Study on Selection and Concentration Based on Text Mining of Patent Publication

本研究的目的在于发现文本挖掘技术，以帮助日本企业规划有效的“选择和集中”战略。通常，文本挖掘技术的研究面临着少数重要或关键的词被隐藏在大量的噪声或无意义的词中的困难。为了解决这一难题，我们的研究提出了两种思路或技术。首先，为了使挖掘更加准确，原始文本数据通过附加词集、下划线、链接原始文本数据与其他数据等标注，在数字空间中被赋予了各种附加值。第二，词集在数字空间或关系数据库中朝着明确的方向发展。最后，对符合矩阵进行了多元分析和网络分析。我们的技术有望发现大量的知识，这些知识将在日本规划“选择和集中”战略中发挥重要作用。

The aim of this study is located in discovering the technique of text mining, which contributes to planning valid "selection and concentration" strategy of Japanese companies. In general, studies of text mining technique face to difficulties that a few important or critical words are hidden in a huge quantity of noise or meaningless words. Our study proposed two ideas or techniques in order to solve this difficulty. First, in order to make mining more accuate, original text data are given a variety of value added in digital space by attaching annotation such as words set, underline, linking origina text data with another data and so on. Second, word set are evolved in digital space or relational database in to the unmistaken direction. Finally, multivariate analysis and network analysis are applied to this coincidence matrix. Our technique is expected to discover a lot of knowledge, which will play important role in planning "selection and concentration" strategy in Japan.