OntoSenticNet 2: Enhancing Reasoning Within Sentiment Analysis
Mauro Dragoni, Ivan Donadello, Erik Cambria
Abstract
Sentiment analysis is a trending topic that has not yet exhausted its attractiveness, despite the huge research effort carried out in the last 15 years. One of the most promising directions to investigate is the integration of knowledge-based representations within sentiment analysis systems in order to enhance their expressiveness and, at the same time, to enable reasoning over the relevant information detected within opinion-based sources. In this article, we present an improved version of OntoSenticNet providing: i) an updated definition of concepts, properties, and individuals together with an improved hierarchical organization of such entities; ii) the modeling of the sentic algebra elements for supporting the execution of semantic sentiment operations at reasoning time; and iii) the conceptual model of sentiment dependencies and discovery paths. The process of building OntoSenticNet 2 is discussed and some examples are proposed in order to illustrate the conceptual model.