EliXa: A Modular and Flexible ABSA Platform

Proceedings of the 9th International Workshop on Semantic Evaluation (SemEval 2015), pages 748–752, Denver, Colorado, June 4-5, 2015. c⃝2015 Association for Computational Linguistics EliXa: A modular and flexible ABSA platform I˜naki San Vicente, Xabier Saralegi Elhuyar Foundation Osinalde industrialdea 3 Usurbil, 20170, Spain {i.sanvicente,x.saralegi}@elhuyar.com Rodrigo Agerri IXA NLP Group University of the Basque Country (UPV/EHU) Donostia-San Sebasti´an rodrigo.agerri@ehu.eus Abstract This paper presents a supervised Aspect Based Sentiment Analysis (ABSA) system. Our aim is to develop a modular platform which allows to easily conduct experiments by replacing the modules or adding new features. We obtain the best result in the Opinion Target Extrac- tion (OTE) task (slot 2) using an off-the-shelf sequence labeler. The target polarity classi- fication (slot 3) is addressed by means of a multiclass SVM algorithm which includes lex- ical based features such as the polarity values obtained from domain and open polarity lex- icons. The system obtains accuracies of 0.70 and 0.73 for the restaurant and laptop domain respectively, and performs second best in the out-of-domain hotel, achieving an accuracy of 0.80. 1 Introduction Nowadays Sentiment Analysis is proving very use- ful for tasks such as decision making and market analysis. The ever increasing interest is also shown in the number of related shared tasks organized: TASS (Villena-Rom´an et al., 2012; Villena-Rom´an et al., 2014), SemEval (Nakov et al., 2013; Pon- tiki et al., 2014; Rosenthal et al., 2014), or the SemSA Challenge at ESWC20141. Research has also been evolving towards specific opinion ele- ments such as entities or properties of a certain opin- ion target, which is also known as ABSA. The Se- meval 2015 ABSA shared task aims at covering the 1http://challenges.2014.eswc- conferences.org/index.php/SemSA most common problems in an ABSA task: detect- ing the specific topics an opinion refers to (slot1); extracting the opinion targets (slot2), combining the topic and target identification (slot1&2) and, finally, computing the polarity of the identified word/targets (slot3). Participants were allowed to send one con- strained (no external resources allowed) and one un- constrained run for each subtask. We participated in the slot2 and slot3 subtasks. Our main is to develop an ABSA system to be used in the future for further experimentation. Thus, rather than focusing on tuning the different modules our main goal is to develop a platform to facilitate future experimentation. The EliXa system consists of three independent supervised modules based on the IXA pipes tools (Agerri et al., 2014) and Weka (Hall et al., 2009). Next section describes the ex- ternal resources used in the unconstrained systems. Sections 3 and 4 describe the systems developed for each subtask and briefly discuss the obtained results. 2 External Resources Several polarity Lexicons and various corpora were used for the unconstrained versions of our systems. To facilitate reproducibility of results, every re- source listed here is publicly available. 2.1 Corpora For the restaurant domain we used the Yelp Dataset Challenge dataset2. Following (Kiritchenko et al., 2014), we manually filtered out categories not corre- sponding to food related businesses (173 out …

Content truncated to 500 words