Title: Sentiment analysis of customer reviews for Algerian dialect using the DziriBERT model

Authors: Fateh Bougamouza; Samira Hazmoune

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: The increasing volume of daily comments and tweets presents a valuable resource for improving various processes, from business strategies to service management. However, the Algerian Dialect (AlgD), despite its growing presence on social media, has been overlooked in sentiment analysis. This study addresses this gap by proposing an approach for sentiment analysis of Algerian Dialect feedback, specifically from customers of Algerian telephone operators (Djezzy, Mobilis, and Ooredoo). Leveraging transfer learning, the pre-trained DziriBERT model was fine-tuned, with experiments refining data preprocessing techniques and hyperparameters. The outcome is an impressive 82.01% accuracy rate, offering promising insights into sentiment analysis in the Algerian Dialect and highlighting its potential significance for companies and researchers in the field.

Keywords: sentiment analysis; Algerian Arabic dialect; DziriBERT; transfer learning; Algerian telephone operators; emoji categorisation.

DOI: 10.1504/IJDATS.2024.140647

International Journal of Data Analysis Techniques and Strategies, 2024 Vol.16 No.3, pp.341 - 362

Received: 04 Oct 2023
Accepted: 23 Jan 2024
Published online: 29 Aug 2024 *

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