Combining resources to improve unsupervised sentiment analysis at aspect-level

• Published in: Journal of information science Vol. 42; no. 2; pp. 213 - 229
• Main Authors: Jiménez-Zafra, Salud M., Martín-Valdivia, M. Teresa, Martínez-Cámara, Eugenio, Ureña-López, L. Alfonso
• Format: Journal Article
• Language: English
• Published: London, England SAGE Publications 01-04-2016; Bowker-Saur Ltd
• Subjects: Classification; Classifiers; Consumer goods; Data mining; Linguistics; Product reviews; Purchasing; Searches; Searching; Sentiment analysis; Studies; Target markets; Tasks; User satisfaction; Users; lexicon-based approach; linguistic resources; polarity classification; Aspect-based sentiment analysis; voting classifier system
• ISSN: 0165-5515; 1741-6485
• DOI: 10.1177/0165551515593686

Every day more companies are interested in users’ opinions about their products or services. Also, every day there are more users that search for reviews on the web before purchasing a product. These users and companies are not satisfied with knowing the overall sentiment of a product, they want a finer knowledge of users’ opinions. Owing to this fact, more and more researchers are working on sentiment analysis at aspect-level. This paper describes an unsupervised approach for aspect-based sentiment analysis, which aims to identify the aspects of given target entities and the sentiment expressed for each aspect. We have evaluated several tasks, although perhaps the major novelty is in the classification of the aspects. We employ a lexicon-based method combining different linguistic resources and we conclude that the combination of several classifiers improves the classification significantly. In addition, a comparison with a supervised system is performed in order to determine the strengths and weakness of each of them.