A Comprehensive Parts of Speech Tagger for Automatically Checked Valid Bengali Sentences

Abstract

This paper constructs a method for corroborating Bengali sentences whether they are equitable syntactically, semantically and pragmatically; then a method is devised for detecting parts of speech (POS) from the valid sentences. In our approach, we have analyzed several techniques to check whether construction of inputted Bengali sentence is valid, and finally, we have chosen the best technique among them. Moreover, after analyzing the construction of several Bengali sentences we have designed a rule-based algorithm for detecting POS with a significant accuracy. The idea of ignoring sentences with grammatical mistakes helped significantly to achieve higher accuracy and to reduce execution time. Moreover, our projected method achieved an accuracy of 91.45% which is the highest among similar POS tagger.