Parsing methodology for Bangla natural language sentences

Abstract

The research described in this thesis aims at the design and implementation of a parsing methodology for Bangia natural language sentences that is used as part of a BangIa to English machine translation system. Specifically, we have focused on the development of a system that performs syntactic analysis of the selected texts from newspaper articles and texts of books written in BangIa language. The main component of a J\![Tsystem is the parser. For a parsing system, the source language sentences are taken as input. The task of an automatic parser is to take a grammar and a sentence and apply the grammar to sentence and show how the words are combined into phrases and how the phrases are put together to fOllll larger phrases (including sentences). Phrase-structure rules has been used in context-free and context-sensitive grammars to describe of a given language syntax. This is .'; accomplished by attempting to break language down into its constituent parts or parts of speech including nouns, pronouns, verbs, adjective, and prepositions. This thesis proposed a set of context-free and context-sensitive phrase structure rules to parse all kinds of BangIa natural language sentences including simple, complex and compound sentences. The proposed rules can parse all types of Bangia sentences. The output of a parser has been represented as an SR, which simply records the rules and how they are matched. The result of a parser that produced a list of tokens from a BangIa sentence is then used as input for an MT engine to produce other equivalent sentence of the given BangIa sentence. The parser module has been developed and tested for different kinds of BangIa sentences. The results shown that the parser module parsed the most of the sentence of the articles except interrogative, optative, exclamatory sentences. We got successful parsing output for most of the test cases.