A DEEP LEARNING APPROACH TO SENTIMENT ANALYSIS OF ENGLISH AND HAUSA TWEETS

ABSTRACT

Social media offers an opinionated platform of data contents about many topics of interest, such as product reviews, feedback on purchases or services, political interest, etc., that is dynamically created by users in different languages. Sentiment analysis in many languages is required due to the need for effective classification of these contents. However, a lot of contents are available in low resource languages like Hausa, but the majority of study is done on high-level languages like English, Arabic, German, etc. Similar to this, present models for multilingual sentiment analysis do not take into account the incorporation of language-specific features. The Convolutional Neural Network–Long Short-Term Memory (CNN-LSTM) model is used in this study to present a deep learning technique to multilingual sentiment analysis of the English and Hausa languages. The model incorporates a combination of CNN and LSTM models and pre-trained word embeddings from two languages (English and Hausa). The model has been tested on datasets that are monolingual (pure English and pure Hausa) and multilingual (English and Hausa combined). With 83.5 percent in the multilingual dataset and 81.8 percent in the pure Hausa dataset, the experimental result demonstrates the efficiency of the suggested methodology.

Keywords: Deep learning, social media, multilingual sentiment analysis, and sentiment analysis.