University of Bordj Bou Arreridj
Welcome to the Digital Repository of University Mohamed El Bachir El Ibrahimi - Bordj Bou Arreridj
Communities in University of Bordj Bou Arreridj Repository
Select a community to browse its collections.
Now showing 1 - 7 of 7
- كلية الحقوق والعلوم السياسية
- كلية الآداب واللغات
- كلية الرياضيات و الأعلام الآلي
- كلية علوم الطبيعة و الحياة و علوم الأرض و الكون
- كلية العلوم الاقتصادية والتجارية وعلوم التسيير
- Faculté des sciences et de la technologie
- كلية العلوم الاجتماعية والإنسانية
Recent Submissions
Course Handout Seminars and Literature Review (SLR)
(university of bordj bou arreridj, 2026) Farid BOUMAZA
Sentiment Analysis for Multilingual Datasets Using Transformers
(university of bordj bou arreridj, 2026) Brahimi Zineb; - Khramsia Khawla; Dahmouni Oumaima
The rapid proliferation of social media platforms has generated an unprecedented volume of User-Generated Content (UGC). This vast and diverse data landscape renders many existing Natural Language Processing (NLP) tools fundamentally inadequate for performing accurate automated sentiment analysis across different contexts. Consequently, this dissertation addresses this critical research gap by introducing three complementary contributions.
First, we construct and annotate a novel multilingual sentiment dataset collected from YouTube, focusing on major events that attracted global attention and explicitly designed to capture the complex nature of social media. Second, we conduct a systematic benchmarking study of state-of-the-art Transformer-based pre-trained language models, including mBERT, XLM-R, and DziriBERT, to identify the architecture most robust to the specificities of Algerian dialectal text. Third, we design and deploy a functional Software-as-a-Service (SaaS) platform integrating the best-performing model, offering real-time sentiment analysis via a web interface accessible to researchers and businesses.
SakinaPath: A DBT-InformedWell-being Application for Algerian Adults
(university of bordj bou arreridj, 2026) Ouarkoub Meriem; Benzeghiba Wail; Bendiab Kamel Lotfi Islam; Miloudi Mahdi
Mental health disorders impose a growing burden in Algeria, while access to care remains limited by shortages
of professionals, stigma, geographic barriers, and a multilingual landscape that existing digital tools
rarely address. No publicly identified mobile application offers Algerian adults structured, self-guided mental
health support grounded in an evidence-based framework and adapted to local linguistic and privacy
needs.
This dissertation addresses that gap through the design, implementation, and evaluation planning of SakinaPath,
a cross-platform mobile mental health application based on the principles of Dialectical Behaviour
Therapy (DBT). A review of existing international and regional solutions identified three recurring limitations
— linguistic inaccessibility, insufficient privacy protection, and a lack of structured self-guided skills
training — each informing a specific design decision.
The resulting prototype follows a three-tier architecture: a Flutter cross-platform client, a Supabase
PostgreSQL backend secured through Row-Level Security policies, and two external integrations (Groq API
for a non-clinical AI recommender; Chargily Pay for DZD payment processing). It implements three DBT
modules (mindfulness, distress tolerance, emotion regulation), fourteen psychometric assessments, a safetycritical
SOS pathway, a moderated peer community, and a verified psychiatrist booking and remote consultation
system, supported by a React-based administration panel. The interface supports Arabic, French, and
English, with right-to-left layout for Arabic.
This thesis contributes a contextualised requirements analysis for DBT-based digital mental health tools
in Algeria and a privacy-first architecture centered on database-level access control, demonstrated through a
functional Master’s-level prototype. Limitations include the absence of the interpersonal effectiveness DBT
module, reliance on LLM-based crisis detection, the lack of empirical evaluation results, and unresolved data
residency considerations.
Design and Development of an Intelligent Student Assistant Application
(university of bordj bou arreridj, 2026) Ben Merrouche Aymen; Hammani Abdelhak; Lakhal Zakariya
TalebDZ is an AI-powered assistant chatbot designed to support students at Algerian universities. The application addresses a recurring problem faced by university students: the difficulty of accessing accurate and timely information regarding administrative procedures, academic regulations, orientation, and university services. TalebDZ integrates a Retrieval-Augmented Generation pipeline built on Meta LLaMA 3.3 70B, accessed via the OpenRouter API, with a ChromaDB vector database and a LangChain orchestration layer to deliver grounded, contextually relevant responses. The system supports four languages (Modern Standard Arabic, Algerian Darija, English and French) reflecting the multilingual reality of Algerian higher education. A Flutter-based mobile application provides the user interface, backed by a Flask REST API and a Supabase PostgreSQL database for authentication and persistence. Evaluation results demonstrate a RAG accuracy of 93.3% across a 15-question multilingual benchmark and a mean user satisfaction score of 4.10 out of 5 across 10 participants, confirming the system's potential as a practical academic support tool