Enhanced concept-level sentiment analysis system with expanded ontological relations for efficient classification of user reviews

Abstract

Background/introduction: Concept-level sentiment analysis deals with the extraction and classification of concepts and features from user reviews expressed online about products and other entities like political leaders, government policies, and others. The prior studies on concept-level sentiment analysis have used a limited set of linguistic rules for extracting concepts and their associated features. Furthermore, the ontological relations used in the early works for performing concept-level sentiment analysis need enhancement in terms of the extended set of features concepts and ontological relations. Methods: This work aims at addressing the aforementioned issues and tries to bridge the literature gap by proposing an extended set of linguistic rules for concept-feature pair extraction along with enhanced set ontological relations. Additionally, a supervised a machine learning technique is implemented for performing concept-level sentiment analysis. Results and conclusions: Experimental results depict the effectiveness of the proposed system in terms of improved efficiency (P: 88%, R: 88%, F-score: 88%, and A: 87.5%).

Keywords

• Machine learning techniques
• Support vector machine
• Formal concept analysis (FCA)
• Concept lattice
• Ontological relations