抄録
This paper describes trend analyses of patent applicants and a relevant patent retrieval technique using the trends for higher retrieval accuracy. This method focuses on the difference of the applicants of a query patent and the patent invalidating the query invention. First, we analyze trends of patent applicants from three viewpoints: the attributes of patent documents, the rate of shared terms and the difficulty of relevant patent retrieval. The results show that 21% of the patents invalidating a query have common applicants with their corresponding query and that it is much more difficult to retrieve relevant patents with different applicants from the query. Then, we use four relevant patent retrieval methods to evaluate each behavior of the retrieval accuracy from a viewpoint of the difference of patent applicants. The result shows that the behavior of each method depends on the difference of patent applicants. Finally, we propose a technique to combine relevant patent retrieval methods by using the difference of the behavior of retrieval accuracy in each retrieval method. This combining technique contributes to improving the retrieval accuracy of relevant patents with different applicants from a query.
