A neural approach for sentiment analysis in Algerian dialect
A neural approach for sentiment analysis in Algerian dialect
Fichiers
Date
2025-05-29
Auteurs
HAMADOUCHE Khaoula
Nom de la revue
ISSN de la revue
Titre du volume
Éditeur
Université oran1
Résumé
Our research explores several fundamental aspects of natural language processing for sentiment analysis, with a focus on automatic corpus construction and sarcasm detection within the context of Algerian dialect.Specifically, we have adopted state-of-the-art techniques in deep learning, namely the BERT model (Bidirectional- Encoder Representations from Transformers). This research aims to address key challengesrelatedtounderstandingandprocessing linguistic nuances in Algerian Arabictextsbydesigningaplatformthat collects data from various social media sources and conducts sentiment analysistasks(polaritydeterminationandsarcasm detection).
Methodologically,ourthesisemploys techniques including: (1) corpus extraction from social media, (2) the application of machine learning algorithms for sentiment classification, and (3) the integration of linguistic rules and deep learning models for sarcasm detection.
The approaches we have used have allowed us to achieve satisfactory results that can extend to various applications in social media analysis, opinion extraction, and automated content analysis within arabophone communities, including the study of dialectal variations.
This thesis underscores the importance of linguistic diversity in NLP researchand provides a foundation for future investigations in the field of dialectal sentiment analysis.
Description
Mots-clés
Natural Language Processing ; Sentiment Analysis ; Algerian Dialect ; Corpus Construction ; BERT ; sarcasm; deep learning; machine learning; Arabic natural processing ; WORD emberedding