Estimating Invention Task and Means from Patent Journals Based on the Document Similarity with the Patent Journal Structures

2014 Fiscal Year Final Research Report

• Project/Area Number: 24500121
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Research Field: Media informatics/Database
• Research Institution: Iwate Prefectural University
• Principal Investigator: KUREMATSU Masaki 岩手県立大学, ソフトウェア情報学部, 准教授 (00305286)
• Project Period (FY): 2012-04-01 – 2015-03-31
• Keywords: 情報検索 / 特許情報 / 文書類似度 / 文書分類

Outline of Final Research Achievements

In order to support checking exists patents, I proposed a framework of estimating invention task and means from patent journals based on the document similarity between it and each exist patent given invention task and means by experts. The framework divides a patent into 4 parts based on the patent journal structure initially. It gets the similarity of each part and integrates them and shows invention task and means ranked in descending order by the similarity in the difference 4 orders. I did an experiment with an expert and small data set. In this research, invention task and means have sub categories. I evaluate this system from the viewpoint of the rank of invention task and means given by experts. 17% of the collect　invention task and means with sub categories had entered in the top 10 and 49% of them without sub categories had entered in the top 10. This experimental result shows that is it possible to use my idea to estimate invention task and means from patent journals.

Free Research Field

知能情報学