LIMSI_UPV at SemEval-2020 Task 9: Recurrent Convolutional Neural Network for Code-mixed Sentiment Analysis

Proceedings of the 14th International Workshop on Semantic Evaluation, pages 1281–1287 Barcelona, Spain (Online), December 12, 2020. 1281 LIMSI UPV at SemEval-2020 Task 9: Recurrent Convolutional Neural Network for Code-mixed Sentiment Analysis Somnath Banerjee1, Sahar Ghannay1, Sophie Rosset1, Anne Vilnat1, Paolo Rosso2 1 LIMSI, CNRS, Universit´e Paris-Saclay, Orsay, France 2 Universitat Polit`ecnica de Val`encia, Spain 1firstname.lastname@limsi.fr 2prosso@dsic.upv.es Abstract This paper describes the participation of LIMSI UPV team in SemEval-2020 Task 9: Sentiment Analysis for Code-Mixed Social Media Text. The proposed approach competed in SentiMix Hindi- English subtask, that addresses the problem of predicting the sentiment of a given Hindi-English code-mixed tweet. We propose Recurrent Convolutional Neural Network that combines both the recurrent neural network and the convolutional network to better capture the semantics of the text, for code-mixed sentiment analysis. The proposed system obtained 0.69 (best run) in terms of F1 score on the given test data and achieved the 9th place (Codalab username: somban) in the SentiMix Hindi-English subtask. 1 Introduction In this digital era, users express their personal thoughts and opinions regarding a wide range of topics on social media platforms such as blogs, micro-blogs (e.g., Twitter), and chats (e.g., WhatsApp and Facebook messages). Multilingual societies like India with a decent amount of internet penetration widely adopted such social media platforms. However, the regional language influences the proliferation of the Hindi-English Code-Mixed (CM) data. Sentiment analysis of these end-user data from social media is a crucial resource for commerce and governance. However, in contrast to the classical sentiment analysis methods, which were originally designed for dealing with well-written product reviews, CM texts from social media often contain misspellings (often intentional), badly cased words, letter substitutions, ambiguities, non standard abbreviations, improper use of grammar, etc. CM poses several unseen difficulties to natural language processing (NLP) tasks such as word-level language identification, part-of-speech tagging, dependency parsing, machine translation and semantic processing. In the last few years, a number of workshops such as Linguistic Code-Switching Workshops1 and shared tasks such as Mixed Script Information Retrieval (Banerjee et al., 2020) have been organized due to the emerging popularity of code-mixing. To promote research in this area, Task 9 of SemEval-2020 was devoted to CM sentiment analysis in Twitter. The goal of the task was to automatically classify the polarity of a given CM Twitter post into one of the three predefined categories: positive, negative and neutral. The CM languages are English-Hindi and English-Spanish; for a more detailed description of the task see (Patwa et al., 2020). In this paper, we present a deep learning approach, using a Recurrent Convolutional Neural Network for the task of automatic CM sentiment classification of tweets. The rest of the paper is structured as follows. Section 2 provides background in brief. Section 3 provides the system overview and Section 4 describes our approach in detail. In Section 5, we discuss the analysis and evaluation results for our system. We conclude our work in Section 6. This work is licensed under a Creative Commons Attribution 4.0 International Licence. …

Content truncated to 500 words