Improved Candidate Answers Ranking for QA

Query answering system (QA) has become an important branch of nature language processing (NLP) and information retrieval (IR), it encompasses a wide range of topics related to the storage and retrieval of all kinds of media. This paper presents a new combination of IR grammars, Latent Dirichlet Allocation (LDA) grammars Doc2Vec models and Deep Learning models to improve the value of Mean Reciprocal Rank (MRR) in the factoid question answering task. Empirical study shows the proposed model can effectively position the correct answer and improve the value of MRR when combined with LDA grammars.