| Summary: | 碩士 === 真理大學 === 管理科學研究所 === 95 === Patent is the most import way to protect the new technologies of the factories and stores. The prior art search is necessary when the manufacturerers are proceeding with investment. The retrieval patent documents are useful for the academic research as well. However, in the process of patent retrieval, it has been a choke point to grasp with more exact keywords. Although there are few researches to solve the problems of patent retrieval through text mining, these only focus on the information retrieval and classification, ignoring the minimal yet important keywords.
According to Ohsawa (1998)’s theory of Chance Discovery, it is based on how to explore the rare keyword with important information. However, it usually filters those rare keyword as Noise. Therefore, this research would like to extend Ohsawa’s concept through fixing its structural design. The traditional Key graph cannot incorporate keywords abundantly into it. This research intends to compress the structure of technology. By structure of technology and new keywords links, we can incorporate more unpredictable keywords.
The contribution of this research is to offer a new strategy different form IPC、UPC. The roles of inventor, assignee and agent are reconsidered. In the beginning state with divaricate technological terms, the keywords may be varied for different cultural backgrounds and other reasons. By high frequency keyword to compress structure of DVD technology, this study want to use this method finding the emerging technology.
Making a contrast with TF、TFIDF and BM25, the result of this study can raise the rate of accuracy from 0.449 to 0.571, with the missed 40 patent documents. It proves we can improve patent-researched efficiency practically by extending the theory of Chance Discovery. It is also the way to improve the accuracy of patent research in academic field.
|
|---|
