@article{oai:ipsj.ixsq.nii.ac.jp:02003329,
 author = {堀川,達平 and 北山,大輔 and Tappei Horikawa and Daisuke Kitayama},
 issue = {3},
 journal = {情報処理学会論文誌データベース（TOD）},
 month = {Jul},
 note = {学習や調査など，幅広く知識を獲得するようなオープンエンドの検索タスクとして探索的検索タスクというものが存在する．探索的検索タスクにおいて，ユーザの事前知識が乏しい場合，検索初期に検索対象の概観をとらえそこなうと，得た情報と思いつきによる検索となり，多様な検索を行うことが難しくなる．そこで我々は，探索的検索タスクにおいて，ユーザが検索初期に対象の概観を得ることを目標に，大規模言語モデルが内包している知識をもとに，検索テーマのサブトピックをヒントとして提示する手法を提案する．プロトタイプシステムを構築し，被験者を用いて検索行動を収集し，閲覧したWebページと入力クエリの面から，Web検索行動の多様性が向上するかを評価した．その結果，提案手法により検索行動の多様性が向上したことが示された．, The exploratory search task is a type of open-ended search task, in which users acquire knowledge in a wide range of areas, such as through study or research. In exploratory search tasks, if the user has little prior knowledge, if they fail to get an overview of the search target at the beginning of the search, they will search based on the information they have and their hunches, and it will become difficult to perform a diverse range of searches. Therefore, in exploratory search tasks, we propose a method for presenting subtopics of the search theme as hints based on the knowledge contained in large-scale language models, with the aim of helping users get an overview of the target at the beginning of the search. We built a prototype system, collected search behavior using subjects, and evaluated whether the diversity of web search behavior improved from the perspective of the web pages viewed and the input queries. As a result, it was shown that the diversity of search behavior improved with the proposed method.},
 pages = {1--7},
 title = {大規模言語モデルを用いたサブトピック提示によるWeb検索行動の多様化とその評価},
 volume = {18},
 year = {2025}
}