<?xml version="1.0" standalone="yes"?>
<Paper uid="P85-1035">
  <Title>Native: Learner: Native: Learner: Native: Learner: Native: Remember the s~ory of David and Goliath?</Title>
  <Section position="3" start_page="284" end_page="284" type="metho">
    <SectionTitle>
(3) Pattern Constructor: Learning of phrase patterns
</SectionTitle>
    <Paragraph position="0"> is accomplished by analyzing parsing failures. Each failure situation is associated with a patternmodification action.</Paragraph>
    <Paragraph position="1"> (4) Concept Constructor: Learning of phrase concepts is accomplished by a set of strategies which are selected according to the context.</Paragraph>
    <Paragraph position="2"> Schematically, the program receives a sequence of sentence/contezt pairs from which it refines its current pattern/concept pair. The pattern is derived from the sentence and the concept is derived from the coLtext. However, the two processes are not independent since the context influences construction of patterns while linguistic clues in the sentence influence formation of concepts.</Paragraph>
  </Section>
  <Section position="4" start_page="284" end_page="288" type="metho">
    <SectionTitle>
2. Phrasal Representation of the Lexicon
</SectionTitle>
    <Paragraph position="0"> Parsing in RINA is central since learning is evaluated in terms of parsing ability before and after phrases are acquired. Moreover, learning is accomplished through parsing.</Paragraph>
    <Section position="1" start_page="284" end_page="285" type="sub_section">
      <SectionTitle>
2.1 The Background
</SectionTitle>
      <Paragraph position="0"> RINA combines elements of the following two approaches to language processing: Phra~-bued pattern matching: In the implementation of UC \[Wilensky84\], an intelligent help system for UNIX users, both PHRAN \[AJ'ens82 l, the conceptual analyzer, and PHRED \[Jacobs85\] the generator, share a phrasal lepton. As outlined by Wilensky {Wilensky81\] this lexicon provides a declarative database, being modularly separated from the control part of the system which carries out parsing and generation. This development in representation of linguistic knowledge is paralleled by theories of functional grammars {Kay79\[, and lezicalfunctional grammars \[Bresnan78\].</Paragraph>
      <Paragraph position="1"> Ca~,-b,,-,,ed demon pmming: Boris \[DyerS3 I modeled reading and understanding stories in depth. Its conceptual analyzer employed demon-based templates for parsing and for generation. Demons are used in parsing for two purposes: (1) to implement syntactic and semantic expectations \[Riesbeck74\] and (2) to implement memory operations such as search, match and update.</Paragraph>
      <Paragraph position="2"> This approach implements Schank's \[Schank77\] theory of representation of concepts, and follows case-grammar \[Fillmore681 principles.</Paragraph>
      <Paragraph position="3"> RINA uses a declarative phrasal lexicon as suggested by Wilensky \[Wilensky82\], where a lexical phrase is a pattern-concept pair. The pattern notation is described below and the concept notation is Dyer's \[Dyer83\] i-link notation.</Paragraph>
    </Section>
    <Section position="2" start_page="285" end_page="285" type="sub_section">
      <SectionTitle>
2.2 The Pattern Notation
</SectionTitle>
      <Paragraph position="0"> To span English sentences, RINA uses two kinds of patterns: lezical patterns and ordering patterns \[Arens82\]. In Figure I we show sample lexical patterns (patterns of lexical phrases). Such patterns are viewed as the generic linguistic forms of their corresponding phrases.</Paragraph>
      <Paragraph position="1"> I. ?x: (animate.a~ent) nibble :verb &lt;on ?y: food&gt;  2. ?z: Cpernou.Lgent) tLke:verb on ?y:p,tlent 3. ?x: (person.a~ent) &lt;put:verb foot:body-part do~m&gt;  The notation is explained below: (t) A token is a literal unless otherwise specified. For example, on is a literal in the patterns above. (2) ?x:sort denotes a variable called .~x of a semantic type sort. ?y:food above is a variable which stands for references to objects of the semantic class food.  (3) Act.verb denotes any form of the verb s!lntactic class with the root act. nibble:vet6 above stands for expressions such as: nibbled, hms never nibbled, etc.</Paragraph>
      <Paragraph position="2"> (4) By default, a pattern sequence does not specify the order of its tokens.</Paragraph>
      <Paragraph position="3"> (5) Tokens delimited by &lt; and &gt; are restricted to their specified order. In Pattern I above, on must directly precede ?y:food.</Paragraph>
      <Paragraph position="4"> Ordering patterns pertain to language word-order conventions in general. Some sample ordering patterns are: active: &lt;?x:agenr. ?y: (verb.~tive)&gt; passive: &lt;?x:pattent ?y: (verb.p~,.sPSve)&gt; *&lt;by ?Z : agent&gt; infinitive:&lt;to ?x: verb. active&gt; &amp;quot;?y: Iq~ent  The additional notation introduced here is:  (6) An * preceding a term, such as *&lt;by ?z:~ent&gt; in the first pattern above indicates that the term is optional. null (7) * denotes an omitted term. The concept for Ty in the third example above is extracted from the agent of the pattern including the current pattern.</Paragraph>
      <Paragraph position="5"> (8) By convention, the agent is the case-frame which precedes the verb in the lexical pattern. Notice that the notion of agent is necessary since (a) the agent is not necessarily the subject (i.e., she vu taken) and {b) the agent is not necessarily the actor {i.e., she received the book, he took a blo~), and (c) in the infinitive form, the agent must be referred to since the agent is omitted from the pattern in the lexicon. (9) Uni/ieation \[Kay79\] accounts for the interaction of  lexical patterns with ordering patterns in matching input sentences.</Paragraph>
      <Paragraph position="6"> So far, we have given a declarative definition of our grammar, a definition which is neutral with respect to either parsing or generation. The parsing procedure which is derived from the definitions above still has to be given.</Paragraph>
    </Section>
    <Section position="3" start_page="285" end_page="288" type="sub_section">
      <SectionTitle>
2.3 Parsing Objectives
</SectionTitle>
      <Paragraph position="0"> Three main tasks in phrasal parsing may be identified, ordered by degree of difficulty.</Paragraph>
      <Paragraph position="1">  (1) Phrase dlaambiguation: When more than one lexi null cat phrase matches the input sentence, the parser must select the phrase intended by the speaker. For example, the input the vorkeru took to the streets could mean either &amp;quot;they demonstrated&amp;quot; or &amp;quot;they were fond of the streets'. In this case, the first phrase is selected according to the principle of pattern speci\]icit 9 \[Arens821. The pattern ?X: person taXe:verb &lt;to the streets&gt; is more specific then ?x:person take:verb &lt;to ?y:thing&gt; However, in terms of our pattern notation, how do we define pattern specificity? {2) Ill-formed input comprehension: Even when an input sentence is not well phrased according to textbook grammar, it may be comprehensible by people and so must be comprehensible to the parser. For example, John took Nary school is telegraphic, but comprehensible, while John took Nzry to conveys only a partial concept. Partially matching sentences (or &amp;quot;near misses') are not handled well by syntax-driven pattern matehers. A deviation in a function word (such as the word to above) might inhibit the detection of the phrase which could be detected by a  semantics-driven parser.</Paragraph>
      <Paragraph position="2"> (3) Error-detection: when the hypothesized phrase  does not match the input sentence/context pair, the parser is required to detect the failure and return with an indication of its nature. Error analysis requires that pattern tokens be assigned a casesignificance, as shown in Section 4.</Paragraph>
      <Paragraph position="3"> Compounding requirements--disambiguation plus error-analysis capability-- complicate the design of the parser. On one hand, analysis of &amp;quot;near misses&amp;quot; (they bury a hatchet instead of they buried the hatchet) can  be performed through a rigorous analysis--assuming the presence of a single phrase only. On the other hand, in the presence of multiple candidate phrases, disambiguafinn could be made efficient by organizing sequences of pattern tokens into a discrimination net. However, attempting to perform both disambiguation and &amp;quot;near miss&amp;quot; recognition and analysis simultaneously presents a difficult problem. The discrimination net organization would not enable comparing the input sentence, the &amp;quot;near miss&amp;quot;, with existing phrases.</Paragraph>
      <Paragraph position="4"> The solution is to organize the discrimination sequence by order of generality from the general to the specific. According to this principle, verb phrases are matched by conceptual features first and by syntactic features only later on. For example, consider three initial erroneous hypotheses: (a) bury a hatchet (b) bury the gun, and (c) bury the hash. On hearing the words &amp;quot;bury the hatchet', the first hypothesis would be the easiest to analyze (it differs only by a function word while the second differs by a content-holding word) and the third one would be the hardest (as opposed to the second, huh does not have a common concept with hlttchet).</Paragraph>
    </Section>
    <Section position="4" start_page="288" end_page="288" type="sub_section">
      <SectionTitle>
2.4 Case-Frames
</SectionTitle>
      <Paragraph position="0"> Since these requirements are not facilitated by the representation of patterns as given above, we slightly modify our view of patterns. An entire pattern is constructed from a set of case-/tames where each case-frame is constructed of single tokens: words and concepts.</Paragraph>
      <Paragraph position="1"> Each frame has several slots containing information about the case and pertaining to: (a) its syntactic appearance (b) its semantic concept and (c) its phrase role: agent, patient. Variable identifiers (e.g., ?x. ?y) are used for unification of phrase patterns with their corresponding phrase concepts. Two example patterns are given below: The first example pattern denotes a simple literal  The third case frame in Figure 4 above, the indirect object, does not have any corresponding concept. Rather it is represented as a sequence of words. However the words in the sequence are designated as the marker, the determiner and the word itself.</Paragraph>
      <Paragraph position="2"> Using this view of patterns enables the recognition of &amp;quot;near misses&amp;quot; and facilitate error-analysis in parsing.</Paragraph>
    </Section>
  </Section>
  <Section position="5" start_page="288" end_page="288" type="metho">
    <SectionTitle>
3. Demons Make Patterns Operational
</SectionTitle>
    <Paragraph position="0"> So far, we have described only the linguistic notation and indicated that unification \[Kay79\] accounts for production of sentences from patterns. However, it is not obvious how to make pattern unification operational in parsing. One approach \[Arens82\] is to generate word sequences and to compare generated sequences with the input sentence. Another approach IPereiraS01 is to implement unification using PROLOG. Since our task is to provide lenient parsing, namely also ill-formed sentences must be handled by the parser, these two approaches are not suitable. In our approach, parsing is carried out by converting patterns into demons.</Paragraph>
    <Paragraph position="1"> Conceptual analysis is the process which involves reading input words left to right, matching them with existing linguistic patterns and instantiating or modifying in memory the associated conceptual meanings. For example, assume that these are the phrases for take: in the lexicon: ?x:person take:verb ?y:person ?z:locale John took her to Boston.</Paragraph>
    <Paragraph position="2"> ?x:person take:verb ?y:phys-obj He took the book.</Paragraph>
    <Paragraph position="3"> ?x:person take:verb off ?y:attire He took off his coaL.</Paragraph>
    <Paragraph position="4"> ?x:person take:verb on ?y:person David took on Goliath.</Paragraph>
    <Paragraph position="5"> ?x:person take:verb a bow The actor took a boy.</Paragraph>
    <Paragraph position="6"> ?x:thing take:verb a blow The vail took a blov.</Paragraph>
    <Paragraph position="7"> ?x:person take:verb ~to the streets~ The vorkern ~ok t,o the streets.</Paragraph>
    <Paragraph position="8"> The juvenile took t,o the e~reeCs.</Paragraph>
    <Paragraph position="9">  where variables ?x, :y and ?z also appear in correspondin&amp; concepts (not shown here). How are these patterns  actually applied in conceptual analysis?</Paragraph>
    <Section position="1" start_page="288" end_page="288" type="sub_section">
      <SectionTitle>
3.1 Interaction of Lexlcal and Ordering Patterns
</SectionTitle>
      <Paragraph position="0"> Token order in the lexical patterns themselves (Figure 5) supports the derivation of simple active-voice sentences only. Sentences such as: Msry vas ~,zken on by John.</Paragraph>
      <Paragraph position="1"> A veak contender David might, have left, alone, bu~ Goliath he book on.</Paragraph>
      <Paragraph position="2"> David decPSded to take on Gol'tath.</Paragraph>
      <Paragraph position="3">  cannot be derived directly by the given hxical patterns. These sentences deviate from the order given by the corresponding lexical patterns and require interaction with language conventions such as passive voice and infinitive. Ordering patterns are used to span a wider range of sentences in the language. Ordering patterns such as the one's given in Figure 2 depict the word order involving verb phrases. In each pattern the case-frame preceding the verb is specified. (In active voice, the agent appears imediately before the verb, while in the passive it is the patient that precedes the verb.)</Paragraph>
    </Section>
    <Section position="2" start_page="288" end_page="288" type="sub_section">
      <SectionTitle>
3.2 How Does It All Work?
</SectionTitle>
      <Paragraph position="0"> Ordering patterns are compiled into demons. For example, DAGENT, the demon anticipating the agent of the phrase is generated by the patterns in Figure 2. rt has three clauses: If the verb is in active form then the agent is immediately be/ore the verb If the verb is in passive form then the agent may appear, preceded by by.</Paragraph>
      <Paragraph position="1"> If the verb is in infinitive then the agent is omitted.</Paragraph>
      <Paragraph position="2"> Its concept is obtained from the function verb.  In parsing, this demon is spawned when a verb is encountered. For example, consider the process in parsing the sentence Da.v~.d dec'ideal ~ bake on ~,o\].PSath. Through identifying the verbs and their forms, the protess is: decided (active, simple) Search for the agent before the verb, anticipate an infinitive form.</Paragraph>
      <Paragraph position="3"> talc, (active, infinitive) Do not anticipate the agent. The actor of the &amp;quot;take on&amp;quot; concept which is the agent, is extracted from the agent of &amp;quot;decide'.</Paragraph>
    </Section>
  </Section>
  <Section position="6" start_page="288" end_page="289" type="metho">
    <SectionTitle>
4. Failure-Driven Pattern Construction
</SectionTitle>
    <Paragraph position="0"> Learning of phrases in RINA is an iterative protess. The input is a sequence of sentence-context pairs, through which the program refines its current hypothesis about the new phrase. The hypothesis pertains to both the pattern and the concept of the phrase.</Paragraph>
    <Section position="1" start_page="288" end_page="288" type="sub_section">
      <SectionTitle>
4.2 The Learning Cycle
</SectionTitle>
      <Paragraph position="0"> The basic cycle in the process is:  (a) A sentence is parsed on the background of a conceptual context.</Paragraph>
      <Paragraph position="1"> (b) Using the current hypothesis, either the sentence is comprehended smoothly, or a failure is detected. (c) If a failure is detected then the current hypothesis is  updated.</Paragraph>
      <Paragraph position="2"> The crucial point in this scheme is to obtain from the parser an intelligible analysis of failures. As an example,  consider this part of the first dialog:. 1 Program: tie took on him. He von ~he fight? 2 User:. No. He took him on. Dav'\[d Lt, ta, cked him. 3 Program: He took him on.</Paragraph>
      <Paragraph position="3">  He accepted the challenge? The first hypothesis is shown in Figure 8. pattern: concept: ?x:person take:verb don ?y:person~ ?x win the conflict with ?y  Notice that the preposition on is attached to the object ?y, thus assuming that the phrase is similar to He looked at Iqaar7 which cannot produce the following sentence: H. look.d her at. This hypothesis underlies Sentence 1 which is erroneous in both its form and its meaning. Two observations should be made by comparing this pattern to Sentence 2: The object is not preceded by the preposition on. The preposition on does not precede any object. These comments direct the construction of the new hypothesis: null  where the preposition on is taken as a modifier of the verb itself, thus correctly generating Sentence 3. In Figure 9 the conceptual hypothesis is still incorrect and must itself be modified.</Paragraph>
    </Section>
    <Section position="2" start_page="288" end_page="288" type="sub_section">
      <SectionTitle>
4.3 Learning Strategies
</SectionTitle>
      <Paragraph position="0"> A subset of RINA's learning strategies, the ones used for the David and OoliaCh Dialog (Section 1.1) are described in this section. In our exposition of failures and actions we will illustrate the situations involved in the dialogues above, where each situation is specified by the following five ingredients:  (1) the input sentence (Sentence), (2) the context (not shown explicitly here), (3} the active pattern: either the pattern under construction, or the best matching pattern if this is the first sentence in the dialogue (Patternl).</Paragraph>
      <Paragraph position="1"> (4) the failures detected in the current situation (Failures), (5) the pattern resulting from the application of the ac- null tion to the current pattern (Pattern2).</Paragraph>
      <Paragraph position="2"> Creating a New Phrase A case.role mismatch occurs when the input sentence can only be partially matched by the active pattern. A 9oal mismatch occurs when the concept instantinted by the selected pattern does not match the goal situation in the context.</Paragraph>
      <Paragraph position="3">  tion fails since {1) a location is not found and (2) the action does not match David's goals. If these two failures are encountered, then a new phrase is created. In absence of a better alternative, RINA initially generates David Cook him somevhere.</Paragraph>
      <Paragraph position="4"> Discriminating a Pattern by Freezing a Prepoab tional Phrase A prepoMtional mismatch occurs when a preposition P matches in neither the active pattern nor in one of the lexical prepositional phrases, such as:</Paragraph>
    </Section>
    <Section position="3" start_page="288" end_page="288" type="sub_section">
      <SectionTitle>
Prepositional mismatch
</SectionTitle>
      <Paragraph position="0"> ?x:person take:verb &lt;on ?y:person&gt; The preposition on is not part of the active pattern. Neither does it match any of the prepositional phrases which currently exist for on. Therefore, since it cannot be interpreted in any other way, the ordering of the sub-expression &lt;on ?y,:peraoa&gt; is frozen in the larger pattern, using &lt; and &gt;.</Paragraph>
      <Paragraph position="1"> Two-word verbs present a di~culty to language learners \[Ulm75\] who tend to ignore the separated verb-particle form, generating: take on him instead of cake him o,s. In the situation above, the learner produced this typical error.</Paragraph>
      <Paragraph position="2"> Relaxing an Undergeneralized Pattern Two failures involving on: (1) case-role mismatch (on ?y:p,r6oa is not found)and (2) prepositional mismatch (on appears unmatched at the end of the sentence) are encountered in the situation below: Sentence: Patte~at: Failures: Pattern2: David took him on.</Paragraph>
      <Paragraph position="3"> ?x:person take:verb &lt;on ?y'person Prepositional and case-role mismatch.</Paragraph>
      <Paragraph position="4"> ?x:person take:verb on ?y:person The combination of these two failures indicate that the pattern is too restrictive. Therefore, the &lt; and &gt; freezing delimiters are removed, and the pattern may now account for two-word verbs. In this case on can be separated from C/,&amp;ke.</Paragraph>
    </Section>
    <Section position="4" start_page="288" end_page="289" type="sub_section">
      <SectionTitle>
Generaiising a Semantic Restriction
</SectionTitle>
      <Paragraph position="0"> A semantic mismatch is marked when the semantic class of a variable in the pattern does not subsume the class of the corresponding concept in the sentence.</Paragraph>
      <Paragraph position="1">  ?x:person take:verb on ?y:task As a result, the type of ?y in the pattern is generalized to include both cases.</Paragraph>
      <Paragraph position="2">  Freezing a Reference Which Relates to a Metaphor An unrelated reference is marked when a reference in the sentence does not relate to the context, but rather it relates to a metaphor (see elaboration in \[Zernik85\] ). The reference his fooc cannot be resolved in the context, rather it is resolved by a metaphoric gesture.  Her father put his foot down.</Paragraph>
      <Paragraph position="3"> ?x:person put:verb down ?y:phys-obj Goal mismatch and unrelated reference ?x:person put:verb down foot:body-part Since, (I) putting his foot on the floor does not match any of the goals of Jenny's father and (2) the reference his foot is related to the domain of metaphoric gestures rather than to the context. Therefore, foot becomes frozen in the pattern. This method is similar to a method suggested by Fuss and Wilks \[Fuss83\]. In their method, a metaphor is analyzed when an apparently ill-formed input is detected, e.g.: the car drank ffi lot of gas.</Paragraph>
    </Section>
    <Section position="5" start_page="289" end_page="289" type="sub_section">
      <SectionTitle>
4.4 Concept Constructor
</SectionTitle>
      <Paragraph position="0"> Each pattern has an associated concept which is specified using Dyer's \[Dyer83\] i-link notation. The concept of a new phrase is extracted from the context, which may contain more than one element. For example, in the first dialogue above, the given context contains some salient sto W points \[Wilensky82\] which are indexed in episodic memory as two violated expectations: * David won the fight in spite of Goliath's physical superiority. null * David accepted the challenge in spite of the risk involved. null The program extracts meanings from the given set of points. Concept hypothesis construction is further discussed in \[Zernik85\].</Paragraph>
      <Paragraph position="1"> 5. Previous Work in Language Learning In RINA, the stimulus for learning is comprehension failure. In previous models language learning was ,~lso driven by detection of failures.</Paragraph>
      <Paragraph position="2"> PST \[Reeker76\] learned grammar by acting upon dilfercnces detected between the input sentence and internally generated sentences. Six types of differences were classified, and the detection of a difference which belonged to a class caused the associated alteration of the grammar.</Paragraph>
      <Paragraph position="3"> FOUL-UP \[Granger771 learned meanings of single words when an unknown word was encountered. The meaning was extracted from the script \[Schank77\] which was given as the context. A typical learning situation was The cffir vas driving on Hvy 66, vhen it careened off the road. The meaning of the unknown verb care.ned was guessed from the SACCIDENT script.</Paragraph>
      <Paragraph position="4"> POLITICS \[CarbonellTO\], which modeled comprehension of text involving political concepts, initiated learning when semantic constraints were violated. Constraints were generalized by analyzing underlying metaphors.</Paragraph>
      <Paragraph position="5"> AMBER \[Langley82\] modeled learning of basic sentence structure. The process of learning was directed by mismatches between input sentences and sentences generated by the program. Learning involved recovery from both errors of omission (omitting a function word such as the or is in daddy bouncing ball) and errors of commission (producing daddy is liking dinner).</Paragraph>
      <Paragraph position="6"> Thus, some programs acquired linguistic patterns and some programs acquired meanings from context, but none of the above programs acquired new phrases. Acquisition of phrases involves two parallel processes: the formation of the pattern from the given set of example sentences, and the construction of the meaning from the context. These two processes are not independent since the construction of the conceptual meaning utilizes linguistic clues while the selection of pattern elements of new figurative phrases bears on concepts in the context.</Paragraph>
    </Section>
  </Section>
class="xml-element"></Paper>