Tools
Tools
Waghela, Meet Anil (2025) Multimodal Depression Detection in Overseas Students Using Voice and Text Analytics: A Scalable Cloud-Based Machine Learning Approach. Masters thesis, Dublin, National College of Ireland.
Preview |
PDF (Master of Science)
Download (2MB) | Preview |
Preview |
PDF (Configuration Manual)
Download (508kB) | Preview |
Abstract
The issue of depression among overseas students is an acute systemic issue of mental health, and about 45% of international students experience depressive symptoms in relation to much lower percentages in domestic groups. This study proposes a universal multimodal machine learning framework, which combines voice and text analysis to automate the detection of depression that can work with the international students population. The project achieves a proof of concept project with a combination of state-of-the-art approaches to natural language processing such as BERT embeddings and acoustic representation extraction with Mel-Frequency Cepstral Coefficients (MFCC) and spectral analysis. This system architecture has three different classification models including one individual model based on text and one model on audio and a complex fusion model based on BERT model meta-learning.
A significant degree of performance was obtained when this model was experimentally validated on a stratified dataset of DAIC-WOZ with subset of samples from a comprehensive mental health dataset achieving training accuracy of 93.0% and testing model fusion accuracy of 88.3%, which is a significant improvement over its single-modal text-only (98.2%) and audio-only (84.1%) counterparts. The study offers a scalable cloud-ready framework design, extensive exploratory data analysis that have demonstrated the presence of unique linguistic and prosodic patterns in communications that involve depression, and empirical evidence of the effectiveness of multimodal methods compared to unimodal ones. These results show that automated mental health screening tools have potential in offering early intervention to vulnerable groups of international students but ethical concerns about privacy, bias, and clinical integration are the priorities when considering deployment in a real-life setting using Microsoft Azure.
| Item Type: | Thesis (Masters) |
|---|---|
| Supervisors: | Name Email Agarwal, Bharat UNSPECIFIED |
| Uncontrolled Keywords: | Multimodal Machine Learning; Depression Detection; Natural Language Processing; Voice Analytics; International Students; Cloud Computing; Mental Health Technology |
| Subjects: | T Technology > T Technology (General) > Information Technology > Cloud computing P Language and Literature > P Philology. Linguistics > Computational linguistics. Natural language processing Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning R Medicine > RA Public aspects of medicine > RA790 Mental Health |
| Divisions: | School of Computing > Master of Science in Data Analytics |
| Depositing User: | Ciara O'Brien |
| Date Deposited: | 27 Aug 2026 09:27 |
| Last Modified: | 27 Aug 2026 09:27 |
| URI: | https://norma.ncirl.ie/id/eprint/9683 |
Actions (login required)
 |
View Item |