Design And Implementation Of Science And Technology Policy Analysis System Based On Topic

Science and technology policy is a planned and organized guideline of science and technology designed to standardize science and technology field working normally, reflecting the effective control of the current direction of the science and technology development and industrial scale. With the science and technology policy texts increasing year by year, how to manage science and technology policy texts effective and help researchers obtain the valuable information quickly has become a problem need to solve current. In this paper, based on the actual project "Hebei province Science and Technology Policy Services Platform", due to the shortcomings of dealing with large-scale science and technology policy, introducing topic found and text clustering technology into the science and technology policy texts processing, using Java Web technology to develop the science and technology policy analysis system, and integrating it into the platform. The main work is as follows:(1)Most of science and technology policy texts have big scale and long length. According to the two characteristics of the policy texts, building a model to policy texts using the LDA model, through which text set from high dimension words space was mapped to the low dimensional topic space. Therefore, utilizing LDA model can solve the problem of sparse text representation in high-dimensional space generated by dealing with large-scale data.(2)After founding the hidden topics contained in the science and technology policy set, extracting the release time and the implementation scope of the science and technology policy. Then using the topic strength calculation method this paper puts forward to analyze the variation trend of the topic strength under the different release time and geographical conditions.(3)On the basis of the value of topic similarity, using the k-means algorithm to cluster the science and technology policy, through which the technology policy set is divided into different classes. Aiming at the disadvantages of k-means algorithm, proposing the improved k-means algorithm based on community discovery. It determines the optimal cluster number and initial the cluster center by community discovery and center node selection method in the community. The effectiveness of the method proposed is verified by the experiments finally.(4)According to the theory research, in view of the new requirements with "Hebei province Science and Technology Policy Services Platform", applying the topic analysis and text clustering technology proposed above to the project to develop a science and technology policy analysis system,which realizes the policy text automatic analysis and management.Finally, summarize the points of view and point out the insufficient, to imagine the future research direction.