Information Retrieval on Social Web Sites That Gathered Multiple Opinions

2021 Fiscal Year Final Research Report

• Project/Area Number: 18K18161
• Research Category: Grant-in-Aid for Early-Career Scientists
• Allocation Type: Multi-year Fund
• Review Section: Basic Section 62020:Web informatics and service informatics-related
• Research Institution: Aoyama Gakuin University
• Principal Investigator: Shoji Yoshiyuki 青山学院大学, 理工学部, 助教 (30783039)
• Project Period (FY): 2018-04-01 – 2022-03-31
• Keywords: 情報検索 / レビューサイト / ランキング学習

Outline of Final Research Achievements

We proposed the methods for aggregating and making searchable the opinions of multiple people on a single topic, such as a review of a certain product. As a concrete example, when various contributors with different backgrounds are on a movie review site, and they review movies using different expressions such as "it made me cry" or "it made my handkerchief soggy," existing document search algorithms cannot correctly rank these movies. We proposed a search algorithm that actually sorts movies in order of likelihood of crying based on the keyword "crying." For this purpose, we successfully extracted the characteristics of movies from the review text, considered the characteristics of the author of the review, and actually generated search results ranking using LEarning to Rank-based techniques.

Free Research Field

情報検索

Academic Significance and Societal Importance of the Research Achievements

近年のWebは、ウェアラブルデバイスなどの普及や、より気軽に投稿できるソーシャルメディアの流行に伴い、より断片的で些細な投稿の集積物になりつつある。そのため、これらをトピックごとにまとめ、任意のキーワードでランキング可能にする技術は重要異性が高い。本研究では、例えば、レビューサイトの投稿を映画ごとにまとめることで「最も『泣ける』映画はどれか」という検索を可能にしたり、地図サイト上の投稿を地物ごとにその目的をまとめることで「『ギターの練習』ができそうな場所はどこか」という検索を可能にした。