UNITOR: Aspect Based Sentiment Analysis with Structured Learning

Proceedings of the 8th International Workshop on Semantic Evaluation (SemEval 2014), pages 761–767, Dublin, Ireland, August 23-24, 2014. UNITOR: Aspect Based Sentiment Analysis with Structured Learning Giuseppe Castellucci(†), Simone Filice(‡), Danilo Croce(⋆), Roberto Basili(⋆) (†) Dept. of Electronic Engineering (‡) Dept. of Civil Engineering and Computer Science Engineering (⋆) Dept. of Enterprise Engineering University of Roma, Tor Vergata, Italy {castellucci,filice}@ing.uniroma2.it; {croce,basili}@info.uniroma2.it Abstract In this paper, the UNITOR system partici- pating in the SemEval-2014 Aspect Based Sentiment Analysis competition is pre- sented. The task is tackled exploiting Ker- nel Methods within the Support Vector Machine framework. The Aspect Term Extraction is modeled as a sequential tag- ging task, tackled through SVMhmm. The Aspect Term Polarity, Aspect Category and Aspect Category Polarity detection are tackled as a classification problem where multiple kernels are linearly combined to generalize several linguistic information. In the challenge, UNITOR system achieves good results, scoring in almost all rank- ings between the 2nd and the 8th position within about 30 competitors. 1 Introduction In recent years, many websites started offering a high level interaction with users, who are no more a passive audience, but can actively produce new contents. For instance, platforms like Amazon1 or TripAdvisor2 allow people to express their opin- ions on products, such as food, electronic items or clothes. Obviously, companies are interested in understanding what customers think about their brands and products, in order to implement correc- tive strategies on products themselves or on mar- keting solutions. Performing an automatic analy- sis of user opinions is then a very hot topic. The automatic extraction of subjective information in text materials is generally referred as Sentiment Analysis or Opinion Mining and it is performed 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/ 1http://www.amazon.com 2http://www.tripadvisor.com via natural language processing, text analysis and computational linguistics techniques. Task 4 in SemEval 2014 edition3 (Pontiki et al., 2014) aims at promoting research on Aspect Based Opinion Mining (Liu, 2007), which is approached as a cas- cade of 4 subtasks. For example, let us consider the sentence: The fried rice is amazing here. (1) The Aspect Term Extraction (ATE) subtask aims at finding words suggesting the presence of as- pects on which an opinion is expressed, e.g. fried rice in sentence 1. In the Aspect Term Polarity (ATP) task the polarity evoked for each aspect is recognized, i.e. a positive polarity is expressed with respect to fried rice. In the Aspect Category Detection (ACD) task the cate- gory evoked in a sentence is identified, e.g. the food category in sentence 1). In the Aspect Cat- egory Polarity (ACP) task the polarity of each ex- pressed category is recognized, e.g. a positive category polarity is expressed in sentence 1. Different strategies have been experimented in recent years. Classical approaches are based on machine learning techniques and rely on sim- ple representation features, such as unigrams, bi- grams, Part-Of-Speech (POS) tags (Pang et …

Content truncated to 500 words