Integration of Lexical and Semantic Knowledge for Sentiment Analysis in SMS

Integration of Lexical and Semantic Knowledge for Sentiment Analysis in SMS Wejdene Khiari∗,∗∗,∗∗∗, Asma Bouhafs∗∗, Mathieu Roche∗∗∗ ∗ESC Tunis School of Business, University of Manouba, Tunisia ∗∗University of Carthage, Carthage Presidency, Tunisia ∗∗∗UMR TETIS (Cirad, Irstea, AgroParisTech) & LIRMM (CNRS, Univ. Montpellier), France wijdenkhiari@gmail.com, asma bouhafs@yahoo.com, mathieu.roche@cirad.fr Abstract With the explosive growth of online social media (forums, blogs, and social networks), exploitation of these new information sources has become essential. Our work is based on the sud4science project. The goal of this project is to perform multidisciplinary work on a corpus of authentic SMS, in French, collected in 2011 and anonymised (88milSMS corpus: http://88milsms.huma-num.fr). This paper highlights a new method to integrate opinion detection knowledge from an SMS corpus by combining lexical and semantic information. More precisely, our approach gives more weight to words with a sentiment (i.e. presence of words in a dedicated dictionary) for a classification task based on three classes: positive, negative, and neutral. The experiments were conducted on two corpora: an elongated SMS corpus (i.e. repetitions of characters in messages) and a non-elongated SMS corpus. We noted that non-elongated SMS were much better classified than elongated SMS. Overall, this study highlighted that the integration of semantic knowledge always improves classification. Keywords: Sentiment analysis, SMS corpus, lexical and semantic information. 1. Introduction Internet has evolved boundlessly over the last decade with the advent of the social Web (Web 2.0). This has led to the development of new media such as various social networks ranging from Twitter, Facebook, Google+ and LinkedIn. These web sites offer opportunities for users to express themselves, as well as to exchange opinions and ideas with others through multiple platforms such as microblogs, blogs, web sites, SMS, emails, etc. Automatic analysis of texts generated from these communication modes for opinion detection is a real challenge in the field of opinion mining. Our work is under way in this context. The sms4science project is coordinated by CENTAL (Centre for Natu- ral Language Processing) at the Catholic University of Louvain, Belgium. The goal of the sud4science project (Panckhurst et al., 2013) is to perform multidisciplinary work on a corpus of 88.522 authentic SMS, in French, collected in 2011 and anonymised1 (88milSMS corpus: http://88milsms.huma-num.fr). In this paper, we present an opinion mining process that considers the specificities of SMS. The proposed paper is organized as follows. In Section 2, we describe our work related to sentiment identification in short texts, like those found in SMS and tweets. In Section 3, we detail the overall methodology of our knowledge integration process based on two strategies, automatic and manual, for annotation of the corpus. This annotation process confirmed the difficul- ties involved in sentiment analysis and feature identifica- 1The purpose of this anonymisation is to mask the identity of individuals (Panckhurst et al., 2013). The following tags were used for the anonymisation process: PRE (First Name), NOM (Last Name), SUR (Nickname), ADR (Address), LIE (Place), TEL (Telephone Number), COD (Code), URL (URL), MAR (Brand Name), MEL (Email), Other. tion …

Content truncated to 500 words