Browsing by Author "Supervisor: Said, Gadri"

Now showing 1 - 4 of 4
  • Thumbnail Image
    ItemOpen Access
    Designing and developing an Intelligent Chatbot for Product Inquiry Analysis in E-Stores Using NLP and LLMS Models
    (Mohamed Boudiaf University of M'sila, 2025-06-15) Aymen, Gasmi; Supervisor: Said, Gadri
    E-commerce is undergoing a paradigm shift thanks to the integration of artificial intelligence technologies, particularly intelligent chatbots, which enhance user experience and provide im mediate and efficient customer support. This research presents the development of a multilin gual chatbot, based on Retrieval Augmented Generation (RAG) technology and Large Language Models (LLMs), to provide accurate and contextual answers in Arabic, French, and English. The system contributes to reducing the burden on human support teams and increasing customer sat isfaction, with the ability to dynamically update the knowledge base. The model demonstrates high adaptability, making it applicable to multiple fields beyond e-commerce.
  • Thumbnail Image
    ItemOpen Access
    Facial Emotion Recognition Using Deep Learning Approach
    (Mohamed Boudiaf University of M'sila, 2025-06-15) Douaa Hanan, Kadi; Nada Erayhane, Heltali; Supervisor: Said, Gadri
    This thesis aims to design and implement a system for recognizing human emotions based on facial expressions, using artificial intelligence techniques, particularly Convolutional Neural Networks (CNN). Emotion recognition is an emerging and important field within affective computing, with broad applications in mental health, education, marketing, surveillance systems, and human-computer interaction. In this work, a facial image dataset containing various emotions (such as anger, happiness, sadness, surprise...) was used. The images underwent preprocessing steps such as grayscale conversion and resizing. The CNN model was then trained using environments like JupyterLab and Google Colab, with tools such as TensorFlow and Keras used for model design and evaluation. The system consisted of the following main stages:  Face Detection  Feature Extraction  Emotion Classification The results showed good accuracy in emotion recognition, confirming the effectiveness of the proposed model. A simple application interface was also developed to test the model on both live and stored images, bringing the project closer to real-world applications. Despite the promising results, some challenges remain, such as lighting conditions, facial angle variations, and similarities between emotional expressions. This opens the door for more advanced future work, such as integrating multiple modalities (voice, text, facial expression) or adopting more powerful models trained on more diverse datasets.
  • Thumbnail Image
    ItemOpen Access
    Smart Irrigation Based on Soil Moisture Detection
    (Mohamed Boudiaf University of M'sila, 2025-06-15) Somia, Dilmi; Supervisor: Said, Gadri
    This work presents the development of a smart irrigation system based on monitoring soil moisture and air temperature. The objective is to reduce water waste and improve crop productivity by using an Arduino board and low-cost sensors to automate irrigation. The system delivers water only when needed, based on real-time environmental data. Although artificial intelligence was not fully integrated, the study explores its potential application in future versions. Additionally, a farmer awareness component was proposed to promote the adoption of such technologies. The results show that the system is functional, cost-effective, and suitable for small and medium-sized farms.
  • Thumbnail Image
    ItemOpen Access
    Speech Emotion Recognition Using Deep Learning Models
    (Mohamed Boudiaf University of M'sila, 2025-06-15) Rania, Agoune; Supervisor: Said, Gadri
    This project addresses the recognition of human emotions through speech using artificial intelligence. A CNN-based model was developed to classify emotions like happiness, anger, and sadness from features such as MFCC. Datasets like RAVDESS, SAVEE, and CASIA were combined to enhance diversity. Data augmentation techniques improved model generalization. The model achieved over 96% accuracy. A Flask-based web interface enables training, testing, and real time prediction. The system offers promising applications in mental health, education, and customer service. This work lays the groundwork for emotionally aware and intelligent interaction systems