Research And Design Of FAQ Question Answering System Based On Text Semantic Matching

With the development of the Internet and the explosive growth of information,compared with traditional search engines,question answering systems,have gradually become an application hotspot,as an efficient way to obtain information.As a typical question answering system,the FAQ answering system is often applied to customer service.FAQ answering system can not only meet the demand of users,but also reduce the cost of manual customer service.So it has good prospect and market demand.The important technology of the FAQ answering system is text semantic matching.Text semantic matching models based on the pre-training model,which are commonly used,still have some shortcomings,such as the inability to extract contextual semantic cues well.Moreover,the FAQ answering system is suitable for various professional fields.Therefore,it is necessary to design the development process of the FAQ answering system,in order to facilitate it conveniently for different fields.Based on the Hainan Smart Tourism Project,this thesis researches,designs and implements a FAQ answering system based on text semantic matching.The specific contents are as follows:Firstly,due to the lack of data,I investigate some large-scale tourism websites.After that we utilize crawler technology and design a crawler for tourism Q-A data,which is used to get Q-A data from tourism websites for our project.On this basis,the data ugmentation method based on hybrid methods is designed,and the dataset of Hainan Tourism Question Pair is constructed,which provides enough data for model training.Secondly,FAQ answering system generally uses text semantic matching to retrieve answers.To overcome the shortcomings of the current text semantic matching algorithm based on pre-training model,this thesis proposes a text semantic matching algorithm based on graph convolution network and pre-training language model,which realizes neighborhood propagation through graph convolution.Furthermore,LSTM is used to extract sequence features.The experimental results show that this method is effective and performs better than the baseline model in some Chinese and English datasets.Thirdly,in this thesis we design and realize a FAQ answering system of Hainan tourism,which supports Web access.The system can provide questions answering service to tourists and provide management capabilities to manager.In addition,we also design the development process of FAQ answering system,and provides customized development function in the system,so that people in different professional fields can easily develop FAQ question answering system for different needs.