Project Abstract

This project explores the creation of an AI-powered chatbot designed to support mental health and academic consultation, particularly for university students facing stress or anxiety. As more people seek accessible emotional support, traditional chatbots often fall short in understanding emotions or maintaining long conversations. This work aims to bridge that gap by delivering empathetic, context-aware responses using deep learning. The core proposition of this research lies in combining large language models with intent classification and sentiment analysis to better handle psychological support queries. The goal is to create a chatbot that can recognize emotional cues, accurately classify user intent, and respond appropriately in sensitive contexts. To achieve this, the system integrates a front-end web interface with a back-end powered by Python and Flask. Intent classification was performed using Qwen-based models, while the dialogue generation model was fine-tuned using LLaMAFactory with LoRA on real-world counseling data. A LangChain-based architecture manages conversation flow and ensures smooth user interaction. The system is currently at the prototype stage, with classification accuracy reaching 88% and dialogue generation being evaluated using Rouge-L scores. A web interface enables real-time user interaction, simulating real-world usage scenarios. Ultimately, this project hopes to contribute a scalable, emotionally intelligent chatbot that enhances digital mental health support. If successful, it could improve the way AI is applied in sensitive, human-centered domains like psychological care and education.

Keywords: Multimodal Chatbot, Deep Learning, Mental Health Support

 Conference Details

Session: B

Location: Sir Stanley Clarke Auditorium at 13:30 15:30

Markers: Lu Zhang, Yuanbo Wu

Course: BSc Computer Science 3yr FT

Future Plans: I’m looking for work