会議名: 情報知識学会 第11回(2003年度)研究報告会
We developed a searching system which takes a company profile as an input and retrieves the information of papers or researchers. For the system to extract the keywords from company profiles, papers and researcher information, we focus on complex noun phrases included in the documents. Actually, each keyword is weighted not only according to how often it appears, but according to how often the constitutive nouns appear. We took, furthermore, experiments for the comparison between our system and Namazu, a searching system. As a result, our system is 0.76 time as much as Namazu with regard to recall, while it is 3.4 times with regard to precision.