Real Environment Oriented Financial Question Answering System

With the explosive growth of financial information, financial sector in China has received more and more concern from people. All the shareholders and fund-holders urgently need a system for them to get information of financial field. But in a long time, most question and answering systems only concern questions of fact which only represent a very small part of all questions raised by people in real-life. These systems are far from being able to meet the community's need of question answering about financial information. At the same time, a lot of platforms for online users to ask and answer questions and FAQs (frequently asked questions) services have aroused. Such systems collect all netizens'wisdom to provide knowledge service for those who need. If we can combine the advantages of automatic question and answering system and online user question and answering system, use the contents of the online user Q&A system as knowledge base, and use automatic question answering techniques to find the most suitable answer from knowledge base, in a way, it can meet the ever-increasing demand for financial domain knowledge.As the syntactic analysis technology has become increasingly popular, more and more natural language processing systems are using syntactic analysis techniques. We adopted syntactic parsing technology to the QA system to increase the performance of it.Finally we implemented a financial oriented question and answering system, which is based on the huge data of online user Q&A system and is used to search and return a list of existing question and answer pairs which are semantically closest to the question user need. The main research work include: summarize the special characteristics that questions real environment have, and propose pretreatment of dealing with these characteristics; for of question-answering system and proposed binding processing phase syntactic analysis technology according to the categories of questions as a method of calculation of tag similarity, and did not include a syntactic analysis features of question classification and compare similarity calculation method to prove that the syntactic analysis can improve the classification and the accuracy of the similarity calculation. The technologies used and research results in this system have extensive commonality in limited fields Q&A system.