Automated Language Scoring System by Employing Neural Network Approaches

Present approaches of automated language scoring lack the ability to investigate the multiple-level and several contexts of sequential features which are helpful to examine the language proficiency for the responses (monologue and dialogue). This paper aimed to identify the different levels of sequential features with respect to various contexts to evaluate the speaker's proficiency of a language. We have employed Neural Network based automated assessment. We have combined 3 attention based Bidirectional Long-Short-Term-Memory to effectively consider the three dimensions of a speech as delivery, grammar and content. The resultant outcomes demonstrated that our methodology have outperformed the traditional ways of language scoring.