Multilingual Sentiment Analysis: A Deep Learning Approach

Saad Mboutayeb, Aicha Majda, Nikola S. Nikolov

2021

Abstract

Most user-generated text on social media is not in English but other languages such as Arabic, French, and Portuguese. This makes the text analysis tasks more difficult, especially sentiment analysis, because of its high dependence on the language. On the other hand, building a model for each language is time and resources consuming. In particular, there is a lack of linguistic resources such as datasets. In this paper, we examine if a sentiment analysis model trained on one language can correctly predict the sentiment of text originally written in another language and translated into the model's language. We present experimental results of training CNN, RNN and combined CNN-RNN models on a dataset of multilingual tweets. Our findings suggest that CNN gives the best results with an accuracy of 85.91% and an F1-score of 84.61%. Our best model also achieved high accuracy on unseen tweets in European languages different from the original languages of the tweets used for training.

Paper Citation

in Harvard Style

Mboutayeb S., Majda A. and Nikolov N. (2021). Multilingual Sentiment Analysis: A Deep Learning Approach. In Proceedings of the 2nd International Conference on Big Data, Modelling and Machine Learning - Volume 1: BML, ISBN 978-989-758-559-3, pages 27-32. DOI: 10.5220/0010727700003101

in Bibtex Style

@conference{bml21,
author={Saad Mboutayeb and Aicha Majda and Nikola S. Nikolov},
title={Multilingual Sentiment Analysis: A Deep Learning Approach},
booktitle={Proceedings of the 2nd International Conference on Big Data, Modelling and Machine Learning - Volume 1: BML,},
year={2021},
pages={27-32},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0010727700003101},
isbn={978-989-758-559-3},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 2nd International Conference on Big Data, Modelling and Machine Learning - Volume 1: BML,
TI - Multilingual Sentiment Analysis: A Deep Learning Approach
SN - 978-989-758-559-3
AU - Mboutayeb S.
AU - Majda A.
AU - Nikolov N.
PY - 2021
SP - 27
EP - 32
DO - 10.5220/0010727700003101