Extracting and Normalizing Entity-actions from Users' Comments

Open Access

Abstract

With the growing popularity of opinion-rich resources on the Web, new opportunities and challenges arise and aid people in actively using such information to understand the opinions of others. Opinion mining process currently focuses on extracting the sentiments of the users on products, social, political and economical issues. In many instances, users not only express their sentiments but also contribute their ideas, requests and suggestions through comments. Such comments are useful for domain experts and are referred to as actionable content. Extracting actionable knowledge from online social media has attracted a growing interest from both academia and the industry. We define a new problem in this line which is extracting entity-actionable knowledge from the users' comments. The problem aims at extracting and normalizing the entity-action pairs. We propose a principled approach to solve this problem and detect exactly matched entities with 75.1% F-score and exactly matched actions with 76.43% F-score. We could achieve an average precision of 81.15% for entity-action normalization.