作者: Qaiser Abbas
DOI: 10.3115/V1/W14-4205
关键词:
摘要: In this paper, the development and evaluation of Urdu parser is presented along with comparison existing resources for language variants Urdu/Hindi. This was given a linguistically rich grammar extracted from treebank. context free sufficient encoded information comparable state art parsing requirements morphologically closely related The extended model together provide us promising results both variants. gives 87% f-score, which outperforms multi-path shift-reduce simple Hindi dependency 4.8% 22% increase in recall, respectively.