Anaphora Resolution with Word Sense Disambiguation

Anaphora Resolution with Word Sense Disambiguation J udita Preiss* Computer Laboratory J J Thomson A venue Cambridge CB3 OFD United Kingdom Judita.Preiss@cl.cam.ac.uk Abstract We describe a simple word sense disambiguation system equipped with the Kennedy and Bogu- raev (1996) anaphora resolution algorithm, evaluated on the SENSEVAL-2 English all-words task. The system relies on the structure of the WordNet hierarchy to pick optimal senses for nouns in the text. Since anaphoric refer- ences are known to indicate the topic of the text (Boguraev et al., 1998), they may aid disam- biguation. 1 Introduction We investigate the effect of repeating pronom- inalized nouns in the input to our Word Sense Disambiguation (WSD) algorithm (Preiss, 2001). The WSD algorithm is based on the WordNet 1.7 hierarchy (Miller et al., 1990), and assigns (WordNet) senses to all nouns. The en- riched version we evaluate in this paper makes use of our re-implementation of an anaphora resolution algorithm of Kennedy and Boguraev (1996). If, as claimed by Boguraev et al. (1998), the topic of the discourse is thus repeated, then the main topic words will be more likely to be dis- ambiguated correctly. The subsequent WSD al- gorithm makes use of this extra topic informa- tion, and this will in turn affect the disambigua- tion of all other nouns in the discourse. The system is evaluated on the English all- words task in SENSEVAL-2. 2 2.1 Algorithms Overview of the Algorithm Our WSD algorithm has three components, as depicted in Figure 1. Taking as input the • This work was supported by the EPSRC while the author was at the University of Sheffield. test data parsed using the Briscoe and Car- roll (1993) parser (which uses the grammar de- scribed in Carroll and Briscoe (1996) ), the first step is to identify and discard the pleonastic pronouns. Our pleonastic component is de- scribed in section 2.2. In the next phase (section 2.3), third person pronouns are resolved to a noun antecedent and replaced in the text by the noun antecedent. The purpose of this is to increase the number of topic words in the text, to aid the disam- biguation of other nouns. This approach as- sumes firstly that pronouns refer mainly to topic words, and secondly that repeating topic words in the text helps overall disambiguation. The final phase of the algorithm is the WSD component, described in section 2.4. Using sim- ulated annealing, it attempts to find a sense assignment for every noun that minimizes an overall 'distance' function using the WordNet hierarchy. In addition, for the repeated nouns added in the previous phase, the senses are tied together. This means that if the sense of one word in a tie is changed during simulated an- nealing, the sense of all words in the tie are si- multaneously changed. The advantage of this approach can be shown on the following discourse: The parrot, like the chicken, is kept by people as a domesticated bird. It can speak. Suppose firstly that there is no …

Content truncated to 500 words