File Information
File: 05-lr/acl_arc_1_sum/cleansed_text/xml_by_section/metho/82/c82-1016_metho.xml
Size: 18,184 bytes
Last Modified: 2025-10-06 14:11:23
<?xml version="1.0" standalone="yes"?> <Paper uid="C82-1016"> <Title>REFERENTIAL NETS WITH ATTRIBUTES</Title> <Section position="1" start_page="0" end_page="0" type="metho"> <SectionTitle> REFERENTIAL NETS WITH ATTRIBUTES </SectionTitle> <Paragraph position="0"> Christopher U. Habel * ODe of the essential problems in natural language production and understanding is the problem of processing referential relations. In this paper I describe a model for representing and processing referential relations: referential nets with attributes. Both processes (analyzing and generating referential expressions) are controlled by attributes. There are two types of attributes, on one hand, the ones to the internal substitutes of the objects spoken about, on the other hand, the ones to the descriptions of these objects.</Paragraph> <Paragraph position="1"> I. BASIC NOTIONS: KNOWLEDGE AND SEMANTIC REPRESENTATION It is a well-known fact that (computational) models of human language production and understanding have to take the knowledge of the speaker and the listener into consideration. Thus, one of the main problems is to describe and represent this knowledge and to distinguish relevant subtypes. A suitable classification which I described in detail in Habel \[7\] distinguishes three subtypes: - knowledge of facts or assertions, e.g. of states of the world, events, etc., the &quot;factual knowledge', - knowledge of rules, e.g. of rule-like relations between objects or states of the world (or classes of such entities), the &quot;inferential knowledge', - knowledge of objects, e.g. of persons in the world, the &quot;referential knowledge'.</Paragraph> <Paragraph position="2"> (In this paper I give emphasis to the third type: referential knowledge represented by &quot;Referential Nets').</Paragraph> <Paragraph position="3"> The common core for representing all these types of knowledge is the &quot;semantic representation language&quot; SRL. From the formal point of view SRL is a (propositional) symbolic language as described by Kalish/Montague \[8\]. Well-formed SRL-expressions are generated by normal (recursive) formation rules. I distinguish - as usual - two kinds of meaningful expressions in SRL, terms and formulas. Thus we define: TER - the set of well-formed terms FOR - the set of well-formed formulas</Paragraph> </Section> <Section position="2" start_page="0" end_page="101" type="metho"> <SectionTitle> SRL = <FOR, TER> </SectionTitle> <Paragraph position="0"> For the purpose of knowledge representation SRL contains some specific operators, e.g.</Paragraph> <Paragraph position="1"> on the factual level: tenses, beliefs, ......</Paragraph> <Paragraph position="2"> on the inferential level: entailment relation,... on the referential level: description operators of various types. Since SRL is a representation language, I now have to treat the central question of &quot;representation theory': What kinds of entities are represented by the different entities of SRL? Here I follow - with minor changes, which I will not discuss in this paper Miller \[I0\]: The expressions and entities of SRL represent cognitive concepts. The operators correspond to different types of &quot;concepts&quot; in the Miller-approach, e.g. predicative , nominal and modifying concepts (this third class are the &quot;operators&quot; of Miller.). E.go by means of description operators it is possible to construct nominal concepts from 102 CH. U. HABEL predicative concepts or other nominal concepts.</Paragraph> <Paragraph position="3"> I shall now illustrate SRL with an example. In understanding and representing the sentence (i) I will meet Barbara in the university tomorrow.</Paragraph> <Paragraph position="4"> the act of referring to &quot;I&quot;, &quot;Barbara&quot;, and &quot;the university&quot; is to be processed, i.e. the (obJect-)orientation to them is the interesting problem from the referential point of view. The SRL representation of (i) is (2) time (meeting (l,Barbara, ETA(x) :university(x) ),tomorrow) The arguments of &quot;time&quot; and &quot;meeting&quot; present the most relevant types of referential operators: &quot;I&quot; and &quot;tomorrow&quot; are deictic substitutes (personal or temporal), Barbara is a proper name and the expression &quot;ETA(x): university(x)&quot; is built up from the nominal concept &quot;'university&quot; (representing a class of institutions for education and science) and the variable-binding description operator &quot;ETa'. &quot;ETA(x): p(x)&quot; has the meaning &quot;'an element from the class {x/p(x)>&quot;. ETA is an indefinite analogy to the IOTA-operator of formal logics (Similar operators are introduced by Hilbert as EPSILON- or ETA-operator.). The semantics of the description operators is given by a set of inference and evaluation rules. E.g. there is an inference rule which relates the ETAexpressions to expressions with existential quantification:</Paragraph> <Paragraph position="6"> Some evaluation rules will be described in the following chapters, e.g. &quot;creation of a referential object'. (Both types of rules are described in detail in Habel \[5\], \[6\].) Thus, ETA is the formal representative of one of the meanings of indefinite articles.</Paragraph> <Paragraph position="7"> Further descriptions of the objects mentioned in (i) or in the SRL expression (2) can be generated by (focussing) transformations similar to the solution of algebraic equations: The university mentioned above is also the one described by (4) ETA(x) :time ( meeting(l,Barbara, university(x)),tomorrow ) 2~ or the natural language equivalent &quot;a (the) university where Barbara and I will meet tomorrow&quot;. (The uniqueness of the ETA-description depends on the sltuatlonal/textual context.) Before we pass on to a detailed description of &quot;referential nets', I will make a short remark on discourse and communication. As mentioned above human language production and understanding is based on the speaker-listener's knowledge of the language, the world and particularly the participants in the communication. Thus, I assume a discourse model 3* which is based Dn tripartite knowledge (namely factual, inferential, and referential), each of these types is represented by structured sets of SRL-expressions.</Paragraph> <Paragraph position="8"> \</Paragraph> </Section> <Section position="3" start_page="101" end_page="101" type="metho"> <SectionTitle> 2. THE REFERENTIAL NET </SectionTitle> <Paragraph position="0"> Processing, e.g. storing or retrieving, referential relations is executed by a particular component of the discourse processing system, the &quot;referential procedures&quot; (ReP). The RePs work on a memory structure which is adequate for the representation of knowledge about objects, the &quot;referential net&quot; (RefN) 4*. A RefN consists of entities of three different types: - referential objects (RefOs), which are the internal substitutes for the objects spoken about, - descriptions, i.e. terms of SRL which describe the RefOs, - attributes, i.e. properties of the description-relation between a RefO and one of its descriptions or properties of the RefO (see chap. 3).</Paragraph> <Paragraph position="1"> After detecting a new indefinite description (as ETA(x) : unlversity(x)) ReP creates a new &quot;referential object'&quot; (RefO). During the discours6 (after the identification process) further descriptions of the same RefO will be linked to this RefO. (The relation among the descriptions of the same RefO is the wellknown relation of eoreference,) ~us we have, for example, the following set of descriptions of the RefO mentioned above : (5) r.l~.~ d. II--ETA(x) : university(x) d.12-- ETA(x): time (meeting(l'Barbara'university(x))'tdegmdegrrdegw) d. I3-- IOTA(x) : professor-at(Barbara,university(x))</Paragraph> </Section> <Section position="4" start_page="101" end_page="101" type="metho"> <SectionTitle> REFERENTIAL NETS WITH ATTRIBUTF_~ 103 </SectionTitle> <Paragraph position="0"> For the description of the referential net I use SRL and in addition a &quot;finite but extendable&quot; set, REFO, of &quot;referential objects'. The referential objects are seen as a kind of basic term of SRL. The set REFO is analogous to the set of variables. The phrase &quot;finite but extendable&quot; is used to emphasize the dynamic aspect of the set REFO and the RefN. REFO consists of exactly those RefOs that are actually needed. It is necessary to distinguish two types of terms, one type which consists of the mental representatives, i.e. the RefOs, and the other which contains those terms which function as descriptions. Naming this set by D-TER we have TER - REFO U D-TER. Thus, referential nets, RefNs, can he defined as (6) RefN c D-TER x REFO * For <d-ter, r.i> 6 RefN I also use &quot;d-ter deser r.i'&quot; , &quot;d-ter is a description of r.i&quot;. In this way a description-relation is defined on the basis of the RefN. By means of referential nets it is possible, e.g. to arrive at one object (i.e. RefO) via different aspects, i.e. different descriptions, of this object. It is only by such aspects (cp. Schank's \[II\] way of memory discrimination), that a system is able to choose &quot;'a best description&quot; of the objects which are the theme of the discourse (see below). The referential net (5) is simplified in two crucial points, first, that the descriptions of r.l operate on MSRL ~pressions and not on RefOs, too. But this can be treated similarly to the focussing process of (4), which can be formallzed now: (7) p(r.i) 8--- r.i -- ETA(x) : p(x) Using (7), the descriptions can be solved with respect to all RefOs, i.e. all arguments are filled by RefOs or open terms. Thus we can derive (8) from (5):</Paragraph> <Paragraph position="2"> Secondly, ReP has tO explicate the referential links of deictlc substitutes, e.g.</Paragraph> <Paragraph position="3"> &quot;r&quot; refers to the speaker, etc.. &quot; &quot; stands for the specific ReP which evaluates deictlc expressions, i.e. which ins~antiates deictic ~xpressions with RefOs, e.g.</Paragraph> <Paragraph position="4"> &quot;I'&quot; - SPEAKER(l) (cp. fn 6.). By the same methods new RefOs for &quot;we&quot; can be created, e.g. in the processing of &quot;There we will discuss some papers.&quot; (We = Barbara + I)</Paragraph> </Section> <Section position="5" start_page="101" end_page="101" type="metho"> <SectionTitle> 3. ATTRIBUTES IN REFERENTIAL NETS </SectionTitle> <Paragraph position="0"> I will now extend the basic two-place relation between RefOs and descriptions to the more adequate concept of a many-place referential relation. The additional places will be called &quot;attributes'. In the first step of extention I give emphasis to &quot;attributes of descriptions'. They represent properties of the description relation, i.e. of pairs <d, r> E RefN. Thus, the extention to an attributed RefN (ARefN) uses a set ATT of attributes and changes (6) to (9) ARefN c ATT x D-TER x REFO .</Paragraph> <Paragraph position="1"> For <att, d, r > C ARefN I also say &quot;'att is an attribute of d with regard to r&quot;. Some possible attributes are (Note the fact that this list is not complete and not fully adequate): - syntactic and semantic features of the description, e.g. gender, number, sexus, etc.</Paragraph> <Paragraph position="2"> - numerical values, e.g. &quot;grades of relevance of a description&quot;, &quot;degrees of &quot;being the TOPIC / being in the FOCUS&quot; &quot;, &quot;recency&quot;. 5* A simple strategy for de-/coding is: '~se/try the description or RefO with highest degrees !'&quot; - names of persons, e.g. participants of earlier discourses. By this attribute it is possible to find a description relevant to speaker and listener (Cp. Clark / Marshall's \[I\] co-presence triples).</Paragraph> <Paragraph position="3"> - situations in which the description / RefO Is relevant or was introduced (cp. Webber's \[12\] &quot;evokedeg-predicate).</Paragraph> <Paragraph position="4"> - links to the factual knowledge; thus we can answer such questions as &quot;What will Barbara do tomorrow?&quot; 104 CH. U. HABEL Before I explain the concept of &quot;attributes to descriptions&quot; by some examples, I will new sketch some aspects of function and statue of the attributes. Firstly, in both processes, analysis as well as generation of referential expressions, it is necessary to take into consideration a set of alternatives. All of them are possible candidates to be the referent (in analyzing) or to be a good description (in producing) respectively. The goal of the ReP is to choose exactly one from the alternatives, namely &quot;the best'. It is obvious that trying to choose the best can fail. Then the decision has to be revised afterwards. (But this is a total different problem.) In other words, the function of the attributes is to point to the most appropriate (or being more careful: to that which seems to be the most appropriate) of the alternatives. Secondly, I think that, on one hand, knowing and processing attributes is part of human natural language processing, and therefore, these attributes have to be part of the discourse processing model. On the other hand, knowledge of attributes is different (cp. the list of possible attributes, above) from the factual or referential knowledge. Thus, I propose a strict separation between descriptions and their attributes. This strict distinction contrasts to Webber's \[12\] solution: her use of the &quot;evoke'-predieate mixes the concepts of descriptions and their attributes in a cognitive inadequate manner. Now I shall continue with examples of the use of attributes * Let us suppose a situation in which a questioner Q asks: (I0) Where can I study Computational Linguistics? and the answerer A knows that the university which is (internally) represented by r.l is a good university to study CL. The selection of the best description of r.l depends on Q and A (and the situational context). I assume the following attributes to the descriptions of r.l (from (5), (8)): (ii) descr, general relevant to: time of relevance:</Paragraph> <Paragraph position="6"> (II) describes a situation, in which the name / location of the university in question is not represented, e.g. since the information about the possibility to study CLwas given by an earlier utterance of Barbara &quot;We have some courses in CL&quot; but A (whose RefN is represented by (ii)) does not know where Barbara teaches. The best description can be chosen by comparing the attributes of Q with those of the descriptions of r.l. If Q knows Barbara, .i.e. <KNOW(Barbara)> ATT (Q) 7*, then the best description will be: (12) d.13 - &quot;The university where Barbara teaches (as a professor).&quot; Ranking d.13 higher / better than d.12 depends on the dimension &quot;time of relevance'. Furthermore, if Q knows more about r.l it is possible that Q will react with, (13) &quot;Well, at Amherst.'&quot; and now A is able to extend hls/her RefN with a further description of r.l (14) r.l - d.14 - Amherst -- &quot;general high relevance&quot; &quot;always&quot; (This attribute depends on the fact that proper names have high relevance in general.) Anagously we can use attributes in the decoding process. Look at the following pair of sentences (similar to (I)).</Paragraph> <Paragraph position="7"> (15) a. I will meet the head of the linguistic department tomorrow. b. She is a specialist for Montague grammar.</Paragraph> <Paragraph position="8"> To solve this referential problem ( and not to be puzzled by the pronoun &quot;she&quot; ) we need further background knowledge, e.g. &quot;Barbara is the head of the linguistic department.&quot;. With this prior knowledge and the additional (theoretical) concept of &quot;attributes to RefOs&quot; 8* as described in (16) &quot;female&quot; &quot;local TOPIC'- r.3-- d.31- Barbara d.32- IOTA(x) : head-of(x ,linguist. dept.) the referential problem of (15) is easy to solve. A system (human listener or machine) has to match the pattern corresponding to the pronoun &quot;she&quot; ~Ich is induced by the context (15.5) against the descriptions, their attributes and the attributes of the RefO's. The crucial attribute is here &quot;female&quot; for both &quot;she&quot; and &quot;r.3&quot;.</Paragraph> <Paragraph position="9"> This type of congruence between gender of the pronoun and sexus of the RefO, i.e.</Paragraph> </Section> <Section position="6" start_page="101" end_page="101" type="metho"> <SectionTitle> REFERENTIAL NETSWITHATTRIBUTES 105 </SectionTitle> <Paragraph position="0"> between attributes of descriptions and those of the RefOs, is very important in German. In spite of the syntaetlc ineongruence (neutr. vs. fem.) it is possible (in colloquial German) to say: (17) Ich traf gestern ein Maedchen. Sle hatte rote Haare.</Paragraph> <Paragraph position="1"> I met yesterday a glrl . She had red hair .</Paragraph> <Paragraph position="2"> neutr. ~ fem.</Paragraph> </Section> <Section position="7" start_page="101" end_page="101" type="metho"> <SectionTitle> 4. CONCLUDING REMARKS </SectionTitle> <Paragraph position="0"> By means of an extended referential net (w~th attributes both of descriptions and RefOs) it is possible to describe how to store and process referential entities (and expressions) in a cognitive adequate and powerful way.</Paragraph> <Paragraph position="1"> Following the principles and concepts described above a system for anaphora resolution has been implemented (Guenther \[3\]), which will be a component of the next version of the BACON-system (cp. fn 3).</Paragraph> </Section> class="xml-element"></Paper>