|
Hindi Word Sense Disambiguation using variants of Simplified Lesk measure |
|---|---|
| รหัสดีโอไอ | |
| Creator | Satyendr Singh |
| Title | Hindi Word Sense Disambiguation using variants of Simplified Lesk measure |
| Contributor | Goldie Gabrani, Tanveer Siddiqui |
| Publisher | Sirindhorn International Institute of Technology, Bangkadi Campus (SIIT-BKD) |
| Publication Year | 2560 |
| Journal Title | Journal of Intelligent Informatics and Smart Technology |
| Journal Vol. | 2 |
| Page no. | 1-6 |
| Keyword | Hindi word sense disambiguation, Dictionary based Word Sense Disambiguation, Lesk-based Word Sense Disambiguation |
| URL Website | https://ph05.tci-thaijo.org/index.php/JIIST |
| Website title | Journal of Intelligent Informatics and Smart Technology |
| ISSN | 2586-9167 |
| Abstract | This paper evaluates the performance of a Lesk-like algorithm for Hindi Word Sense Disambiguation (WSD). The algorithm uses the similarity between the sense definition and the context of ambiguous word for disambiguation. Three different scoring functions have been investigated for measuring the similarity: direct overlap, frequency of matching words and frequency of matching words excluding the target word. We evaluate the effects of context window size, stop word elimination and stemming on Hindi WSD task. We also investigate the effect of number of senses on Hindi WSD task. The evaluation has been carried out on a manually created sense inventory consisting of 60 polysemous Hindi nouns. The maximum overall precision of 54.54% was ob-served for the case when both stemming and stop word removal was performed and frequency based scoring excluding the target word was used. The best case results in a significant improvement of 10.4% in precision and 21.3% in recall over the baseline performance. In general, we obtained decrease in precision with increasing number of senses. |
