Research On Architecture And Key Algorithm Of Question Answering For Science And Technology Policies

STP(Science and Technology Policy)is an important part of the national innovation and development strategy,and it is vital for the coordinated development of society,economy and science.With the development of emerging technologies such as big data and artificial intelligence,the current situation of multi-source and redundancy of STP has accelerated public pursuit of professional and precise information.How to provide the public with efficient,accurate,and valuable policy knowledge has become an urgent problem for the government in the field of policy knowledge services.At present,the policy Q&A(Question Answering)function is embedded in e-service platform by many government departments in China to strengthen public interaction and exchange.However,the Q&A principle of the existing policy knowledge service is similar to the keyword matching of the search engine,and the answer result is unsatisfactory or irrelevant.In view of this,this thesis focused on the STP knowledge service field.With the help of data mining,natural language processing and other technologies,this thesis designed a one-stop Q&A architecture for knowledge integration and management.By improving the function of STP knowledge service,and optimizing the Q&A strategy and technology,the system provides efficient and accurate knowledge service for the public to "understand policy,comprehend policy and make good use of policy".The main research work of this thesis is as follows:Firstly,the relevant knowledge background was discussed,including the relevant research and service status of STP,the concept and characteristics of knowledge service,as well as the classification,module and key algorithm of Q&A.These knowledge backgrounds laid the theoretical foundation for constructing the Q&A architecture of STP and improving key algorithm.Secondly,through the research of STP service platform and user demand,this thesis make clear the advantages and disadvantages of the policy Q&A platform,and grasps the characteristics of the enterprises' demand for STP knowledge.Based on the theory of knowledge service and the analysis of status and demand,the architecture of STP knowledge service was proposed,which takes Q&A technology as the core.In addition,the elements,characteristics and levels of the architecture were pointed out,and the Q&A strategy including chat,retrieval and recommendation was designed for the technology layer.Thirdly,the key algorithm of technology layer in architecture was improved in this thesis.The existing Smooth Inverse Frequency algorithm only pays attention to some features of sentences and ignores other valuable information contained in sentences.Therefore,on the basis of Smooth Inverse Frequency algorithm,this thesis introduceed two features of word order and semantic dependency,and the algorithm was validated by the data of science and technology policy Q&A.The experimental results showed that the F-metric value of the proposed algorithm was improved by 9.21% compared with the Smooth Inverse Frequency algorithm.Finally,summarizing the research content,suggestions to promote the realization of STP Q&A were put forward,and the shortcomings of this study were described.