Volta

Naïve Bayes classifier for Hindi word sense disambiguation

This paper investigates Naïve Bayes (NB) classifier for Hindi Word Sense Disambiguation (WSD) utilizing eleven features. The features used in the experiment includes local context, collocations, unordered list of words, nouns and vibhaktis. Evaluation is done on a manually created sense annotated Hindi corpus consisting of 60 polysemous Hindi nouns. A precision of 77.52% was observed, using unordered list of words in feature vector. We obtained maximum precision of 86.11% by utilizing nouns in feature vector after applying morphology. The experimental results demonstrates that by adding more rich features, WSD accuracy of NB classifier can be significantly improved over unordered list of words. We obtained precision of 56.49%, by utilizing vibhaktis in the feature vector.

Reference & gravity metrics

Secondary-market trade history

References (19)

Cited by (1)