JSAI2019

Presentation information

General Session

General Session » [GS] J-9 Natural language processing, information retrieval

[2I5-J-9] Natural language processing, information retrieval: classification and evaluation

Wed. Jun 5, 2019 5:20 PM - 7:00 PM Room I (306+307 Small meeting rooms)

Chair:Hirotoshi Taira Reviewer:Kugatsu Sadamitsu

6:20 PM - 6:40 PM

[2I5-J-9-04] The Simplest Document Classifier

〇Yoshitaka Shirai1, Yutaka Hirata1 (1. Chubu Univercity)

Keywords:Document classification, Analysis, simplification

Document Classification (DC) is the task to assign a document to a specific category. Deep convolutional neural network (DCNN) outperforms humans in DC in terms of speed and accuracy. However, the internal classification process of DCNN-based classifiers is in a black box. Here we propose an extremely simple algorithm for DC that is very fast and highly accurate, and allows us to examine classification processes. We demonstrate the validity of the proposed method, using major document data bases such as 20 Newsgroups, Livedoor-news, IMDB, and Twitter.