UWB: Machine Learning Approach to Aspect-Based Sentiment Analysis

Proceedings of the 8th International Workshop on Semantic Evaluation (SemEval 2014), pages 817–822, Dublin, Ireland, August 23-24, 2014. UWB: Machine Learning Approach to Aspect-Based Sentiment Analysis Tom´aˇs Brychc´ın NTIS – New Technologies for the Information Society, Faculty of Applied Sciences, University of West Bohemia, Univerzitn´ı 8, 306 14 Plzeˇn Czech Republic brychcin@kiv.zcu.cz Michal Konkol NTIS – New Technologies for the Information Society, Faculty of Applied Sciences, University of West Bohemia, Univerzitn´ı 8, 306 14 Plzeˇn Czech Republic konkol@kiv.zcu.cz Josef Steinberger Department of Computer Science and Engineering, Faculty of Applied Sciences, University of West Bohemia, Univerzitn´ı 8, 306 14 Plzeˇn Czech Republic jstein@kiv.zcu.cz Abstract This paper describes our system partici- pating in the aspect-based sentiment anal- ysis task of Semeval 2014. The goal was to identify the aspects of given tar- get entities and the sentiment expressed to- wards each aspect. We firstly introduce a system based on supervised machine learning, which is strictly constrained and uses the training data as the only source of information. This system is then ex- tended by unsupervised methods for latent semantics discovery (LDA and semantic spaces) as well as the approach based on sentiment vocabularies. The evaluation was done on two domains, restaurants and laptops. We show that our approach leads to very promising results. 1 Introduction The majority of current sentiment analysis ap- proaches tries to detect the overall polarity of a sentence (or a document) regardless of the target entities (e.g. restaurants) and their aspects (e.g. food, price). By contrast, the ABSA (aspect based sentiment analysis) task is concerned with identi- fying the aspects of given target entities and esti- mating the sentiment polarity for each mentioned aspect. The aspect scenario can be decomposed into two tasks: aspect extraction and aspect sentiment classification (Liu, 2012). The task of aspect extraction is to recognize aspects of the entity and more generally can be seen as an information extraction task. The ba- sic approach is finding frequent nouns and noun 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 de- tails: http://creativecommons.org/licenses/ by/4.0/ phrases (Liu et al., 2005; Blair-Goldensohn et al., 2008; Moghaddam and Ester, 2010; Long et al., 2010). Aspect extraction can be also seen as a special case of the general information extrac- tion problem. The most dominant methods are based on sequential learning (e.g. HMM – Hidden Markov Models (Rabiner, 2010) or CRF – Condi- tional Random Fields (Lafferty et al., 2001)). An- other group of methods use topic models (Mei et al., 2007; Titov and McDonald, 2008; Blei et al., 2003). Aspect sentiment classification determines whether the opinions on different aspects are positive, negative, or neutral. While lexicon-based approaches use a list of aspect-related sentiment phrases as the core resource (Ding et al., 2008; Hu and Liu, 2004), the key issue for learning methods is to determine the scope of each sentiment expression, i.e., whether it covers the aspect in the sentence (Jiang et …

Content truncated to 500 words