Bilingual Experiments with an Arabic-English Corpus for Opinion Mining

Proceedings of Recent Advances in Natural Language Processing, pages 740–745, Hissar, Bulgaria, 12-14 September 2011. Bilingual Experiments with an Arabic-English Corpus for Opinion Mining Mohammed Rushdi-Saleh SINAI research group University of Jaén msaleh@ujaen.es M. Teresa Martín-Valdivia SINAI research group University of Jaén maite@ujaen.es L. Alfonso Ureña-López SINAI research group University of Jaén laurena@ujaen.es José M. Perea-Ortega SINAI research group University of Jaén jmperea@ujaen.es Abstract Recently, Opinion Mining (OM) is receiving more at- tention due to the abundance of forums, blogs, e- commerce web sites, news reports and additional web sources where people tend to express their opinions. There are a number of works about Sentiment Analysis (SA) studying the task of identifying the polarity, whether the opinion expressed in a text is positive or negative about a given topic. However, most of research is focused on English texts and there are very few re- sources for other languages. In this work we present an Opinion Corpus for Arabic (OCA) composed of Arabic reviews extracted from specialized web pages related to movies and films using this language. Moreover, we have translated the OCA corpus into English, generating the EVOCA corpus (English Version of OCA). In the experiments carried out in this work we have used dif- ferent machine learning algorithms to classify the polar- ity in these corpora showing that, although the experi- ments with EVOCA are worse than OCA, the results are comparable with other English experiments, since the loss of precision due to the translation is very slight. 1. Introduction Nowadays, the interest in Opinion Mining (OM) has grown significantly due to different factors. On the one hand, the rapid evolution of the World Wide Web has changed our view of the Internet. It has turned into a collaborative framework where technological and social trends come together, resulting in the over exploited term Web 2.0. On the other hand, the tre- mendous use of e-commerce services has been ac- companied by an increase in freely available online reviews and opinions about products and services. A customer who wants to buy a product usually searches information on the Internet trying to find other con- sumer analyses. In fact, web sites such as Amazon1, Epinions2 or IMDb3, can affect the customer decision. 1 http://www.amazon.com 2 http://www.epinions.com Moreover, the automatic Sentiment Analysis (SA) is useful not only for individual customer but also for any company or institution. However, the huge amount of information makes necessary to accom- plish new methods and strategies to tackle the prob- lem. Thus, SA is becoming one of the main research areas that combines Natural Language Processing (NLP) and Text Mining (TM). This new discipline attempts to identify and analyze opinions and emo- tions. It includes several subtasks such as subjectivity detection, polarity classification, review summariza- tion, humor detection, emotion classification, senti- ment transfer, and so on [9]. However, most of works related to OM are oriented to use English language. Perhaps due to the novelty of the task, there are very few papers analyzing the opinions …

Content truncated to 500 words