roberta-depression-detection
| Property | Value |
|---|---|
| Base Model | cardiffnlp/twitter-roberta-base |
| Task | Depression Detection Classification |
| Author | paulagarciaserrano |
| Model Hub | Hugging Face |
What is roberta-depression-detection?
This model is a specialized implementation of RoBERTa, fine-tuned for detecting signs of depression in social media text. It was trained on data from the Shared task on Detecting Signs of Depression from Social Media Text at LT-EDI 2022-ACL 2022, achieving a macro F1-score of 0.54 on the development set. The model performs multiclass classification, categorizing text into three depression severity levels: not depression, moderate, and severe.
Implementation Details
The model was trained using carefully selected hyperparameters, including a learning rate of 2e-05, weight decay of 0.01, and ran for 5 epochs. Training was conducted with batch sizes of 8 for both training and evaluation, using epoch-based evaluation and saving strategies.
- Training dataset includes 53,909 sentences across three classes
- Evaluation dataset comprises 73,414 sentences
- Implements advanced text classification architecture based on RoBERTa
Core Capabilities
- Multi-class depression severity classification
- Processes social media text effectively
- Easy integration using Hugging Face Transformers pipeline
- Handles varying text lengths (from short to long posts)
Frequently Asked Questions
Q: What makes this model unique?
This model specializes in detecting depression signs from social media text, offering three-level classification (not depression, moderate, severe) with balanced training across different text lengths and complexity levels. It's particularly noteworthy for its training on a diverse dataset with varying document lengths and word counts per class.
Q: What are the recommended use cases?
The model is ideal for analyzing social media content for depression indicators, mental health research, and automated mental health screening tools. However, it should be used as a supplementary tool rather than a primary diagnostic instrument, and always in conjunction with professional medical evaluation.
