🧠 Finetuned BERT for Mental Health Text Classification

This model is a fine-tuned version of bert-base-uncased for detecting mental health-related categories from textual content. It classifies user-generated posts into five categories:

🟦 Depression
🟨 Anxiety
🔴 Suicidal
🟩 Addiction
🟪 Eating Disorder

It is designed to support research, digital therapy tools, and emotion-aware systems.

📁 Model Details

Base model: bert-base-uncased
Fine-tuned on: A custom dataset derived from multiple Kaggle sources
Classification type: Single-label (one of five categories)
Loss function: Cross-entropy
Framework: PyTorch, Hugging Face Transformers

🧪 Dataset

The dataset used for training and testing was compiled from multiple Kaggle sources involving real-world discussions related to mental health. It contains posts categorized into the five emotion/mental-health topics.

Training samples were selected from five original CSV files and combined into a single file.
Testing data was kept separate and sourced similarly.

📦 You can find the dataset here: Noobie314/mental-health-posts-dataset

🛠️ How to Use

from transformers import AutoTokenizer, AutoModelForSequenceClassification

model_name = "Noobie314/finetuned-emotion-bert"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForSequenceClassification.from_pretrained(model_name)

text = "I'm feeling hopeless and tired of everything..."
inputs = tokenizer(text, return_tensors="pt")
outputs = model(**inputs)

predicted_class = outputs.logits.argmax(dim=1).item()

📊 Evaluation

The model was evaluated on a test set of 10,000 rows with the following results:

Category	Precision	Recall	F1-Score	Support
Addiction	0.914	0.911	0.912	1000
Anxiety	0.831	0.808	0.819	1996
Depression	0.772	0.621	0.688	3990
Eating Disorder	0.916	0.921	0.919	1000
Suicidal	0.531	0.752	0.622	1994
Accuracy	0.744
Macro avg	0.793	0.803	0.792	9980
Weighted avg	0.764	0.744	0.747	9980

Accuracy: 74.36%
Macro Average F1-Score: 0.792
Weighted Average F1-Score: 0.747

These results indicate the model performs relatively well, with higher precision and recall for categories like Eating Disorder and Addiction.

✅ Intended Uses

This model is intended for:

NLP research in mental health domains
Emotion-aware digital assistants
Content moderation tools focused on emotional tone and risk

⚠️ Disclaimer: This model is not suitable for clinical or medical decision-making. It does not replace licensed mental health professionals.

📜 License

This model is licensed under the Apache 2.0 License.

📬 For questions, suggestions, or collaborations, feel free to open an issue or contact via the Hugging Face Hub.

Noobie314
/

finetuned-emotion-bert