Sentiment Classification in Swahili Language Using Multilingual BERT

TL;DR

This paper applies multilingual BERT to perform sentiment classification on Swahili social media data, achieving high accuracy despite limited resources for African languages.

Contribution

It demonstrates the effectiveness of multilingual BERT for sentiment analysis in Swahili, a low-resource language, by creating and annotating a new dataset and fine-tuning the model.

Findings

Achieved 87.59% accuracy in sentiment classification.

Created and annotated 8.2k Swahili social media reviews.

Showed multilingual BERT's effectiveness for African languages.

Abstract

The evolution of the Internet has increased the amount of information that is expressed by people on different platforms. This information can be product reviews, discussions on forums, or social media platforms. Accessibility of these opinions and peoples feelings open the door to opinion mining and sentiment analysis. As language and speech technologies become more advanced, many languages have been used and the best models have been obtained. However, due to linguistic diversity and lack of datasets, African languages have been left behind. In this study, by using the current state-of-the-art model, multilingual BERT, we perform sentiment classification on Swahili datasets. The data was created by extracting and annotating 8.2k reviews and comments on different social media platforms and the ISEAR emotion dataset. The data were classified as either positive or negative. The model was fine-tuned and achieve the best accuracy of 87.59%.