Update README.md
Browse files
README.md
CHANGED
|
@@ -5,4 +5,42 @@ language:
|
|
| 5 |
pipeline_tag: text-classification
|
| 6 |
tags:
|
| 7 |
- sentiment-analysis
|
| 8 |
-
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 5 |
pipeline_tag: text-classification
|
| 6 |
tags:
|
| 7 |
- sentiment-analysis
|
| 8 |
+
---
|
| 9 |
+
|
| 10 |
+
---
|
| 11 |
+
language: en
|
| 12 |
+
license: apache-2.0
|
| 13 |
+
pipeline_tag: text-classification
|
| 14 |
+
tags:
|
| 15 |
+
- sentiment-analysis
|
| 16 |
+
- mental-health
|
| 17 |
+
- transformers
|
| 18 |
+
- huggingface
|
| 19 |
+
- ai-chatbot
|
| 20 |
+
---
|
| 21 |
+
|
| 22 |
+
# 🧠 MindGuardAI – Sentiment Analysis Model
|
| 23 |
+
|
| 24 |
+
This model is part of the **MindGuardAI** project, a mental health–focused chatbot system. It performs **sentiment analysis** on user input text to classify emotional tone as **positive** or **negative**, helping assess user mood and emotional state in real time.
|
| 25 |
+
|
| 26 |
+
## ✨ Model Highlights
|
| 27 |
+
|
| 28 |
+
- **Task**: Sentiment classification (binary: `Positive`, `Negative`)
|
| 29 |
+
- **Trained On**: Social media mental health datasets (Twitter, Reddit-style samples)
|
| 30 |
+
- **Model Type**: Fine-tuned BERT-based model using the Hugging Face `transformers` library
|
| 31 |
+
- **Use Case**: Integrates with mental health chatbot systems for mood tracking and emotional awareness
|
| 32 |
+
|
| 33 |
+
---
|
| 34 |
+
|
| 35 |
+
## 🚀 Quick Start
|
| 36 |
+
|
| 37 |
+
### 🧪 Inference API
|
| 38 |
+
|
| 39 |
+
Try it out instantly using Hugging Face's hosted API 👇
|
| 40 |
+
|
| 41 |
+
```python
|
| 42 |
+
from transformers import pipeline
|
| 43 |
+
|
| 44 |
+
pipe = pipeline("text-classification", model="hasanmustafa0503/MindGuardAI")
|
| 45 |
+
result = pipe("I feel really low and unmotivated today.")
|
| 46 |
+
print(result)
|