Extraction of Implicit Aspects for Opinion Mining on User Reviews

2022 Fiscal Year Final Research Report

• Project/Area Number: 20K11950
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Review Section: Basic Section 61030:Intelligent informatics-related
• Research Institution: Japan Advanced Institute of Science and Technology
• Principal Investigator: Shirai Kiyoaki 北陸先端科学技術大学院大学, 先端科学技術研究科, 准教授 (30302970)
• Project Period (FY): 2020-04-01 – 2023-03-31
• Keywords: オピニオンマイニング / 属性抽出 / 暗黙的属性

Outline of Final Research Achievements

When users express their opinion toward an aspect of a product (e.g. “battery” or “price” of a mobile phone) in a review, they sometimes write opinions implicitly without using explicit words. In this study, such an implicitly expressed aspect is called “implicit aspect”. We propose a method to extract implicit aspects automatically. First, by extracting explicit aspects from a large amount of reviews, sentences containing or not containing explicit aspects are obtained. The latter is regarded as an implicit sentence, which may includes a certain implicit aspect. Next, the implicit aspect of the implicit sentence is determined by searching the similar sentence that includes the explicit aspect and propagating that explicit aspect as the implicit aspect of the sentence. Finally, a model to extract implicit aspects is trained by deep learning using the labeled sentences as training data.

Free Research Field

自然言語処理

Academic Significance and Societal Importance of the Research Achievements

従来の属性抽出に関する研究の多くは明示的属性を対象としていたのに対し，本研究では暗黙的属性を抽出の対象とする点に特徴がある．近年の自然言語処理技術は深層学習による手法が主流だが，最新の深層学習モデルを適用するために，暗黙的属性がラベル付けされたデータセットを自動的に構築する点に学術的意義がある．
ユーザレビューを分析し，製品やサービスに対する評判を明らかにするオピニオンマイニングは，ユーザにとって有益な情報をもたらす技術である．本研究の成果により，明示的属性だけではなく暗黙的属性を分析することが可能になり，オピニオンマイニングの精緻化が促進されるため，その社会的意義は大きい．