A gloss-centered algorithm for disambiguation

A Gloss-centered Algorithm for Disambiguation Ganesh Ramakrishnan Dept. of C.S.E IIT Bombay India - 400076 B. Prithviraj Dept. of C.S.E IIT Bombay India - 400076 � hare,prithvir,pb � @cse.iitb.ac.in Pushpak Bhattacharyya Dept. of C.S.E IIT Bombay India - 400076 Abstract The task of word sense disambiguation is to assign a sense label to a word in a passage. We report our algorithms and experiments for the two tasks that we participated in viz. the task of WSD of Word- Net glosses and the task of WSD of English lexical sample. For both the tasks, we explore a method of sense disambiguation through a process of “compar- ing” the current context for a word against a reposi- tory of contextual clues or glosses for each sense of each word. We compile these glosses in two differ- ent ways for the two tasks. For the first task, these glosses are all compiled using WordNet and are of various types viz. hypernymy glosses, holonymy mixture, descriptive glosses and some hybrid mix- tures of these glosses. The “comparison” could be done in a variety of ways that could include/exclude stemming, expansion of one gloss type with another gloss type, etc. The results show that the system does best when stemming is used and glosses are expanded. However, it appears that the evidence for word-senses ,accumulated through WordNet, in the form of glosses, are quite sparse. Generating dense glosses for all WordNet senses requires a massive sense tagged corpus - which is currently unavail- able. Hence, as part of the English lexical sample task, we try the same approach on densely popu- lated glosses accumulated from the training data for this task. 1 Introduction The main idea behind our approach for both the WSD tasks is to use the context of a word along with the gloss or description of each of its senses to find its correct sense. The similarity between the context and each sense of the word is measured and the word-sense with the highest similarity measure is picked as most appropriate, that with second high- est similarity is ranked second and so on. Glosses have been used by authors in the past for WSD (Lesk, 1986). The novelty in our approach, for the task of disambiguation of extended Word- Net is in the way we generate our descriptions or glosses. Also, an additional novelty in the sec- ond task, is in our use of textual proximity between words in the neighborhood of the word to be disam- biguated and the words in the glosses of each of its senses. 2 Glosses 2.1 Descriptive glosses A word, with its associated part of speech and an associated sense number, has a description. Description for fifth noun sense of “tape” memory device consisting of a long thin plastic strip coated with iron oxide; used to record audio or video signals or to store computer information Figure 1: An example descriptive-gloss for “tape” from WordNet We call these descriptions - descriptive glosses. For word-senses picked …

Content truncated to 500 words