Title: Emoji translation for sentiment analysis in Algerian Arabic dialect

Authors: Samira Hazmoune; Fateh Bougamouza

Addresses: Faculty of Sciences, Department of Computer Science, University of 20 août 1955-Skikda, Skikda, 21000, Algeria ' Faculty of Sciences, Department of Computer Science, University of 20 août 1955-Skikda, Skikda, 21000, Algeria

Abstract: Sentiment analysis (SA) is an important natural language processing (NLP) field that involves extracting sentiments and opinions from text data. Although SA has advanced significantly, its application to dialectal Arabic text presents challenges due to linguistic nuances and resource constraints. This research investigates the incorporation of emojis into SA for Algerian Arabic dialect (AAD), marking the first exploration of its kind in this area. Specifically, we focus on emoji translation, building upon prior studies highlighting emojis, potential in SA and their translation into meaningful words or sentences as a preprocessing approach. We evaluate the impact of this approach on enhancing sentiment classification in AAD text, specifically focusing on customer reviews of Algerian telephone operators. After preprocessing, including various emoji translation techniques, we employ transfer learning by fine-tuning DziriBERT model on a compiled Algerian dialect dataset. Our results demonstrate promising outcomes and offer novel conclusions and perspectives in AAD sentiment analysis.

Keywords: sentiment analysis; emoji translation; DziriBERT; AAD; Algerian Arabic dialect; transfer learning; emoji categorisation; emoji handling; customer reviews.

DOI: 10.1504/IJDATS.2025.148561

International Journal of Data Analysis Techniques and Strategies, 2025 Vol.17 No.3, pp.216 - 237

Received: 12 May 2024
Accepted: 21 Jul 2024
Published online: 12 Sep 2025 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article