Probabilistic Approach of Parsing Bengali Sentences

Statistical parsing denotes the technique of syntactical analysis that is based on probabilistic inference from the corpus. In order to remove the structural ambiguity of sentences, a probabilistic technique of parsing Bengali sentences is proposed in this paper. In this paper, we use the CYK algorithm for parsing and apply binarization technique to improve the parsing efficiency. For this purpose, a set of probabilistic context-free grammars (PCFGs) is proposed based on intonation of the sentences and structures which are used for Bayesian inference. The proposed system is tested with 3500 sentences (10648 words) and achieved 85% overall accuracy for all kinds of Bengali sentences.