1"> 3. Getiernl Lnboratories and Clintcs Propam Area - Plans and directs the Institute's general (as distinguished from specifically targeted) laboratory and c1inlca.l reseqch activities.</Paragraph>
    <Paragraph position="2">  Interests. ~evelo~rpent of techniques, instrumentatlon and other resources which will improve the prevention, diagnosis or treatment of glaucoma, retinal and choroidd disease, corneal dlsease, cataract, congenital and developmental ab~~ormalities and other vlsion related health problems.</Paragraph>
    <Paragraph position="3">  materials that are compat~bre wth blood and other tissues, instrumeatation for improved diagnosis and monitoring of patients, and devlces for the treatment of patients with mad&lt;quate or fading circulatory or pulmonary functions; cause, treatment and prevention of myocar&amp;al infarction; supply,. safety, utdization, distribution and use of blood; respiratory chseases; and sickle cell anemia.  canon and computer techniques to meet the information needs of health care professionals. Inforrnatlon storage and retrieval systems, micro&lt;dm systems, clearinghouses, data banks, time-sharing computer networks. Cable, microwave and satellite com~unications networks for voice, video, and data transmission. Computer and communicatiqns terminals, facsimile, slowscan vldeo Evaluation of interation of communications technology and health care systems. - null lritcrcsts: Axds institution ~n thc cc-nstructlorl of tlcw and icrnodclccl n~cdic,~l rcscnrch facilitics: llclps thcrn cstablisl~ .~ttd opetatc .I tlut11bcr of gellcr.d clinical rese'u-ch centers: su y plics a va~iety of unuru,d .uid liigllly special resea~cll scraces and cquipn~crlt such as computers: suyports regional prirn'xto research centers, and stir~~ulates well-balanced institutional rescarch programs through grants for genernl blomedicd research support mntical and other computer-related sciences. l~lcludln~ inforsilation proces~r;~, in support of NIH programs.</Paragraph>
    <Paragraph position="4"> *-l~l~ll~~$.c; 131 cctor, Chvislon of Co~nputcr Rcsc u '11  Interests:, Plans and conducts research, developmental, and demonstration programs in mathe-</Paragraph>
  </Section>
  <Section position="24" start_page="63" end_page="63" type="metho">
    <SectionTitle>
HEALTH RESOURCES ADMINISTRATION
</SectionTitle>
    <Paragraph position="0"> Bureau of Health Services Research emphasis in five broad categories:</Paragraph>
    <Paragraph position="2"> Plnnnlnq, L regulation, and licensurc: (2) Interests: Supports research concerning the Productlv~t~ of health services. (3) Econorganization, financing, and ntlllzatlon of omic analysis: (4) Quality of health health services research. Interested in new cnre; 5) Health care ddta and informamethods or improving old methods. Pnmary tion systems.</Paragraph>
    <Paragraph position="3"> The Bureau is divided into the. following major divisions:  Interests: Research includes nursing practice, nursing educatlonal programs, organization ind delivery of nursing services to patients, and nursing as an occupation.</Paragraph>
    <Paragraph position="4"> ~esearch also includes studles 111 various n~cdical and other spccialitles related to nursing care and related health programs.</Paragraph>
    <Paragraph position="5">  concerlung health care delivery, educatlonal technology, ct,.trriculum development, and rnstitutional development among others.</Paragraph>
    <Paragraph position="6"> Confac t. Dl. ~obert Kn~uss Tcl. 301, 496-5608 Acfdress* Scc below lDivrsion of Comprehensive Hedth planning Irltcraats. Supports rssearch in two pajor areas (I) technical assistance: (2) health plannm g technical devel~prnen t. ~evelo~mei-i t approach to using new tools and methodolog)  emphasizes development of new methods of treatment fcr addiction of a broad range of National Institute of Alcohol Abuse and A~co- addicuve agents; (2) Training and public holism education: supports efforts to develop and install systems related to the Integrated drug Interests. Supports research concerning treat- abuse management information systems. ment evaluation, clinical afid non-clinical studies relating to alcohol and health, problems, pro- Address: Contracting Officer jects and special studies on the alcohol con- Contracts Management Branch suming populaaon such as drunken drivers, ~ationd Institute on Drug Abuse poverty-stricken alcoholics, convict alcoholics, 11400 ~ockville Pike and the inebriated.</Paragraph>
    <Paragraph position="7"> Rockville, Maryland 208 52</Paragraph>
  </Section>
  <Section position="25" start_page="63" end_page="63" type="metho">
    <SectionTitle>
FWD AND DRUG ADMINISTRATION
</SectionTitle>
    <Paragraph position="0"> Interests: Primarily concerned with research Contact: Mr. ~loyd Sundquist and development studies in radiological health, Director of Contract ab j Grant Administration biologics, drugs, veterinary medicine, medical ~ood and Drug Admin~strationl devices, and cosmetics. Colleges, universit~es, 5600 ish hers Lane and other education institutions reptesent most ~ockdle, Maryland 20852 FDA contracts. Tel: 301, 44316890</Paragraph>
  </Section>
  <Section position="26" start_page="63" end_page="63" type="metho">
    <SectionTitle>
OFFICE OF EDUCATION
</SectionTitle>
    <Paragraph position="0"> Interests: Research and development in the Address: field of educational msthodology and tech-Director, Contracts and ;rants Division niques, incluhng curriculum development,</Paragraph>
  </Section>
  <Section position="27" start_page="63" end_page="63" type="metho">
    <SectionTitle>
SOCIAL AND REHABILITATION SERVICE
</SectionTitle>
    <Paragraph position="0"> Interests kesearch , development, and evdua- Address* Associate Administrator for hon o[ socid programs in the areas of aging, Planning, ~esearch and Trainlng rehabilitauon, community services, Medicaid - Social and Rehabilitation Servlce youth development and delinquency prevention. Room 5050, HEW South Budding</Paragraph>
  </Section>
  <Section position="28" start_page="63" end_page="63" type="metho">
    <SectionTitle>
DEPARTMENT OF HOUSING AND URBAN DEVELOPMENT
</SectionTitle>
    <Paragraph position="0"> Interests: ~esearch , development and evalua- Address: Assistant Secretary for Research tion of new technology leadmg to the solu~oian</Paragraph>
    <Section position="1" start_page="63" end_page="63" type="sub_section">
      <SectionTitle>
and Technology
</SectionTitle>
      <Paragraph position="0"> of problems of housing, urban development and</Paragraph>
    </Section>
    <Section position="2" start_page="63" end_page="63" type="sub_section">
      <SectionTitle>
Department of Housing and Urban Develop-
</SectionTitle>
      <Paragraph position="0"/>
    </Section>
  </Section>
  <Section position="29" start_page="63" end_page="63" type="metho">
    <SectionTitle>
DEPARTMENT OF THE INTERIOR
</SectionTitle>
    <Paragraph position="0"/>
    <Section position="1" start_page="63" end_page="63" type="sub_section">
      <SectionTitle>
Geological Survey
</SectionTitle>
      <Paragraph position="0"> Interest= Fundamental research in topography, Address Director, U.S. Dept. of the Inter101 geocliemistry , hydrology, geology, geophysics, Geolopcal Survey related science; and .new techn~ques and meth12201 Sunnse Valley Drive ods for appraising and conserving minerals and Reptofi, Va. 22092 water, Tel: 703, 860-6111 Bureau of Mas Itlterests: Basic and applied research in earaction. processing, use, reuse, and disposal of minads and mineral fuels: recovering and recycling mineral substances from wastes including urban pf~~sc; research in improving mining practices ad dewsing new systems that will reduce health and safety halards to worlie~s in the metal, nonmetal, and coal producing indust~ies; and environmental quality research and studies associated with problems of surface disturbance, air,;and water pollution, and waste disposal that are common to the processes used in elitracting minerals fr m the earth.</Paragraph>
      <Paragraph position="1"> Address: Director, Bureau of Mines &amp;quot;C&amp;quot;' Street between Eighteenth and Nineteenth</Paragraph>
    </Section>
    <Section position="2" start_page="63" end_page="63" type="sub_section">
      <SectionTitle>
Streets
</SectionTitle>
      <Paragraph position="0"> Washington, D. C. 20240 I'el: 202, 1834815 Office of Water Resources Research Interests: Encourages and supports research investigations ded~g with major problem areas and which hold promise of contributing to the solution of important water problems.</Paragraph>
      <Paragraph position="1"> Address. Director, office of Water Resources</Paragraph>
    </Section>
    <Section position="3" start_page="63" end_page="63" type="sub_section">
      <SectionTitle>
Minerds Research
</SectionTitle>
      <Paragraph position="0"> Interests: Encouraga and support elecmcd power transrmssion and &amp;stribution research to improve the cost for electkical cables. Work in cooperation with the Electnc Power Commission in ~difomia.</Paragraph>
      <Paragraph position="1">  blterests. Encourages and supports research dealing with environmental studies, wldlife studies, historical architecture studies, site survey studies, natural hlstory studies, and social history studies. The NPS sup ports ecological studies dealing with any relationships of a concentrated number of  persons affecti~ig ecology, such as visitors, etc.</Paragraph>
      <Paragraph position="2"> Address.</Paragraph>
      <Paragraph position="3">  to award grants. They do, h~wever, contract w~th universities and colleges for research interests, Basic research is conducted and is necessary for the preservation and enhancement of natural resources.</Paragraph>
      <Paragraph position="4"> Address: Dr. Eugene Hester, Associate</Paragraph>
    </Section>
    <Section position="4" start_page="63" end_page="63" type="sub_section">
      <SectionTitle>
Criminal Justice
</SectionTitle>
      <Paragraph position="0"> Interests: The NILECJ conducts research pro- (5) Correction - a) xqconceptualization grams for the LEAA in seven categones: toward worklug for new methods of rehabilitation; b) accessment of adult  (I) Community Crlrne Prevention - violence, correction; c) offender classificatiorl. crime- victms, environmental design, white collar crlrne and fencing, etc. (6) Advanced Technology Division ~eals with hardware - pollce armor, (2) ~uvenile Delinquency - juvenile incarceration, body equipment, police vehicles, closedemployment, and youth cnme, etc. circuit televksion ballist~c shields, etc. (3) Police - police patrol, police integriry-and (7) Education and Manpower professional resp~asibllit~, police administration, a) - Education Programs under LEAA forensic science, personnel utdization. b) - Visiting el lows Programs c) - Manpower Surveys (4) Courts - measures for courts, sentencing, court administration, diversion. ~d&amp;ress Director of Research Programs</Paragraph>
    </Section>
  </Section>
  <Section position="30" start_page="63" end_page="63" type="metho">
    <SectionTitle>
DEPARTMENT OF LABOR
</SectionTitle>
    <Paragraph position="0"> Office of Administrative Services - Office Occupational Safety and Health Adminisof the Assistant Secretary tration Itlterests: The office of Administrative Services directs and controls a11 of the following Department of Labor facilities and services relating to procurement.</Paragraph>
    <Paragraph position="1"> Ill terests: Assists and encourages the states in their efforts to assure safe and healthful worklng conditions. Provides research, inforrnatlon education and training m the field of occupational safety and health.</Paragraph>
    <Paragraph position="2"> Address: See below Address: See below.</Paragraph>
    <Section position="1" start_page="63" end_page="63" type="sub_section">
      <SectionTitle>
Bureau of Labor Statistics
Enlployment Standards ~dministration
</SectionTitle>
      <Paragraph position="0"> Interests: Conducts a program of statistical sweys and economic analyses on laborrelated issues including manpower and employment, prices and living conditions, wages and industrial relations, produc tivlty and technological developments, occup atiohd safety and health, structure and growth of the economy, povmy, urban conditions and related socio-economic issues, and international labor comparisons.</Paragraph>
      <Paragraph position="1"> Address. See below</Paragraph>
    </Section>
    <Section position="2" start_page="63" end_page="63" type="sub_section">
      <SectionTitle>
Labor - Management Services Administra-
</SectionTitle>
      <Paragraph position="0"> tion Interests: Dcvelops labor-manageme nt policy, assists national interest collective bargainmg and counsels state and local govemments on labor-management relations. blterests: Advlse and assist the Secretary on all matters related to Employment Standards programs, policies, and systems, and provides executive direction for the Nation's wage-hour, employee's compensation, wage determination, federal contract compliance, programs for women, and all other related programs.</Paragraph>
      <Paragraph position="1">  Interests: Administers the manpower research Director, office of Manpower, R &amp; D and development program and the review and Room 9100, Patrick Henry Building analysis of manpower program performance. Washington, D.C. 20213</Paragraph>
    </Section>
  </Section>
  <Section position="31" start_page="63" end_page="63" type="metho">
    <SectionTitle>
BUREAU OF INTERNATIONAL AFFAIRS,
</SectionTitle>
    <Paragraph position="0"> Itzterests: Adminls~ers the international labor Address: Bureau of International ~ffairs and manpower activities of the Department of</Paragraph>
  </Section>
  <Section position="32" start_page="63" end_page="63" type="metho">
    <SectionTitle>
DEPARTMENT OF TRANSPORTATION
OFFICE OF THE SECRETARY FOR UNIVERSITY RESEARCH
</SectionTitle>
    <Paragraph position="0"> Interests: The office of the Secretary for University Research is the central office of DOT concerned with the procurement of services and studies to unive+dties and colleges relating to the following individual offices: Office of the .Assistant Secretary for Policy and International Affairs Interests. The procuremehts for this office include aqdies and analyses conducted on a broad range of transpoitation policy, economic and environmental problems to determine the national transportatioh requuements; and studies aimed at removing the impedbents to the efficient flow of passengers and cargo.</Paragraph>
    <Paragraph position="1"> Office of the Assistant Secretary for Environment and Urban&amp;quot; Systems Interests: The procurements for this office include transportanon policy studies relatlng to the environment and urban systems and includes research to minimize the environmental and urban impact of transportation, improve the intermodal urban- and regional transportation planning process, and coordinate and strengthen transportation policy effecting the environmental and urban systems.</Paragraph>
    <Paragraph position="2"> office of the Assistant Secretary for Systems Development and ~echllokogy Interests: The procuremcnts fdr this office pclude analytical studies to support establishment of R &amp; D gods, objectives, policies, plar and information systems; and R &amp; D studies and hardware related to noise abatement, pollution, telecommunications and advanced rnulti modal programs; Office of the Assistant Secretary for safety and Consumer ~ffalrs Interem: The procvrements for this officc include a broad range of services and items relating to programs concer~ing pipeline safety transportation of hazardous materials, safety program coordination, transportation secunty and consumer affairs.</Paragraph>
    <Paragraph position="3"> Any iqz formation concerning procurement carned on wrthin the Office of the Secretary of Transportation may be obtalned by contac ting the followtng central ofice</Paragraph>
  </Section>
  <Section position="33" start_page="63" end_page="63" type="metho">
    <SectionTitle>
FEDERAL AVIATION ADMINISTRAT ION
</SectionTitle>
    <Paragraph position="0"> Systems Research and Development Service National Aviation Facilities Experimental Center Interests: Design for the National Airspace System and research and development activities associated with programs pertaining to air traffic control, air navigation, communications, airports, avianon weather, aeronautics (airframe structures and materials, aircraft flight systems, propulson) and in aircraft engineering and safety, including airborne devices, equipment, and systems other than navigation and communications. Special emphasis is placed on the development of simplified, low cost, functional general aviation cockpit instrumentation and control systems. Also conducts close liaison \nth the Department of Defense on the dcvclopment of V/STOL and other subsonic aircraft for application to civil requirements. null Address. Director, Systems ~esearch and</Paragraph>
    <Section position="1" start_page="63" end_page="63" type="sub_section">
      <SectionTitle>
Development Servlce
Federal Aviation Administration
</SectionTitle>
      <Paragraph position="0"> Washington, D.C. 20590 Tcl: 202, 426-3577 Interests: Conducts all research, experimcntation, testing and evaluation of aviation research and development programs under the direction of the Systems ResParch and Development S'ervice. null  Interests. Conducts a medical research ,program that deals in clinical medic in^ and human factors. The program is directed toward the solution of current and anticipated problems in the memcal aspects of amation.</Paragraph>
    </Section>
  </Section>
  <Section position="34" start_page="63" end_page="63" type="metho">
    <SectionTitle>
FEDERAL RAILROAD ADMINISTRATION
</SectionTitle>
    <Paragraph position="0"> Interests: Research and development in rd- Address.</Paragraph>
    <Paragraph position="1"> road and high speed ground transportation, Chief, Contracts and Procurement vision including but not limited to aerodynamics, Federal Rdroad ~drniniskation vehicle control, communications, and guideways.</Paragraph>
    <Section position="1" start_page="63" end_page="63" type="sub_section">
      <SectionTitle>
Department of Transportation
</SectionTitle>
      <Paragraph position="0"> It also contracts for demonstrations to deter- 400 Seventh Street, S.W.</Paragraph>
      <Paragraph position="1"> mine contributions that high speed ground Washington, D.C. 20590 transportation could make to effect an effic~ent Tel: 202, 426-0872 and cconornical intercity transportanon system, and surveys to measure public response to changes in service.</Paragraph>
      <Paragraph position="2"> NATIONAL HIGHWAY TRAFFIC SAFETY ADMlNlSTRATlON I~rterests. The National Highway ~rafftc safety Administration is responsible for programs to reduce deaths, injuries, and property damage resulting from traffic accidents. It is also responsible for the establishtncnt and enforcement of safety performance standards fo~ motor vclliclcs and ass~ciatcd cquipmcnt; the cfbctuation of supporting research, includins thc acquisition of rcqui~cd testing faciliti(*s: '~nd the adminlstratio~l of the National Drivcr Kcgstcr.</Paragraph>
      <Paragraph position="3"> The Administration furtllcr provides for a co-ordinated national highway safcty grant program to bc carricd out on a matching fund basis with thc Statcs, and authorizes supporting highC null way safety rescarch, dcvclopment, and dcmonstration programs, Thc NHTSA enters into contracts with private industry, educational institutions, rton-profit organizat~ons, and State and local governments for defects investigations, crashworthlncss progra IYIS, system operations, einergency medical services dcmonstratio~~ projects, safcty manpower development, drivcr/vehicle intcrac tion, expcrimcntal safety vchicles, test and cvaluatlon, vcliicle structures, occupant. packaging, bio~~~cchanics, passive rcstraln t tests, cornputel support, nlanagchent studics, and data aciluisition. null</Paragraph>
    </Section>
  </Section>
  <Section position="35" start_page="63" end_page="63" type="metho">
    <SectionTitle>
URBAN MASS TRANSPORTATION ADMINISTRATION
</SectionTitle>
    <Paragraph position="0"> Interests. Thc Urban Mass Transportation ~ddevelopment of improved managerial proministration (UMTA) is responsible for the con- cedures in the areas of maintenance, routduct of a program to assist in the development ing, scheduling, accounting and reporting[) of improved mass transportation through financial ad to State and local governments and Address: Mr. Philip Hughes their instrumcntalitics, and research and develop- Urban Mass Transportation Administration Inent. Principal procurement activities arc cen - 2100 Second Street, S.W.</Paragraph>
    <Paragraph position="1"> tered around the UMTA office of Research De- Washington, D.C. 20590 velopment and Demonstration. Research con- Tcl: 202, 426-0090 tracts may be for either hardware (such as the U.S. COAST GUARD ~nvironmentd Transportation and ~echnology warnlng systems for buoys, and other basic Division ocean engineering systems.</Paragraph>
    <Paragraph position="2"> Interests Conducts the full range of R &amp; D Address chief Commandant (G-det-62) activities within two main branches:  (1) pol- null office of Research and Development lution - sewage and waste water treatment, re- United States Coast Guard mote detection of oil and other such hazards, 400 Seventh Street, S.W. (2) alds to navigation - light power sources, Washington, D.C. 20590 Tel: 202, 426-1042</Paragraph>
    <Section position="1" start_page="63" end_page="63" type="sub_section">
      <SectionTitle>
Marine Safety i&amp;quot;eu$mology Division
</SectionTitle>
      <Paragraph position="0"> Ititerests: Conducts research in two main Address: areas: (I.) rescue and search projects in Chief Commandant (Gdet-62) helicopter operations, detection of people in Office of Research and Development water, communications, satellite communications, U.S. Coast Guard domestic ice-breaking, and high performance 400 Scventh Street, S,W, watercraft;' (2) research in co~nrnercial visull Washington, D.C. 20590 safety projects in fue and explosion safety, Tel: 202, 426-1008 research to minimize collisions, investigation of struc turd problems, survival problems, and crew safety problems.</Paragraph>
    </Section>
  </Section>
class="xml-element"></Paper>