著者

齋藤 準樹 湯川 高志

出版者

情報処理学会

雑誌

研究報告情報基礎とアクセス技術(IFAT) (ISSN:21862583)

巻号頁・発行日

vol.2011, no.2, pp.1-8, 2011-03-21

---

本稿では,ソーシャルブックマークに含まれるタグの共起関係から階層的な類語辞書を作成し,それを用いてユーザの興味語を抽出する手法および興味語の類似性によりユーザ推薦を行う手法を提案する.また提案した手法について,Twitter を対象として興味抽出および推薦の精度と意外性に関する評価実験を実施し,有用性の確認を行った結果についても述べる.In the present paper, as a means of extracting user interest for the purpose of user recommendation, the authors propose a method for constructing the hierarchy of words based on SBM tags and to emphasize characteristic word by using this relation. Additionally, the user recommendation system based on this interest extraction is proposed. As a result of a survey on Twitter, the authors discovered that the tags in SBM and their hierarchy have a rich vocabulary for extracting the interests of Twitter users. Moreover, experimental results have indicated that the user recommendation system attains approximately 0.41-0.48 precision if the friend relations in SNS are also utilized as a user preference data.