SemEval-2014 Task 4: Aspect Based Sentiment Analysis

Proceedings of the 8th International Workshop on Semantic Evaluation (SemEval 2014), pages 27–35, Dublin, Ireland, August 23-24, 2014. SemEval-2014 Task 4: Aspect Based Sentiment Analysis Maria Pontiki Institute for Language and Speech Processing, “Athena” Research Center mpontiki@ilsp.gr Haris Papageorgiou Institute for Language and Speech Processing, “Athena” Research Center xaris@ilsp.gr Dimitrios Galanis Institute for Language and Speech Processing, “Athena” Research Center galanisd@ilsp.gr Ion Androutsopoulos Dept. of Informatics Athens University of Economics and Business ion@aueb.gr John Pavlopoulos Dept. of Informatics, Athens University of Economics and Business annis@aueb.gr Suresh Manandhar Dept. of Computer Science, University of York suresh@cs.york.ac.uk Abstract Sentiment analysis is increasingly viewed as a vital task both from an academic and a commercial standpoint. The majority of current approaches, however, attempt to detect the overall polarity of a sentence, paragraph, or text span, irrespective of the entities mentioned (e.g., laptops) and their aspects (e.g., battery, screen). SemEval- 2014 Task 4 aimed to foster research in the field of aspect-based sentiment analysis, where the goal is to identify the aspects of given target entities and the sentiment expressed for each aspect. The task pro- vided datasets containing manually anno- tated reviews of restaurants and laptops, as well as a common evaluation procedure. It attracted 163 submissions from 32 teams. 1 Introduction With the proliferation of user-generated content on the web, interest in mining sentiment and opinions in text has grown rapidly, both in academia and business. Early work in sentiment analysis mainly aimed to detect the overall polarity (e.g., positive or negative) of a given text or text span (Pang et al., 2002; Turney, 2002). However, the need for a more fine-grained approach, such as aspect-based (or ‘feature-based’) sentiment analysis (ABSA), soon became apparent (Liu, 2012). For example, laptop reviews not only express the overall senti- ment about a specific model (e.g., “This is a great This work is licensed under a Creative Commons At- tribution 4.0 International Licence. Page numbers and pro- ceedings footer are added by the organisers. Licence details: http://creativecommons.org/licenses/by/4.0/ laptop”), but also sentiments relating to its spe- cific aspects, such as the hardware, software, price, etc. Subsequently, a review may convey opposing sentiments (e.g., “Its performance is ideal, I wish I could say the same about the price”) or objective information (e.g., “This one still has the CD slot”) for different aspects of an entity. ABSA is critical in mining and summarizing opinions from on-line reviews (Gamon et al., 2005; Titov and McDonald, 2008; Hu and Liu, 2004a; Popescu and Etzioni, 2005). In this set- ting, ABSA aims to identify the aspects of the en- tities being reviewed and to determine the senti- ment the reviewers express for each aspect. Within the last decade, several ABSA systems of this kind have been developed for movie reviews (Thet et al., 2010), customer reviews of electronic products like digital cameras (Hu and Liu, 2004a) or net- book computers (Brody and Elhadad, 2010), ser- vices (Long et al., 2010), and restaurants (Ganu et al., 2009; Brody and Elhadad, 2010). Previous publicly …

Content truncated to 500 words