|
|
--- |
|
|
library_name: transformers |
|
|
tags: |
|
|
- text-classification |
|
|
- healthcare |
|
|
- intent-detection |
|
|
- transformers |
|
|
license: apache-2.0 |
|
|
datasets: |
|
|
- BinKhoaLe1812/MedDialog-EN-100k |
|
|
language: |
|
|
- en |
|
|
metrics: |
|
|
- accuracy |
|
|
- recall |
|
|
- f1 |
|
|
base_model: |
|
|
- FacebookAI/roberta-base |
|
|
--- |
|
|
|
|
|
# Model Card for Model ID |
|
|
|
|
|
<!-- Provide a quick summary of what the model is/does. --> |
|
|
**Dala Intent Model** 🧠💬 |
|
|
Model Overview |
|
|
|
|
|
The Dala Intent Model is a Transformer-based classifier that maps patient symptom queries to predefined intents. |
|
|
It is designed as part of the Dala AI Symptom Checker to help structure healthcare conversations for further reasoning. |
|
|
|
|
|
### Model Description |
|
|
This is the model card of a 🤗 transformers model that has been pushed on the Hub. This model card has been automatically generated. |
|
|
|
|
|
- **Developed by:** @jaywestty |
|
|
- **Model type:** BERT-like transformer (fine-tuned for text classification) |
|
|
- **Language(s) (NLP):** English |
|
|
- **License:** Apache-2.0 |
|
|
## Uses |
|
|
✅ **Intended Uses**: |
|
|
* Classifying patient symptom descriptions into healthcare intents |
|
|
* Assisting conversational AI in guiding users toward possible next steps |
|
|
|
|
|
⚠️ **Limitations**: |
|
|
* Not a diagnostic tool |
|
|
* Should not replace professional medical advice |
|
|
* Performance may vary on domains outside the training dataset |
|
|
|
|
|
## How to Get Started with the Model |
|
|
<pre> ```from transformers import AutoTokenizer, AutoModelForSequenceClassification |
|
|
import torch |
|
|
|
|
|
model_name = "Jayywestty/dala-intent-model" |
|
|
tokenizer = AutoTokenizer.from_pretrained(model_name) |
|
|
model = AutoModelForSequenceClassification.from_pretrained(model_name) |
|
|
|
|
|
text = "I have chest pain and shortness of breath" |
|
|
inputs = tokenizer(text, return_tensors="pt") |
|
|
outputs = model(**inputs) |
|
|
predicted_class = torch.argmax(outputs.logits, dim=-1).item() |
|
|
|
|
|
print("Predicted intent:", predicted_class)```</pre> |
|
|
|
|
|
### Training Data |
|
|
* **Dataset**: Proprietary healthcare dataset (10k examples) |
|
|
* **Split**: 80% train / 10% validation / 10% test |
|
|
* **Optimizer**: (lr = 3e-5) |
|
|
* **Batch size**: 16 |
|
|
* **Epochs**: 3 |
|
|
* **Evaluation metrics**: Accuracy, F1 |
|
|
|
|
|
|
|
|
## Evaluation |
|
|
|
|
|
| Metric | Score | |
|
|
| -------- | ----- | |
|
|
| Accuracy | 0.85 | |
|
|
| F1 Score | 0.86 | |
|
|
|
|
|
|
|
|
|
|
|
## Citation |
|
|
@misc{dala-intent-model, |
|
|
|
|
|
author = {Fadairo, Oluwajuwon}, |
|
|
|
|
|
title = {Dala Intent Model: Transformer for Healthcare Intent Classification}, |
|
|
|
|
|
year = {2025}, |
|
|
|
|
|
publisher = {Hugging Face}, |
|
|
|
|
|
howpublished = {\url{https://huggingface.co/Jayywestty/dala-intent-model}} |
|
|
} |
|
|
|
|
|
|
|
|
## Model Card Authors |
|
|
Author: Fadairo Oluwajuwon |
|
|
|
|
|
## Model Card Contact |
|
|
Email: juwonfadairo13@gmail.com |
|
|
GitHub: [jaywestty](https://github.com/jaywestty) |
