Workshop Proceedings of the 14th International AAAI Conference on Web and Social Media

Workshop: 5th Worksop on Social Sensing (SocialSens 2020)

DOI: 10.36190/2020.24

Published: 2020-06-05
Linguistic Rules for Fine-Grained Opinion Extraction
Francielle Alves Vargas, Thiago Alexandre Salgueiro Pardo

Opinion mining applications are usually classified according to the granularity level they tackle: the document, sentence and aspect levels. In our work, we focus on aspect-based opinion mining, more specifically, on the aspect extraction task, which is essential to perform richer text analyses and produce better results in the area. To the best of our knowledge, no previous systematic effort exists on extraction explicit and implicit aspects for Portuguese. To fill this important gap, in this paper we explore the opinion mining aspect extraction task for Brazilian Portuguese. We performed a linguistic empirical study on relevant proprieties in narratives of product reviews and annotated a corpus manually that we test on three different domains (smartphones, digital cameras and books). Based on this data, we propose linguistic rules for extracting explicit and implicit aspects, which we categorize as psychological verb-based and semantic relation-based rules.