Critical Analysis of Existing Punjabi Grammar Checker and a Proposed Hybrid Framework Involving Machine Learning and Rule-Base Criteria

Abstract

An important area of research involving Artificial Intelligence (AI) is Natural Language Processing (NLP). The objective of training a machine is to imitate and manipulate text and speech of humans. Progressive research is undertaken to find connections between humans and their usage of language commonly used being referred as Natural Language. Various tools for different languages have been developed for operating the natural languages widely used by public. NLP integrates various disciplines and works cohesively for processing text, Information Retrieval, AI and so on. One such tool used for checking the accuracy of a given sentence in any language is referred to as a Grammar Checker. So a Grammar checker of a particular language explores grammatical errors (if any) and provides remedial suggestions for correction of the same. Such feature is imbibed by virtue of Natural Language Processing using Computational Linguistics. We have justified the need of an emerging Machine Learning technique by critically evaluating the existing Punjabi Grammar checker that was developed earlier in light of certain real-time cases. This process is accomplished by critically evaluating the output of each phase and identifying the component accountable for generating maximum errors and false alarms. Based on this analysis, we have proposed a hybrid framework as an efficient way of analyzing correction in sentences. This is attainable through the said booming technique of Machine Learning explicitly using Deep Neural Networks in combination with the existing rule-based approach. It's a novel approach as no work using machine learning has been done earlier in Punjabi Grammar Checker.