Twitter上での発話履歴の時系列パターンに基づく特定発話行動予測手法の検討 An Analysis for Developing User Behavior Analytic Model Construction Method on Twitter

この論文にアクセスする

この論文をさがす

著者

抄録

Twitter をはじめとする SNS では,ユーザの日々の興味や関心がテキストやその発話行動として可視化されている.そのなかで,企業をはじめとして,興味・関心のあるユーザを特定する要求が高まりつつある.しかしながら,単に発信情報を受信するユーザが真に興味・関心を持っているかどうかを的確に判断することは困難が伴う.本研究では,Twitter 上でリツイートと呼ばれるユーザが受信した情報を再発信する行動に着目し,過去の発話履歴の内容に基づいて再発信行動の予測を行うモデル構築手法の開発を目指す.本稿では,インターネット上で大規模な通信販売を行うサイトの発信したテキストとそのテキストを受信するフォロワーについて,再発信されたテキストとフォロワーの日々の発話テキストから得られた特徴語の比較を行う.さらに,行動予測モデル構築のため,フォロワーの発話テキストから得られた特徴語の使用頻度に基づく評価指標の時系列パターン抽出の可能性について検討する.According to popularization of many SNS such as Twitter, enterprise users and some other people want to reach interested users more efficiently. However, it is difficult to detect more deep interests of the users without considering the users' behavior. In this study, I focus on a characteristic behavior of the users on Twitter, called retweet by followers. By taking followers' tweet history, a method constructing analytic models based on temporal patterns of term evaluation indices is described in this paper. The method combines ordinary used characteristic term extraction and a temporal pattern extraction of the terms usages. In the experiment, both of the characteristic terms in the retweeted text and the followers' tweet history is shown on the three major e-commerce accounts in Japan. Based on the results, some temporal patterns of the term importance indices are also shown to discuss the feasibility for predicting the retweet behavior of the followers.

According to popularization of many SNS such as Twitter, enterprise users and some other people want to reach interested users more efficiently. However, it is difficult to detect more deep interests of the users without considering the users' behavior. In this study, I focus on a characteristic behavior of the users on Twitter, called retweet by followers. By taking followers' tweet history, a method constructing analytic models based on temporal patterns of term evaluation indices is described in this paper. The method combines ordinary used characteristic term extraction and a temporal pattern extraction of the terms usages. In the experiment, both of the characteristic terms in the retweeted text and the followers' tweet history is shown on the three major e-commerce accounts in Japan. Based on the results, some temporal patterns of the term importance indices are also shown to discuss the feasibility for predicting the retweet behavior of the followers.

収録刊行物

  • 研究報告知能システム(ICS)  

    研究報告知能システム(ICS) 2015-ICS-178(11), 1-5, 2015-02-23 

    一般社団法人情報処理学会

各種コード

  • NII論文ID(NAID)
    110009882536
  • NII書誌ID(NCID)
    AA11135936
  • 本文言語コード
    JPN
  • 資料種別
    Technical Report
  • ISSN
    09196072
  • データ提供元
    IPSJ 
ページトップへ