概要:

In this study, text-mining calculate the similarity between the syllabi of the special subjects in the information course in the department of electrical and computer engineering in National Institute of Technology, Gifu College. From the similarity between syllabi, you can understand the relationship between subjects. As a method of calculating similarity between syllabi, the new method was verified. The method is similarity calculation method using clustering words by concept distance in Japanese WordNet. The words in the syllabi are divided into the clusters and word-document matrix is tranformed to cluster-document matrix. It is difficult to judge whether the result of this method is correct. Therefore, experiments are conducted by comparing cosine similarity calculation method and similarity calculation method using LSI, which exist conventionally. As a result, it was found that similarity calculation using word clustering was not very suitable compared to LSI method. There is room for improvement in similarity calculation by clustering words.





Deguchi Lab. 2017年3月6日