|
--- |
|
library_name: transformers |
|
license: creativeml-openrail-m |
|
base_model: |
|
- dmis-lab/biobert-v1.1 |
|
pipeline_tag: question-answering |
|
--- |
|
|
|
library_name: transformers |
|
tags: [biomedical, question-answering, healthcare] |
|
|
|
--- |
|
|
|
# Model Card for Online Doctor Model |
|
|
|
This model is a fine-tuned version of the `dmis-lab/biobert-large-cased-v1.1-squad` model. It is designed to answer questions related to diseases based on symptom descriptions, providing a question-answering pipeline to help healthcare professionals and users. This model has been trained on a custom dataset of diseases and their symptoms for predictive question answering. |
|
|
|
## Model Details |
|
|
|
### Model Description |
|
|
|
This is a question-answering model fine-tuned using the `BioBERT` architecture, specifically adapted for healthcare-related questions. The model is designed to extract answers from a disease-symptom dataset based on user-inputted symptoms or queries. |
|
|
|
- **Developed by:** Ayamba Victor Ndoma |
|
- **Model type:** Question Answering |
|
- **Language(s) (NLP):** English |
|
- **License:** Apache 2.0 |
|
- **Finetuned from model:** `dmis-lab/biobert-large-cased-v1.1-squad` |
|
|
|
### Model Sources |
|
|
|
- **Repository:** [Hugging Face Repository](https://huggingface.co/your-username/online-doctor-model) |
|
- **Demo:** [Link to Demo (optional)] |
|
|
|
## Uses |
|
|
|
This model can be used in the following cases: |
|
|
|
### Direct Use |
|
|
|
- Answering healthcare-related questions based on symptom input from users. |
|
- Assisting medical professionals in preliminary diagnosis based on reported symptoms. |
|
|
|
### Downstream Use |
|
|
|
- Can be further fine-tuned or extended for more specific disease or symptom-related tasks. |
|
- Integrated into chatbot systems for medical consultation services. |
|
|
|
### Out-of-Scope Use |
|
|
|
- The model is not intended for use in making definitive medical diagnoses without human supervision. |
|
- It is not suitable for predicting non-health-related issues. |
|
|
|
## Bias, Risks, and Limitations |
|
|
|
- **Bias:** The model is trained on a custom dataset with potentially limited diversity in disease-symptom pairs. |
|
- **Risks:** Incorrect predictions might occur when symptoms overlap across multiple diseases. |
|
- **Limitations:** The model is constrained to the diseases and symptoms available in the training dataset and may not generalize to all medical conditions. |
|
|
|
### Recommendations |
|
|
|
This model should be used with caution, and its answers should be reviewed by qualified healthcare professionals. |
|
|
|
## How to Get Started with the Model |
|
|
|
Use the following code to get started with the model: |
|
|
|
```python |
|
from transformers import pipeline |
|
qa_pipeline = pipeline("question-answering", model="your-username/online-doctor-model") |
|
|
|
# Example question and context |
|
question = "What are the symptoms of diabetes?" |
|
context = "Diabetes: increased thirst, frequent urination, hunger, fatigue, blurred vision." |
|
|
|
result = qa_pipeline(question=question, context=context) |
|
print(result['answer']) |
|
``` |
|
|
|
## Training Details |
|
|
|
### Training Data |
|
|
|
The model is fine-tuned on a custom dataset containing diseases and their respective symptoms. |
|
|
|
### Training Procedure |
|
|
|
- **Preprocessing:** Text cleaning and tokenization were applied to ensure proper context and symptom pairing. |
|
- **Training regime:** The model was trained using mixed-precision FP16 on a single GPU. |
|
|
|
#### Training Hyperparameters |
|
|
|
- **Epochs:** 3 |
|
- **Batch size:** 16 |
|
- **Learning rate:** 3e-5 |
|
|
|
## Evaluation |
|
|
|
### Testing Data, Factors & Metrics |
|
|
|
#### Testing Data |
|
|
|
The model was evaluated on a held-out portion of the custom disease-symptom dataset. |
|
|
|
#### Factors |
|
|
|
- Subpopulation: Various diseases ranging from common illnesses to rare conditions. |
|
- Domains: Medical text and descriptions of symptoms. |
|
|
|
#### Metrics |
|
|
|
The model was evaluated using the SQuAD metrics, including F1 score and Exact Match (EM). |
|
|
|
### Results |
|
|
|
- **F1 score:** 0.82 |
|
- **Exact Match (EM):** 0.78 |
|
|
|
#### Summary |
|
|
|
The model performs well on the task of extracting relevant symptoms and disease-related answers based on the question provided. However, its performance is limited to the diseases and symptoms present in the training data. |
|
|
|
## Environmental Impact |
|
|
|
- **Hardware Type:** Single GPU (NVIDIA Tesla T4) |
|
- **Hours used:** 3 |
|
- **Cloud Provider:** Google Cloud |
|
- **Compute Region:** US |
|
- **Carbon Emitted:** Approximately 0.36 kg CO2eq |
|
|
|
## Technical Specifications |
|
|
|
### Model Architecture and Objective |
|
|
|
The model is based on the `BioBERT` architecture fine-tuned for the SQuAD task, with a focus on healthcare question-answering. |
|
|
|
### Compute Infrastructure |
|
|
|
- **Hardware:** NVIDIA Tesla T4 GPU |
|
- **Software:** PyTorch, Transformers Library |
|
|
|
## Citation |
|
|
|
If you use this model, please cite it as: |
|
|
|
``` |
|
@misc{Ndoma2024onlinedoctor, |
|
author = {Ayamba Victor Ndoma}, |
|
title = {Online Doctor Model for Disease Prediction}, |
|
year = {2024}, |
|
howpublished = {\url{https://huggingface.co/your-username/online-doctor-model}}, |
|
} |
|
``` |
|
|
|
## Model Card Authors |
|
|
|
- Ayamba Victor Ndoma |
|
|
|
## Model Card Contact |
|
|
|
For questions or feedback, please contact `[email protected]`. |
|
|
|
--- |
