Local and Global Feature Extraction for Senses and its application to Topic Tracking

2015 Fiscal Year Final Research Report

• Project/Area Number: 25330255
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Research Field: Intelligent informatics
• Research Institution: University of Yamanashi
• Principal Investigator: FUKUMOTO Fumiyo 山梨大学, 総合研究部, 教授 (60262648)
• Co-Investigator(Renkei-kenkyūsha): SUZUKI TOMOHIRO 山梨大学, 大学院総合研究部, 准教授 (70235977)
• Project Period (FY): 2013-04-01 – 2016-03-31
• Keywords: 分野語義辞書 / 転移学習 / 素性選択 / 文書分類 / 続報記事抽出

Outline of Final Research Achievements

This study proposed a method for lexical semantic extraction which is effective for topic tracking and text categorization that training data may derive from a difference time period from the test data. We present a method that minimizes the impact of temporal effects by using term smoothing and transfer learning techniques. The results showed that integrating term smoothing and transfer learning improves overall performance of topic tracking and text categorization, especially it is effective when the creation time period of the test data differs greatly from the training data.

Free Research Field

自然言語処理