Research on Template-Based Factual Automatic Question Answering Technology

With the increase of people’s demand for information retrieval, question answering system as a new generation of retrieval methods has attracted more and more attention. This paper proposes an automated template generation method based on remote supervision. In the process of generating template, in order to identify the keywords express the semantics of the problem, this paper designs an algorithm to mapping entity relationship to keywords by using pattern mining and statistical methods. This method implements automation of template generation, and makes the templates more accurate. For the template matching, by Deep-learning-based template matching algorithm, this paper use vector to represent words, and designs a tree-based convolutional neural network to extract features of the dependency syntax information. Building a neural network model for template prediction. Experimental results show that the above method can effectively improve the matching accuracy compared with the traditional template matching method.