This repository contains code for training and using deep learning models for mental disorder detection in conversational domain.
The required libraries are:
- tensorflow2, keras
- numpy, nltk, sklearn, pandas
The full list of packages and versions I used is found in requirements.txt (may contain some unnecessary ones)
See scripts folder. Main flow is precompute representation and then apply on your dataset.
If using this resource, please cite the relevant papers (bib format will be provided as soon as possible):
- Detecting early signs of depression in the conversational domain: The role of transfer learning in low-resource scenarios
- Petr Lorenc, Ana Sabina Uban, Paolo Rosso, and Jan Šedivý, NLDB 2022
The code and algorithm was inspired by:
@InProceedings{10.1007/978-3-030-80599-9_27,
author="Uban, Ana Sabina
and Chulvi, Berta
and Rosso, Paolo",
title="On the Explainability of Automatic Predictions of Mental Disorders from Social Media Data",
booktitle="Natural Language Processing and Information Systems",
year="2021",
publisher="Springer International Publishing",
pages="301--314",
isbn="978-3-030-80599-9"
}
All experiment were conducted during internship (May - September 2021) of Petr Lorenc at Universitat Politècnica de València supervised by professor Paolo Rosso