Sudha Bhingardive

Unsupervised Most Frequent Sense Detection using Word Embeddings

An acid test for any new Word Sense Disambiguation (WSD) algorithm is its performance against the Most Frequent Sense (MFS). The field of WSD has found the MFS baseline very hard to beat. Clearly, if WSD researchers had access to MFS values, their striving to better this heuristic will push the WSD frontier. However, getting MFS values requires sense annotated corpus in enormous amounts, which is out of bounds for most languages, even if their WordNets are available. In this paper, we propose an unsupervised method for MFS detection from the untagged corpora, which exploits word embeddings. We compare the word embedding of a word with all its sense embeddings and obtain the predominant sense with the highest similarity. We observe significant performance gain for Hindi WSD over the WordNet First Sense (WFS) baseline. As for English, the SemCor baseline is bettered for those words whose frequency is greater than 2. Our approach is language and domain independent.

Word Sense Disambiguation Using IndoWordNet

Word sense disambiguation (WSD) is considered as one of the toughest problems in the field of natural language processing. IndoWordNet is a linked structure of WordNets of major Indian languages. Recently, several IndoWordNet-based WSD approaches have been proposed and implemented for Indian languages. In this chapter, we present the usage of various other features of IndoWordNet in performing WSD. Here, we have used features such as linked WordNets and lexico-semantic relations. We have followed two unsupervised approaches, viz. (1) use of IndoWordNet in bilingual WSD for finding the sense distribution with the help of expectation maximization algorithm and (2) use of IndoWordNet in WSD for finding the most frequent sense using word and sense embeddings. Both these approaches justify the importance of IndoWordNet for word sense disambiguation for Indian languages, as the results are found to be promising and can beat the baselines.