FII-UAIC at SemEval-2020 Task 9: Sentiment Analysis for Code-Mixed Social Media Text Using CNN

Proceedings of the 14th International Workshop on Semantic Evaluation, pages 928–933 Barcelona, Spain (Online), December 12, 2020. 928 FII-UAIC at SemEval-2020 Task 9: Sentiment Analysis for CodeMixed Social Media Text using CNN Lavinia Aparaschivei1, Andrei Palihovici1, Daniela Gîfu1,2 1Faculty of Computer Science, “Alexandru Ioan Cuza” University of Iasi 2Institute of Computer Science, Romanian Academy - Iasi Branch { lavinia.n.aparaschivei, andrei.r.palihovici, daniela.gifu}@info.uaic.ro Abstract The “Sentiment Analysis for Code-Mixed Social Media Text” task at the SemEval 2020 com- petition focuses on sentiment analysis in code-mixed social media text1, specifically, on the combination of English with Spanish (Spanglish) and Hindi (Hinglish). In this paper, we pre- sent a system able to classify tweets, from Spanish and English languages, into positive, nega- tive and neutral. Firstly, we built a classifier able to provide corresponding sentiment labels. Besides the sentiment labels, we provide the language labels at the word level. Secondly, we generate a word-level representation, using Convolutional Neural Network (CNN) architec- ture. Our solution indicates promising results for the Sentimix Spanglish-English task (0.744), the team, Lavinia_Ap, occupied the 9th place. However, for the Sentimix Hindi-English task (0.324) the results have to be improved. 1 Introduction The explosive growth of social media (SM) platforms offers a new type of user, multilingual, that can alternate two languages or more in the same conversation (Udupa and Khapra, 2010). In linguistics, it is call Code-Switching (Adel et al., 2013) or language alternation. Any virtual communication channel creates the possibility for people from different countries to share their sentiments without restrictions, about anything (Gîfu and Cioca, 2014), often using multiple languages. This explosion of sentiments has aroused the interest of many researchers for sentiment analysis (SA) for code-mixed SM message. Research on SA has focused, especially, on understanding the dynamics of sentiment in SM, most of them choosing Twitter because it provides free API very useful for data retrieval goal. That facility allows the developer to find real time tweets from different multilingual users. Actually, the analysis of tweets, which includes language alternation, is a challenging Natural Language Processing (NLP) is- sue. Code-mixing (CM) has several challenges to apply NLP techniques, such as word-level language identification or semantic processing (Myers-Scotton, 1993). The goal of this paper is to implement a model for CM content on Twitter, which imply two objec- tives: first, the classification of positive, negative, and neutral tweets by generating word-level repre- sentation, using Convolutional Neural Network (CNN). The Spanish-English code-mixed content has become ubiquitous on the Internet, creating the need to process this form of natural language. A code- mixed sentence retains the underlying grammar and script of one of the languages it is comprised of. The legitimate question of this survey is: Can we achieve results comparable with those of our peers using more standard, less customized techniques? This work is licensed under a Creative Commons Attribution 4.0 International Licence. Licence details: http://creativecommons.org/licenses/by/4.0/. 929 The rest of the paper is structured as follows: section 2 describes other works related to senti- ment analysis for code-mixed …

Content truncated to 500 words