Patent Text Clustering Analysis And Visualization

In the 17^th CPC National Congress Report, there are some proposals like "Perfect the Patent Protection System", "Encourage Innovation of Science and Technology", "Master core technologies and Own independent patents in key technologies and new fields". In order to obtain the core technologies and foster the core capability of Chinese enterprises, we should accelerate the construction of Patent System and Patent Capability. And this has been the important strategy for China becoming a new industrialized country. However, our enterprises haven't become the main body of independent innovation progress. Many enterprises made mistakes in choosing the Field, Direction and Project of Innovation, and it lead to a lot of repeat researches. In the new product development, they lack the accurate forecast of future development trend. And in the product and technology innovation, they can't find a direct innovation solution fast. Most of Chinese enterprise don't have a perfect Intellectual Property strategy, and are not good at knowing competitors and protecting themselves through Patent Information. As a result, how to analyze the Patent Information effectively and assist the innovation decision-making of enterprise, and help Chinese enterprise settle the problems in the innovation activity, has become a very important research project with practical significance.At present, foreign research institute and software corporation has developed some mature analysis software. However, those software are developed mainly for foreign patents. In our country, there are some research results, but most of them only provide the function of classification, retrieval, management, statistic based on the patent external features. The analysis of patent content is just in the starting stage.Considering the research status, this paper made a deep research on the text clustering and patent analysis, and then made a general summarization of the patent analysis technology application in the enterprises. On the basis of it, the paper proposed a decision support matrix for enterprise technology innovation based on the patent mining, and also designed the realization approach. Based on the discussion of the key techniques in the text clustering: term extraction, dimension reduction, cluster algorithm optimizing, cluster result visualization, this paper proposed a patent text clustering method system from the angle of subjects and classification based on the patent content(title, abstract), and then designed and realized the patent text clustering system. At last, taking the LCD patents as an example, this paper validated the validity of this system, and also made an analysis of cluster results.