Agenda Anaphor Resolution at the Right Frontier - Is the RFC effective? Research question 1. Theoretical Background 2. Method 3. Anke Holler Results 4. in collaboration with Lisa Irmen Discussion 5. University of Heidelberg Holler CoGETI-Workshop, 24.11.06 2 Research question Previous Results Does discourse-relational information affect Various linguistic factors influence accessibility intersentential anaphor resolution? of information Empirical assessment of Right Frontier Constraint (RFC) Substantial empirical research has shown that Polanyi (1988); Asher (1993); Asher & Lascarides (2003) phonologic and morpho-syntactic as well as Hypothesis: semantic and pragmatic information guides the way an anaphor may find its antecedent: Readers are more likely to resolve anaphora to antecedents that are perceived as discourse-structurally salient. Aspects of the complexity of the anaphor 1. (cf. Ariel, 2001) Questionnaire-based experiment Properties of an anaphor’s potential antecedents 2. that affect their salience Holler CoGETI-Workshop, 24.11.06 3 Holler CoGETI-Workshop, 24.11.06 4 Linguistic factors for salience Previous psycholinguistic research Morpho-syntactic information Resolution of anaphora depends on what gender entities are currently in the focus of attention, number congruency cf. Gordon, Grosz & Gilliom (1993), Hudson-d’Zmura & Certain semantic inferences Tanenhaus (1998) Recency effect: = Syntagmatic distance between anaphor and antecedent Grammatical function (or obliqueness) The influence of discourse relations on the preference for subjects salience of potential antecedents of anaphora parallel function assignment strategy has not been in focus. Further semantic aspects animateness topicality Information structure: new vs. familiar information Holler CoGETI-Workshop, 24.11.06 5 Holler CoGETI-Workshop, 24.11.06 6 1
SDRT – Fundamental assumption SDRT – Formal setting Discourse consists of a set of discourse units, which A discourse structure or SDRS is a triple < A , F , LAST>, where: are connected by two sorts of discourse relations. A is a set of labels; Subordinating relations: LAST is a label in A; and F is a function which assigns each member of A a one constituent discourse unit dominates another e.g. Elaboration, Explanation ( Asher & Vieu, 2005) member of Φ , which is the set of well-formed SDRS- formulae. Coordinating relations: An SDRS can be converted into a graph. no constituent discourse unit dominates another Each subordinating relation creates a downward edge. e.g. Narration, Contrast ( Asher & Vieu, 2005) Each coordinating relation creates a horizontal edge. Holler CoGETI-Workshop, 24.11.06 7 Holler CoGETI-Workshop, 24.11.06 8 Right-frontier rule for attachment Example π 1 : Hans hatte eine schlaflose Nacht, New information must either attach to the last Hans had a sleepless night, entered constituent β in a discourse structure or π 2 : denn sein Computer j hatte ernsthafte Probleme. to some constituent γ such that ( β , γ ) is in the because his computer had severe problems. π 3 : Die Software k stürzte dauernd ab. transitive closure of the subordination relation. The software permanently crashed. (Asher, 1993) π 4 : Die Taste für den Buchstaben A war kaputt. The key of letter A was defect. π 5 : Das Display i flackerte. The antecedent for an anaphoric expression is The display jittered. π 6 : Es i muss dringend repariert werden. accessible only at the right hand side of any It must be repaired urgently. level of a linearly ordered discourse parse tree. π 6 ’: Er j muss dringend repariert werden. It must be repaired urgently. π 6 ’’: # Sie k muss neu installiert werden. It must be newly installed. Holler CoGETI-Workshop, 24.11.06 9 Holler CoGETI-Workshop, 24.11.06 10 Example – Graph Method – Materials Experimental passages – General structure π 1 Six lines Explanation Pronominal anaphor in the last line Two potential antecedents in the preceding text, one in the first, one in the fourth line (‚the computer‘) π 2 Relative position of antecedents to Right Elaboration Frontier (RF) Discourse relation between first and second (‚the software‘) (‚the display‘) π 3 π 4 π 5 antecedent Narration Narration Filler information interposed between the second antecedent and the anaphor Holler CoGETI-Workshop, 24.11.06 11 Holler CoGETI-Workshop, 24.11.06 12 2
Am Morgen ging die Studentin in die Universität ( π 1 ), denn es Method – Materials war mal wieder an der Zeit, die Vorlesung über die Vor- und Nachteile von Kants Kategorischem Imperativ zu besuchen. ( π 2 ) Im Hörsaal war es sehr voll. ( π 3 ) Three types of items In the morning the student went to the university because it was time to Type A: only 1st antecedent at RF attend the lecture on advantages and disadvantages of Kant’s categorical imperative. The lecture hall was busy. Type B: only 2nd antecedent at RF Type C: both antecedents at RF A Die Kommilitonin war wie immer schlecht gelaunt ( π 4 ), und es hörte niemand zu. ( π 5 ) The fellow student was as always in a bad mood and nobody listened. Each structural type in two versions B Die Kommilitonin war stattdessen in der Bibliothek ( π 4 ), denn dort Both antecedents and pronoun of the same war es sehr ruhig. ( π 5 ) The fellow student however was in the library because it was quiet there. grammatical gender C Die Kommilitonin war wie immer schlecht gelaunt ( π 4 ), denn es Antecedents of different gender hörte niemand zu. ( π 5 ) The fellow student was as always in a bad mood + pronoun only congruent to first antecedent because nobody listened. Nachmittags musste sie noch viel erledigen. In the afternoon she still had many things to do. Holler CoGETI-Workshop, 24.11.06 13 Holler CoGETI-Workshop, 24.11.06 14 Method – Materials Method – Materials Type A: First antecedent at RF Type B: Second antecedent at RF π 1 π 4 π 1 Contrast Explanation Elaboration (‚however‘) Explanation Explanation Elaboration (‚because‘) π 2 π 3 π 4 π 5 π 2 π 3 π 5 Narration Narration (‚and‘) Holler CoGETI-Workshop, 24.11.06 15 Holler CoGETI-Workshop, 24.11.06 16 Method – Materials Method – Procedure Questionnaire containing 18 experimental passages Type C: Both antecedents at RF six of each type three items with equal gender antecedents three with unequal gender antecedents (only the first π 1 antecedent was gender congruent) Explanation Elaboration Participants’ task to name the antecedent of the π 2 pronominal anaphor π 3 π 4 Narration In the afternoon she still had many things to do Explanation Rephrasing question: (‚because‘) π 5 Who was the one who had to do many things? Holler CoGETI-Workshop, 24.11.06 17 Holler CoGETI-Workshop, 24.11.06 18 3
Predictions Results – RFC Table 1. Participants’ choices in case of antecedents with unequal gender, If RF constrains anaphor resolution in the separated by type of item outlined way participants should Choices 1. 2. tend to choose the 1st antecedent in Type A Antecedent Antecedent Total tend to choose the 2nd antecedent in Type B Type A 110 2 112 (1. antecedent at RF) No such difference is expected if other factors Type B 111 2 113 (2. antecedent at RF) are more influential. Type C 110 3 113 (both antecedents at RF) Recency favors the 2nd antecedent in all types of passages. Total 331 7 338 Gender congruency favors the 1st antecedent in case of unequal gender antecedents . ( χ 2 (2) < 1) Holler CoGETI-Workshop, 24.11.06 19 Holler CoGETI-Workshop, 24.11.06 20 Results – RFC Results – Alternative Accounts Table 2. Participants’ choices in case of equal gender antecedents, separated by type of item. Structural Parallelism as primary influence on Choices 1. 2. ambiguous pronoun resolution (Chambers & Smyth, Antecedent Antecedent Total 1998) Type A 72 38 110 (1. antecedent at RF) Subject role for both antecedents and pronoun Type B 51 56 107 (2. antecedent at RF) in 16 out of 18 items Type C 70 41 111 (both antecedents at RF) Total 193 135 328 Experimental materials largely parallel in syntactic structure ( χ 2 (2) = 8.323, p = .016) Holler CoGETI-Workshop, 24.11.06 21 Holler CoGETI-Workshop, 24.11.06 22 Results – Alternative Accounts Results – Alternative Accounts Changes only in Versions B Situation models as moderators of antecedent → discourse units of antecedent 1 and 2 in coordinate accessibility (Anderson, Garrod & Sanford, 1983; Morrow, discourse relation Greenspan & Bower, 1987) 7 items constant, 7 changing (Antecedent 1 is in the lecture hall, antecedent 2 is in the library...), 4 items in Local character’s accessibility as antecedents between (e.g. antecedent 2 just leaving) declines after substantial contextual changes (e.g. in space or time ) Analyses with items with unambiguous change or constancy Holler CoGETI-Workshop, 24.11.06 23 Holler CoGETI-Workshop, 24.11.06 24 4
Recommend
More recommend
