Branco Anaphora: Analysis, Algorithms and Applications

6th Discourse Anaphora and Anaphor Resolution Colloquium, DAARC 2007, Lagos Portugal, March 29-30, 2007, Selected Papers
2007
ISBN: 978-3-540-71412-5
Verlag: Springer
Format: PDF
Kopierschutz: 1 - PDF Watermark

E-Book, Englisch, Band 4410, 202 Seiten, eBook

Reihe: Lecture Notes in Computer Science

This book constitutes revised selected papers of the 6th Discourse Anaphora and Anaphor Resolution Colloquium, DAARC 2007, held in Lagos, Portugal in March 2007.

The 13 revised full papers presented were carefully reviewed and selected from 60 initial submissions during two rounds of reviewing and improvements. The papers are organized in topical sections on human processing and performance, language analysis and representation, resolution methodology and algorithms, as well as computational systems and applications.

Keywords: NLP, algorithmic learning, automatic abstracting, computational linguistics, document analysis, human language technologies, information extraction, knowledge processing, machine learning, machine translation, natural language processing, natural language programming, parsing, query processing, question answering, semantic analysis, speech processing, text classification, text processing

Zielgruppe

Research

Autoren/Hrsg.

Branco, António

Inhaltsverzeichnis

Human Processing and Performance.- Nuclear Accent Placement and Other Prosodic Parameters as Cues to Pronoun Resolution.- Empirically Assessing Effects of the Right Frontier Constraint.- Pronoun Resolution and the Influence of Syntactic and Semantic Information on Discourse Prominence.- Language Analysis and Representation.- Anaphora Resolution as Equality by Default.- Null Subjects Are Reflexives, Not Pronouns.- Using Very Large Parsed Corpora and Judgment Data to Classify Verb Reflexivity.- An Empirical Investigation of the Relation Between Coreference and Quotations: Can a Pronoun Located in Quotations Find Its Referent?.- Resolution Methodology and Algorithms.- Applying Backpropagation Networks to Anaphor Resolution.- Improving Coreference Resolution Using Bridging Reference Resolution and Automatically Acquired Synonyms.- Evaluating Hybrid Versus Data-Driven Coreference Resolution.- Computational Systems and Applications.- Automatic Anaphora Resolution for Norwegian (ARN).- “Who Are We Talking About?” Tracking the Referent in a Question Answering Series.- Anaphora Resolution: To What Extent Does It Help NLP Applications?.

Leseproben

Empirically Assessing Effects of the Right Frontier Constraint (p. 15-16)

Anke Holler 1 and Lisa Irmen 2
1 University of Heidelberg Institute of General and Applied Linguistics
2 Institute of Psychology

Abstract. In a questionnaire study the effects of discourse structural information on resolving inter-sentential anaphora were investigated. The Right Frontier Constraint, first proposed by Polanyi (1988), states that potential antecedents of an anaphor that are placed at the right frontier of a discourse unit can be accessed more easily than antecedents that are placed somewhere else. Participants (N=36) received written experimental passages of six lines each that contained a pronominal anaphor in the last line and two potential antecedents in the preceding text, one introduced in the first, one in the fourth line of a passage. Antecedents' relative position to the right frontier was manipulated through the discourse relation between the first and the second antecedent and through the filler information interposed between the second antecedent and the anaphor. The two potential antecedents either had the same or different grammatical gender. In the latter case only the first antecedent was gender congruent to the anaphor. Participants' task was to name the anaphor's antecedent. Results show that in case of unequal gender antecedents, participants almost always chose the gender congruent first antecedent, irrespective of its position relative to the Right Frontier. In case of equal gender antecedents choice patterns point to an influence of an antecedent's position relative to the Right Frontier. Alternative theoretical approaches such as centering theory or situational models cannot account for the found results. The findings in the same gender antecedent condition are therefore interpreted as an effect of the Right Frontier Constraint.

1 Introduction

It is a well-known fact that the resolution of anaphora is made up of many processes operating at different linguistic and non-linguistic levels. Even if we confine ourselves to looking at linguistic processes, there are various factors influencing the accessibility of information. Substantial empirical research has shown that phonologic and morpho-syntactic as well as semantic and pragmatic information guides the way an anaphor may find its antecedent. Even subtle changes in the grammatical form of sentences may influence anaphor resolution as Klin, Guzmán, Weingartner and Ralano (2006) have recently shown. Whereas one thread of research concentrates on aspects of the complexity of the anaphor (see for instance Ariel (2001)), another strand of research deals with the question of which properties of an anaphor’s potential antecedents affect their salience. Our work addresses this latter aspect.

There is ample evidence that morpho-syntactic information such as gender or number congruency disambiguates the relation between an anaphor and its antecedent. In addition, it has been argued that anaphor resolution is subject to certain semantic inferences that also help to clearly determine the accessible antecedents of an anaphor. But in all cases where this kind of information is not available or does not vary with regard to a set of potential antecedents other linguistic factors come into play. Although a diversity of criteria for the selection of the antecedent of an anaphor has been provided depending on the respective theoretical background, there is wide agreement that the following factors are relevant for anaphor resolution: First, the syntagmatic distance between anaphor and antecedent, which is also known as the recency effect. Second, the grammatical function (or obliqueness), which on the one hand condensed in a subject assignment strategy, whereupon ambiguous pronouns will be assigned to antecedents which function as subjects. On the other hand, the parallel function assignment strategy refers to this factor as it proposes that pronouns will be assigned to antecedents with identical grammatical functions. The latter strategy is also semantically reformulated in terms of thematic role assignment. Third, further semantic aspects such as whether the potential antecedent is animate or not, and whether it functions as a topic or not. Fourth, the information-structural status of a potential antecedent in terms of providing new or familiar information seems to be a relevant factor. See Garnham (2001) for an overview of factors usually accepted as influencing anaphor resolution. These factors determining the salience of nominal antecedents of anaphora have been subject to a variety of psycholinguistic research. Several empirical studies aimed at investigating how an appropriate antecedent is assigned to potentially ambiguous pronouns during interpretation. In the course of this research it is generally agreed that the resolution of anaphora depends on what entities are currently in the focus of attention, e.g. Gordon, Grosz and Gilliom (1993), Hudson-d’Zmura and Tanenhaus (1998). Although it has been shown, that all of the aforementioned factors are certainly involved in establishing the preferred antecedent, it is still an open question inasmuch further factors have to be considered. To the best of our knowledge, the influence of discourse relations on the salience of potential antecedents of anaphora has not been in focus of psycholinguistics, yet.

The study presented here investigates inasmuch discourse structural information affects the way inter-sentential anaphora are resolved. It aims at an empirical verification of the so-called Right Frontier Constraint first proposed by Polanyi (1988). Following this constraint, it is hypothesized that potential antecedents of an anaphor that are placed at the right frontier of a discourse unit can be accessed more easily than antecedents being placed somewhere else. In other words, readers are more likely to resolve anaphora that are perceived as discourse-structurally salient. We examined this hypothesis in a questionnaire study.