File Information
File: 05-lr/acl_arc_1_sum/cleansed_text/xml_by_section/metho/98/p98-2204_metho.xml
Size: 23,357 bytes
Last Modified: 2025-10-06 14:14:59
<?xml version="1.0" standalone="yes"?> <Paper uid="P98-2204"> <Title>Never Look Back: An Alternative to Centering</Title> <Section position="3" start_page="0" end_page="1251" type="metho"> <SectionTitle> 2 A Look Back: Centering </SectionTitle> <Paragraph position="0"> The centering model describes the relation between the focus of attention, the choices of referring expressions, and the perceived coherence of discourse.</Paragraph> <Paragraph position="1"> The model has been motivated with evidence from preferences for the antecedents of pronouns (Grosz et al., 1983; 1995) and has been applied to pronoun resolution (Brennan et al. (1987), inter alia, whose interpretation differs from the original model).</Paragraph> <Paragraph position="2"> The centering model itself consists of two constructs, the backward-looking center and the list of forward-looking centers, and a few rules and constraints. Each utterance Ui is assigned a list of forward-looking centers, C f (Ui), and a unique backward-looking center, Cb(Ui). A ranking imposed on the elements of the Cf reflects the assumption that the most highly ranked element of C f (Ui) (the preferred center Cp(Ui)) is most likely to be the Cb(Ui+l). The most highly ranked element of Cf(Ui) that is realized in Ui+x (i.e., is associated with an expression that has a valid interpretation in the underlying semantic representation) is the Cb(Ui+l). Therefore, the ranking on the Cf plays a crucial role in the model. Grosz et al. (1995) and Brennan et al. (1987) use grammatical relations to rank the Cf (i.e., subj -.< obj -< ...) but state that other factors might also play a role.</Paragraph> <Paragraph position="4"> For their centering algorithm, Brennan et al.</Paragraph> <Paragraph position="5"> (1987, henceforth BFP-algorithm) extend the notion of centering transition relations, which hold across adjacent utterances, to differentiate types of shift (cf. Table 1 taken from Walker et al. (1994)).</Paragraph> <Paragraph position="7"/> </Section> <Section position="4" start_page="1251" end_page="1253" type="metho"> <SectionTitle> CONTINUE SMOOTH-SHIFT RETAIN ROUGH-SHIFT </SectionTitle> <Paragraph position="0"> Brennan et al. (1987) modify the second of two rules on center movement and realization which were defined by Grosz et al. (1983; 1995): Rule 1: If some element of Cf(Ui-1) is realized as a pronoun in Ui, then so is Cb(Ui).</Paragraph> <Paragraph position="1"> Rule 2&quot; Transition states are ordered. CONTINUE is preferred to RETAIN is preferred to SMOOTH-SHIFT is preferred to ROUGH-SHIFT.</Paragraph> <Paragraph position="2"> The BFP-algorithm (cf. Walker et al. (1994)) consists of three basic steps: 1. GENERATE possible Cb-Cfcombinations.</Paragraph> <Paragraph position="3"> 2. FILTER by constraints, e.g., contra-indexing, sortal predicates, centering rules and constraints. null 3. RANK by transition orderings.</Paragraph> <Paragraph position="4"> To illustrate this algorithm, we consider example (1) (Brennan et al., 1987) which has two different final utterances (ld) and (ld~). Utterance (ld) contains one pronoun, utterance (ld t) two pronouns. We look at the interpretation of (ld) and (ldt). After step 2, the algorithm has produced two readings for each variant which are rated by the corresponding transitions in step 3. In (ld), the pronoun &quot;she&quot; is resolved to &quot;her&quot; (= Brennan) because the CONTINUE transition is ranked higher than SMOOTH-SHIFT in the second reading. In (ld~), the pronoun &quot;she&quot; is resolved to &quot;Friedman&quot; because SMOOTH- null SHIFT is preferred over ROUGH-SHIFT.</Paragraph> <Paragraph position="5"> (1) a. Brennan drives an Alfa Romeo.</Paragraph> <Paragraph position="6"> b. She drives too fast.</Paragraph> <Paragraph position="7"> c. Friedman races her on weekends.</Paragraph> <Paragraph position="8"> d. She goes to Laguna Seca.</Paragraph> <Paragraph position="9"> d.' She often beats her.</Paragraph> <Paragraph position="10"> 3 An Alternative to Centering</Paragraph> <Section position="1" start_page="1251" end_page="1251" type="sub_section"> <SectionTitle> 3.1 The Model </SectionTitle> <Paragraph position="0"> The realization and the structure of my model departs significantly from the centering model: discourse entities which are realized in the current and the previous utterance.</Paragraph> <Paragraph position="1"> * The elements of the S-list are ranked according to their information status. The order among the elements provides directly the preference for the interpretation of anaphoric expressions. In contrast to the centering model, my model does not need a construct which looks back; it does not need transitions and transition ranking criteria. Instead of using the Cb to account for local coherence, in my model this is achieved by comparing the first element of the S-list with the preceding state.</Paragraph> </Section> <Section position="2" start_page="1251" end_page="1252" type="sub_section"> <SectionTitle> 3.2 S-List Ranking </SectionTitle> <Paragraph position="0"> Strube & Hahn (1996) rank the Cfaccording to the information status of discourse entities. I here generalize these ranking criteria by redefining them in Prince's (1981; 1992) terms. I distinguish between three different sets of expressions, hearer-old discourse entities (OLD), mediated discourse entities (MED), and hearer-new discourse entities (NEW).</Paragraph> <Paragraph position="1"> These sets consist of the elements of Prince's familiarity scale (Prince, 1981, p.245). OLD consists of evoked (E) and unused (U) discourse entities while NEW consists of brand-new (BN) discourse entities. MED consists of inferrables (I), containing inferrables (I c) and anchored brand-new (BN A) discourse entities. These discourse entities are discourse-new but mediated by some hearer-oM discourse entity (cf. Figure 1). I do not assume any difference between the elements of each set with respect to their information status. E.g., evoked and unused discourse entities have the same information status because both belong to OLD.</Paragraph> <Paragraph position="2"> For an operationalization of Prince's terms, I stipulate that evoked discourse entitites are co-referring expressions (pronominal and nominal anaphora, previously mentioned proper names, relative pronouns, appositives). Unused discourse entities are proper names and titles. In texts, brand-new proper names are usually accompanied by a relative clause or an appositive which relates them to the hearer's knowledge. The corresponding discourse entity is evoked only after this elaboration. Whenever these linguistic devices are missing, proper names are treated as unused I . I restrict inferrables to the particular subset defined by Hahn et al. (1996). Anchored brand-new discourse entities require that the anchor is either evoked or unused.</Paragraph> <Paragraph position="3"> I assume the following conventions for the ranking constraints on the elements of the S-list. The 3-tuple (x, uttx, posz) denotes a discourse entity x which is evoked in utterance uttz at the text position posz. With respect to any two discourse entities (x, uttz,posz) and (y, utty,pOSy), uttz and utty specifying the current utterance Ui or the preceding utterance U/_ 1, I set up the following ordering constraints on elements in the S-list (Table 2) 2 . For any state of the processor/hearer, the ordering of discourse entities in the S-list that can be derived from the ordering constraints (1) to (3) is denoted by the precedence relation --<.</Paragraph> <Paragraph position="4"> (I) If x E OLD and y E MED, then x -~ y.</Paragraph> <Paragraph position="5"> Ifx E OLD and y E NEW, then x -< y.</Paragraph> <Paragraph position="6"> lfx E MED and y E NEW, then x -< V.</Paragraph> <Paragraph position="7"> (2) If x, y E OLD, or x, v E MED, or x, y E NEW, then if uttx >- utt~, then x -< y, if uttz = utt~ and pos~ < pos~, then x -< y.</Paragraph> <Paragraph position="8"> Summarizing Table 2, I state the following preference ranking for discourse entities in Ui and Ui-l: hearer-oM discourse entities in Ui, hearer-old discourse entities in Ui-1, mediated discourse entities in Ui, mediated discourse entities in Ui-1, hearer-new discourse entities in Ui, hearer-new discourse entities in Ui-1. By making the distinction in (2) ~For examples of brand-new proper names and their introduction cf., e.g., the &quot;obituaries&quot; section of the New York Times. 2The relations >- and = indicate that the utterance containing x follows (>-) the utterance containing y or that x and y are elements of the same utterance (=).</Paragraph> <Paragraph position="9"> between discourse entities in Ui and discourse entities in Ui-1, I am able to deal with intra-sentential anaphora. There is no need for further specifications for complex sentences. A finer grained ordering is achieved by ranking discourse entities within each of the sets according to their text position.</Paragraph> </Section> <Section position="3" start_page="1252" end_page="1253" type="sub_section"> <SectionTitle> 3.3 The Algorithm </SectionTitle> <Paragraph position="0"> Anaphora resolution is performed with a simple look-up in the S-list 3. The elements of the S-list are tested in the given order until one test succeeds. Just after an anaphoric expression is resolved, the S-list is updated. The algorithm processes a text from left to fight (the unit of processing is the word): 1. If a referring expression is encountered, (a) if it is a pronoun, test the elements of the S-list in the given order until the test succeeds4; null (b) update S-list; the position of the referring expression under consideration is determined by the S-list-ranking criteria which are used as an insertion algorithm.</Paragraph> <Paragraph position="1"> 2. If the analysis of utterance U 5 is finished, re- null move all discourse entities from the S-list, which are not realized in U.</Paragraph> <Paragraph position="2"> The analysis for example (1) is given in Table 3 6. I show only these steps which are of interest for the computation of the S-list and the pronoun resolution. The preferences for pronouns (in bold font) are given by the S-list immediately above them. The pronoun &quot;she&quot; in (lb) is resolved to the first element of the S-list. When the pronoun &quot;her&quot; in (lc) is encountered, FRIEDMAN is the first element of the S-list since FRIEDMAN is unused and in the current utterance. Because of binding restrictions, &quot;her&quot; cannot be resolved to FRIEDMAN but tO the second element, BRENNAN. In both (ld) and (ld ~) the pronoun &quot;she&quot; is resolved to FRIEDMAN. The difference between my algorithm and the BFP-algorithm becomes clearer when the unused discourse entity &quot;Friedman&quot; is replaced by a brand-new discourse entity, e.g., &quot;a professional driver ''7 (cf. example (2)). In the BFP-algorithm, the ranking of the Cf-list depends on grammatical roles. Hence, DRIVER is ranked higher than BRENNAN in the Cf(2c). In (2d), the pronoun &quot;she&quot; is resolved to BRENNAN because of the preference for CONTINUE over RETAIN. In (2d~), &quot;she&quot; is resolved to DRIVER because SMOOTH-SHIFT is preferred over ROUGH-SHIFT. In my algorithm, at the end of (2c) the evoked phrase &quot;her&quot; is ranked higher than the brand-new phrase &quot;a professional driver&quot; (cf. Table 4). In both (2d) and (2d ~) the pronoun &quot;she&quot; is resolved to BRENNAN.</Paragraph> <Paragraph position="3"> (2) a. Brennan drives an Alfa Romeo.</Paragraph> <Paragraph position="4"> b. She drives too fast.</Paragraph> <Paragraph position="5"> c. A professional driver races her on weekends.</Paragraph> <Paragraph position="6"> d. She goes to Laguna Seca.</Paragraph> <Paragraph position="7"> d/ She often beats her.</Paragraph> <Paragraph position="8"> Example (3) 8 illustrates how the preferences for intra- and inter-sentential anaphora interact with the information status of discourse entitites (Table 5). Sentence (3a) starts a new discourse segment. The phrase &quot;a judge&quot; is brand-new. &quot;Mr. Curtis&quot; is mentioned several times before in the text, Hence, 7I owe this variant Andrew Kehler. -This example can misdirect readers because the phrase &quot;'a professional driver&quot; is assigned the &quot;default&quot; gender masculine. Anyway, this example - like the original example - seems not to be felicitous English and has only illustrative character.</Paragraph> <Paragraph position="9"> Sin: The New York Times. Dec. 7, 1997, p.A48 (&quot;Shot in head, suspect goes free, then to college&quot;).</Paragraph> <Paragraph position="10"> the discourse entity CURTIS is evoked and ranked higher than the discourse entity JUDGE. In the next step, the ellipsis refers to JUDGE which is evoked then. The nouns &quot;request&quot; and &quot;prosecutors&quot; are brand-new 9. The pronoun &quot;he&quot; and the possessive pronoun &quot;his&quot; are resolved to CURTIS. &quot;Condition&quot; is brand-new but anchored by the possessive pronoun. For (3b) and (3c) I show only the steps immediately before the pronouns are resolved. In (3b) both &quot;Mr. Curtis&quot; and &quot;the judge&quot; are evoked. However, &quot;Mr. Curtis&quot; is the left-most evoked phrase in this sentence and therefore the most preferred antecedent for the pronoun &quot;him&quot;. For my experiments I restricted the length of the S-list to five elements. Therefore &quot;prosecutors&quot; in (3b) is not contained in the S-list. The discourse entity SMIRGA is introduced in (3c). It becomes evoked after the appositive. Hence SM1RGA is the most preferred antecedent for the pronoun &quot;he&quot;. (3) a. A judge ordered that Mr. Curtis be released, but e agreed with a request from prosecutors that he be re-examined each year to see if his condition has improved.</Paragraph> <Paragraph position="11"> b. But authorities lost contact with Mr. Curtis after the Connecticut Supreme Court ruled in 1990 that the judge had erred, and that prosecutors had no right to re-examine him.</Paragraph> <Paragraph position="12"> c. John Smirga, the assistant state's attorney in charge of the original case, said last week that he always had doubts about the psychiatric reports that said Mr. Curtis would never improve.</Paragraph> <Paragraph position="13"> 9I restrict inferrables to the cases specified by Hahn et al. (1996). Therefore &quot;prosecutors&quot; is brand-new (cf. Prince (1992) for a discussion of the form of inferrables).</Paragraph> <Paragraph position="14"> 1254 (3a) A judge S: \[JUDGEBN: judge\] ordered that Mr. Curtis S: \[CURTISE: Mr. Curtis, JUDGEBN: judge\] be released, but e S: \[CURTISE: Mr. Curtis, JUDGEE: e\] agreed with a request S: \[CURTISE: Mr. Curtis, JUDGEE: e, REQUESTBN: request\] from prosecutors S: \[CURTISE: Mr. Curtis, JUDGEE: e, REQUESTBN: request, PROSECUTORSBN: prosecutors\] that he S: \[CURTISE: he, JUDGEE: e, REQUESTBN: request, PROSECUTORSBN: prosecutors\] be re-examined each year S: \[CURTISE: he, JUDGEE: ~, REQUESTBN: request, PROSECUTORSBN: prosecutors, YEARBN: year\] to see if his S: \[CURTISE: his, JUDGEE: ~, REQUESTBN: request, PROSECUTORSBN: prosecutors, YEARBN: year\] condition S: \[CURTISE: his, JUDGEE: e, CONDITIONBNA : condition, REQUESTBN: request, PROSECUTORSBN: prosec.\] has improved.</Paragraph> <Paragraph position="15"> S: \[CURTISE: his, JUDGEE: e, CONDITIONBNA: condition, REQUESTBN: request, PROSECUTORSBN: prosec.\] (3b) But authorities lost contact with Mr. Curtis after the Connecticut Supreme Court ruled in 1990 that the judge had erred, and that prosecutors had no right S: \[CURTISE: his, CS COURTu: CS Court, JUDGEE: judge, CONDITIONBNA: condition, AUTH.BN: auth.\] to re-examine him.</Paragraph> <Paragraph position="16"> S: \[CURTISE: him, CS COURTu: CS Court, JUDGEE: judge, CONDITIONBNA: condition, AUTH.BN: auth.\] (3c) John Smirga, the assistant state's attorney in charge of the original case, said last week S: \[SMIRGAE: attorney, CASEE: case, CURTISE: him, CS COURTu: CS Court, JUDGEE: judge \] that he had doubts about the psychiatric reports that said Mr. Curtis would never improve. S: \[SMIRGAE: he, CASEE: case, REPORTSE: reports, CURTISE: Mr. Curtis, DOUBTSBN: doubts\]</Paragraph> </Section> </Section> <Section position="5" start_page="1253" end_page="1255" type="metho"> <SectionTitle> 4 Some Empirical Dat:i </SectionTitle> <Paragraph position="0"> In the first experiment, I compare my algorithm with the BFP-algorithm which was in a second experiment extended by the constraints for complex sentences as described by Kameyama (1998).</Paragraph> <Paragraph position="1"> Method. I use the following guidelines for the hand-simulated analysis (Walker, 1989). I do not assume any world knowledge as part of the anaphora resolution process. Only agreement criteria, binding and sortal constraints are applied. I do not account for false positives and error chains. Following Walker (1989), a segment is defined as a paragraph unless its first sentence has a pronoun in subject position or a pronoun where none of the preceding sentence-internal noun phrases matches its syntactic features. At the beginning of a segment, anaphora resolution is preferentially performed within the same utterance. My algorithm starts with an empty S-list at the beginning of a segment.</Paragraph> <Paragraph position="2"> The basic unit for which the centering data structures are generated is the utterance U. For the BFPalgorithm, I define U as a simple sentence, a complex sentence, or each full clause of a compound sentence. Kameyama's (1998) intra-sentential centering operates at the clause level. While tensed clauses are defined as utterances on their own, untensed clauses are processed with the main clause, so that the Cf-list of the main clause contains the elements of the untensed embedded clause.</Paragraph> <Paragraph position="3"> Kameyama distinguishes for tensed clauses further between sequential and hierarchical centering. Except for reported speech (embedded and inaccessible to the superordinate level), non-report complements, and relative clauses (both embedded but accessible to the superordinate level; less salient than the higher levels), all other types of tensed clauses build a chain of utterances on the same level.</Paragraph> <Paragraph position="4"> According to the preference for inter-sentential candidates in the centering model, I define the following anaphora resolution strategy for the BFPalgorithm: (1) Test elements of Ui-1. (2) Test elements of Ui left-to-right. (3) Test elements of Cf(Ui-2), Cf(Ui-3) .... In my algorithm steps (1) and (2) fall together. (3) is performed using previous states of the system.</Paragraph> <Paragraph position="5"> Results. The test set consisted of the beginnings of three short stories by Hemingway (2785 words, 153 sentences) and three articles from the New York Times (4546 words, 233 sentences). The resuits of my experiments are given in Table 6. The first row gives the number of personal and possessive pronouns. The remainder of the Table shows the results for the BFP-algorithm, for the BFP-algorithm extended by Kameyama's intra-sentential specifications, and for my algorithm. The overall error rate of each approach is given in the rows marked with wrong. The rows marked with wrong (strat.) give the numbers of errors directly produced by the algorithms' strategy, the rows marked with wrong (ambig.) the number of analyses with ambiguities generated by the BFP-algorithm (my approach does not generate ambiguities). The rows marked with wrong (intra) give the number of errors caused by (missing) specifications for intra-sentential anaphora. Since my algorithm integrates the specifications for intra-sentential anaphora, I count these errors as strategic errors. The rows marked with wrong (chain) give the numbers of errors contained in error chains. The rows marked with wrong (other) give the numbers of the remaining errors (consisting of pronouns with split antecedents, errors because of segment boundaries, and missing specifications for event anaphora).</Paragraph> <Paragraph position="6"> Hem. NYT Pron. and Poss. Pron. 274 302 BFP-Algo.</Paragraph> <Paragraph position="7"> BFP/Kam.</Paragraph> <Paragraph position="8"> Interpretation. The results of my experiments showed not only that my algorithm performed better than the centering approaches but also revealed insight in the interaction between inter- and intra-sentential preferences for anaphoric antecedents. Kameyama's specifications reduce the complexity in that the Cf-lists in general are shorter after splitting up a sentence into clauses. Therefore, the BFP-algorithm combined with her specifications has almost no strategic errors while the number of ambiguities remains constant. But this benefit is achieved at the expense of more errors caused by the intra-sentential specifications. These errors occur in cases like example (3), in which Kameyama's intra-sentential strategy makes the correct antecedent less salient, indicating that a clause-based approach is too fine-grained and that the hierarchical syntactical structure as assumed by Kameyama does not have a great impact on anaphora resolution.</Paragraph> <Paragraph position="9"> I noted, too, that the BFP-algorithm can generate ambiguous readings for Ui when the pronoun in Ui does not co-specify the Cb(Ui-1). In cases, where the Cf(Ui-1) contains more than one possible antecedent for the pronoun, several ambiguous readings with the same transitions are generated.</Paragraph> <Paragraph position="10"> An exampleldeg: There is no Cb(4a) because no element of the preceding utterance is realized in (4a). The pronoun &quot;them&quot; in (4b) co-specifies &quot;deer&quot; but the BFP-algorithm generates two readings both of which are marked by a RETAIN transition.</Paragraph> <Paragraph position="11"> (4) a. Jim pulled the burlap sacks off the deer b. and Liz looked at them.</Paragraph> <Paragraph position="12"> In general, the strength of the centering model is that it is possible to use the Cb(Ui-t) as the most preferred antecedent for a pronoun in Ui. In my model this effect is achieved by the preference for hearer-old discourse entities. Whenever this preference is misleading both approaches give wrong results. Since the Cb is defined strictly local while hearer-old discourse entities are defined global, my model produces less errors. In my model the preference is available immediately while the BFP-algorithm can use its preference not before the second utterance has been processed. The more global definition of hearer-old discourse entities leads also to shorter error chains. - However, the test set is too small to draw final conclusions, but at least for the texts analyzed the preference for hearer-old discourse entities is more appropriate than the preference given by the BFP- algorithm.</Paragraph> </Section> class="xml-element"></Paper>