Question Answering System with Enhancing Sentence Embedding

Abstract

In order to improve the semantic understanding of the input question in the question answering system, a question answering system based on knowledge representation is constructed, which is composed of named entity recognition and question matching. The named entity recognition method based on Bert+BiLSTM+CRF is used, and the BGCNN model proposed in this paper is used for question matching. BGCNN is a model combining Bert, neural network and Siamese network. The average F1 value of the system on the financial data set is 0.9007, which is not a small improvement compared with the previous model.