Combine multi-features with deep learning for answer selection

Answer selection is an important subtask in open-domain question answering (QA) system, which mainly models for question and answer pairs. In this paper, we first develop a basic framework based on bidirectional long short term memory (Bi-LSTM), and then we extract lexical and topic features in question and answer respectively, finally, we append these features to Bi-LSTM models. Our models experiment on WikiQA dataset, Experimental results show that our models get a slight improvement compared to other published state of the art results.