Rule based dependency parser for Telugu

Abstract

Abstract newlineParsing natural languages has been gaining popularity in recent years and newlineattracted the interest of Natural Language Processing (NLP) researchers around newlinethe world. It is challenging when the language under study is a free-word order newlinelanguage and morphologically rich like Telugu, the south-central Dravidian newlinelanguage. Parsing refers to the process of syntactic analysis of a specific language newlinetext. A parser is an automated tool that dissects sentences to provide newlinesyntactic/syntactico-semantic analysis of relations of words in a sentence. Parsing newlineis useful in the downstream analysis and applications of NLP such as machine newlinetranslation, document classification, dialogue modelling, etc.., newlineThis study adopts a knowledge-driven approach, i.e. a rule-based technique for newlinebuilding parser for Telugu using linguistic cues as rules. The present research newlineadopts the Indian grammatical tradition i.e. P¯an. ini s Grammatical (PG) tradition newlineas the dependency model to parse sentences. A detailed description of mapping newlinesemantic relations to vibhaktis (case suffixes and postpositions) using linguistic newlinecues in Telugu is presented. newlineAn enhanced annotation scheme for Telugu dependency relations is introduced. newlineChallenges faced in parsing ambiguous structures are elaborated alongside newlineproviding enhanced tags to handle them. The study describes the parsing newlinealgorithm and the linguistic knowledge employed while developing the parser. The newlineresearch further provides results, which suggest that enriching the current parser newlinewith linguistic inputs can increase the accuracy and tackle ambiguity better than newlineexisting data-driven methods. Results are encouraging and this parser proves to be newlineefficient for languages like Telugu which can be later extended to other newlinemorphologically-rich languages. newline