Loading...
Please wait, while we are loading the content...
Similar Documents
Dependency Parsing for Telugu Using Data-driven Parsers
| Content Provider | Semantic Scholar |
|---|---|
| Author | Gatla, Praveen |
| Copyright Year | 2019 |
| Abstract | In this paper, we have developed manually annotated Telugu corpora by following DS guidelines (2009) and experimented our Telugu dependency treebank data on the data-driven parsers like Malt (Nivre et al., 2007a) and MST (McDonald et al. 2006) for parsing Telugu sentences. In the dependency, we link the head and dependents with their dependency relations (drels) by giving kāraka and non-kāraka relations to them. Telugu annotated data contains token with their morph information, pos, chunk and the drels. We have used our final Telugu treebank data in CONLL format for parsing in malt and MST parsers. We evaluated the labeled attachment score (LAS), unlabeled attachment score (UAS) and labeled accuracy (LA) for both the parsers and also compared their score in case of dependency relation too. Finally, we evaluated the most frequent errors which occurred after parsing the sentences and explained them with relevant examples with appropriate linguistic analysis, so that we can improve the accuracy of parsers in our future research. |
| File Format | PDF HTM / HTML |
| Alternate Webpage(s) | http://languageinindia.com/jan2019/praveengatladependencyteluguparser1.pdf |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |
