A novel framework using zero shot learning technique for a non-factoid question answering system

Abstract

The non-factoid question answering (QA) is the next generation of textual QA systems that gives passage-level summaries for a natural language query posted by the user. The main issue lies in the appropriateness of the generated summary. This paper proposes a framework for a non-factoid QA system that has three main components: (1) a deep neural network classifier, which produces sentence vector considering word correlation and context; (2) zero shot classifier that uses a multi-channel convolutional neural network (CNN) to extract knowledge from multiple sources in the knowledge accumulator, which acts as a knowledge enhancer that strengthens the passage level summary; (3) summary generator that uses maximal marginal relevance (MMR) algorithm, which computes similarity among the query-related answers and the sentences from the zero shot classifier. This model is applied to the datasets WikiPassageQA and ANTIQUE. The experimental analysis shows that this model gives comparatively better results for WikiPassageQA dataset.