Anaphora Resolution as Lexical Cohesion Identification

ANAPHORA RESOLUTION AS LEXICAL COHESION IDENTIFICATION Samuel W.K. Chan and Benjamin K T'sou Language Information Sciences Research Centre City University of Hong Kong, Hong Kong ABSTRACT Anaphora,. an important indicator in lexical cohesion, is a discourse level linguistic phenomenon. Most theoretical linguistic approaches to the interpretation of anaphoric expressions propose a treatment on the basis of purely syntactic information. In this article, what we proposed is to cast anaphora resolution as a semantic inference process in which combination of multiple strategies, each exploiting a different linguistic knowledge, is employed to resolve anaphora into a coherent one. We also exhibit how to embed an anaphora resolution into a framework which captures all the salient parameters as well as to remedy, to a certain extent, the inadequacies found in any monolithic resolution systems. The effectiveness of the anaphora resolution considered in this work is exemplified through a set of simulations. 1. INTRODUCTION It is realized that pieces of information in a text are highly interconnected and links between them are mainly local, i.e., linguistic items connect adjacent or near adjacent clues. One of the principal devices indicating such interconnections is anaphora. Anaphora is the linguistic phenomenon of abbreviated subsequent reference. It is also a device for referring back to an entity which has been introduced with more fully descriptive phrasing earlier in the text. The entity may be an object, a concept, an individual, or state of being. In the linguistic studies of anaphora, although some comprehensive studies have already presented and a coherent classification of anaphora has been laid, much of them is still in a theoretical basis and no much computational algorithms are released [1,13,15]. Among a few, systematic methods have been proposed for resolving anaphora, however, as in work on other syntactic processing, most anaphora resolution algorithms are conceptualized entirely within symbolic frameworks [3,10,11,9]. Syntactic information plays a central role in establishing appropriate referents, nevertheless, it has long been recognized the traditional symbolic syntactic approach which tries to capture the meaning from antecedent linguistic items in text cannot be translated into a processing theory [2]. Anaphoric relations could be hidden in the context. Interpretation of anaphora is crucial in making use of syntactic and semantic preferences of each linguistic item in sentences, moreover, it is a great necessity to incorporate previously analyzed sentences which is regarded as an important knowledge source in modelling prior specific context effect. Psycholinguistic accounts of discourse comprehension often assume anaphora resolution proceeds by search-and-match. Interpretation of an anaphora is not resolved by decoding, but inferred, by a non- demonstrative inference process – a process of hypothesis formation and evaluation – in which linguistic constituents and contextual assumptions determine the class of possible hypotheses. These hypotheses are then evaluated in the light of certain general principles which authors and readers are expected to obey. The semantic inferences during anaphora resolution are those based on easily available information and those required making statements in a text locally coherent. To acquire a correct interpretation of a sentence, …

Content truncated to 500 words