bert-medical-entity-extraction

Medical Entity Extraction with BERT

📌 Overview

This repository hosts the quantized version of the bert-base-cased model for Medical Entity Extraction using the 'tner/bc5cdr' dataset. The model is specifically designed to recognize entities related to Disease,Symptoms,Drug. The model has been optimized for efficient deployment while maintaining high accuracy, making it suitable for resource-constrained environments.

🏗 Model Details

• Model Architecture: BERT Base Cased
• Task: Medical Entity Extraction
• Dataset: Hugging Face's tner/bc5cdr
• Quantization: Float16
• Fine-tuning Framework: Hugging Face Transformers

---

🚀 Usage

Installation

pip install transformers torch

Loading the Model

from transformers import BertTokenizerFast, BertForTokenClassification
import torch

device = "cuda" if torch.cuda.is_available() else "cpu"

model_name = "AventIQ-AI/bert-medical-entity-extraction"
model = BertForTokenClassification.from_pretrained(model_name).to(device)
tokenizer = BertTokenizerFast.from_pretrained(model_name)

Named Entity Recognition Inference

from transformers import pipeline

ner_pipeline = pipeline("ner", model=model_name, tokenizer=tokenizer)
test_sentence = "An overdose of Ibuprofen can lead to severe gastric issues."
ner_results = ner_pipeline(test_sentence)
label_map = {
    "LABEL_0": "O",  # Outside (not an entity)
    "LABEL_1": "Drug",
    "LABEL_2": "Disease",
    "LABEL_3": "Symptom",
    "LABEL_4": "Treatment"
}

def merge_tokens(ner_results):
    merged_entities = []
    current_word = ""
    current_label = ""
    current_score = 0
    count = 0

    for entity in ner_results:
        word = entity["word"]
        label = entity["entity"]  # Model's output (e.g., LABEL_1, LABEL_2)
        score = entity["score"]

        # Merge subwords
        if word.startswith("##"):
            current_word += word[2:]  # Remove '##' and append
            current_score += score
            count += 1
        else:
            if current_word:  # Store the previous merged word
                mapped_label = label_map.get(current_label, "Unknown")
                merged_entities.append((current_word, mapped_label, current_score / count))
            current_word = word
            current_label = label
            current_score = score
            count = 1

    # Add the last word
    if current_word:
        mapped_label = label_map.get(current_label, "Unknown")
        merged_entities.append((current_word, mapped_label, current_score / count))

    return merged_entities

print("\n🩺 Medical NER Predictions:")
for word, label, score in merge_tokens(ner_results):
    if label != "O":  # Skip non-entities
        print(f"🔹 Entity: {word} | Category: {label} | Score: {score:.4f}")

🔹 Labeling Scheme (BIO Format)

• B-XYZ (Beginning): Indicates the beginning of an entity of type XYZ (e.g., B-PER for the beginning of a person’s name).
• I-XYZ (Inside): Represents subsequent tokens inside an entity (e.g., I-PER for the second part of a person’s name).
• O (Outside): Denotes tokens that are not part of any named entity.

---

📊 Evaluation Results for Quantized Model

🔹 Overall Performance

• Accuracy: 93.27% ✅
• Precision: 92.31%
• Recall: 93.27%
• F1 Score: 92.31%

---

🔹 Performance by Entity Type

---

⏳ Inference Speed Metrics

• Total Evaluation Time: 15.89 sec
• Samples Processed per Second: 217.26
• Steps per Second: 27.18
• Epochs Completed: 3

---

Fine-Tuning Details

Dataset

The Hugging Face's tner/bc5cdr dataset was used, containing texts and their ner tags.

📊 Training Details

• Number of epochs: 3
• Batch size: 8
• Evaluation strategy: epoch
• Learning Rate: 2e-5

⚡ Quantization

Post-training quantization was applied using PyTorch's built-in quantization framework to reduce the model size and improve inference efficiency.

---

📂 Repository Structure

.
├── model/               # Contains the quantized model files
├── tokenizer_config/    # Tokenizer configuration and vocabulary files
├── model.safetensors/   # Quantized Model
├── README.md            # Model documentation

---

⚠️ Limitations

• The model may not generalize well to domains outside the fine-tuning dataset.
• Quantization may result in minor accuracy degradation compared to full-precision models.

---

🤝 Contributing

Contributions are welcome! Feel free to open an issue or submit a pull request if you have suggestions or improvements.