AventIQ-AI/Bert-base-uncased-intent-classification

BERT-Base-Uncased Fine-Tuned Model for Intent Classification on CLINC150 Dataset

This repository hosts a fine-tuned BERT model for multi-class intent classification using the CLINC150 (plus) dataset. The model is trained to classify user queries into 150 in-scope intents and handle out-of-scope (OOS) queries.

Model Details

• Model Architecture: BERT Base Uncased
• Task: Multi-class Intent Classification
• Dataset: CLINC150 (plus variant)
• Quantization: Float16
• Fine-tuning Framework: Hugging Face Transformers

---

Installation

pip install transformers datasets scikit-learn evaluate

---

Loading the Model

from transformers import AutoModelForSequenceClassification, AutoTokenizer
import torch

# Load tokenizer and model
model_path = "bert-base-uncased"
tokenizer = AutoTokenizer.from_pretrained(model_path)
model = AutoModelForSequenceClassification.from_pretrained(model_path)
# Define test sentences
test_sentences = [
    "Can you tell me the weather in New York?",
    "I want to transfer money to my friend",
    "Play some relaxing jazz music",
]


# Tokenize and predict
def predict_intent(sentences, model, tokenizer, id2label_fn, device="cpu"):
    if isinstance(sentences, str):
        sentences = [sentences]

    model.eval()
    model.to(device)

    inputs = tokenizer(sentences, padding=True, truncation=True, return_tensors="pt").to(device)

    with torch.no_grad():
        outputs = model(**inputs)
        logits = outputs.logits
        predictions = torch.argmax(logits, dim=-1)

    return [id2label_fn(label.item()) for label in predictions]

---

Performance Metrics

• Accuracy: 0.947097
• Precision: 0.949821
• Recall: 0.947097
• F1 Score: 0.945876

---

Fine-Tuning Details

Dataset

The CLINC150 (plus) dataset contains 151 intent classes (150 in-scope + 1 out-of-scope) for intent classification in English utterances. It includes 15k training, 3k validation, and 4.5k test examples with diverse user queries.

Training

• Epochs: 5
• Batch size: 16
• Learning rate: 2e-5
• Evaluation strategy: epoch

---

Quantization

Post-training quantization was applied using PyTorch’s half() precision (FP16) to reduce model size and inference time.

---

Repository Structure

.
├── quantized-model/               # Contains the quantized model files
│   ├── config.json
│   ├── model.safetensors
│   ├── tokenizer_config.json
│   ├── vocab.txt
│   └── special_tokens_map.json
├── README.md                      # Model documentation

---

Limitations

• The model is trained specifically for multi classification on CLINIC150 Dataset.
• FP16 quantization may result in slight numerical instability in edge cases.

---

Contributing

Feel free to open issues or submit pull requests to improve the model or documentation.