|
|
--- |
|
|
language: en |
|
|
license: mit |
|
|
tags: |
|
|
- conversational-ai |
|
|
- question-answering |
|
|
- nlp |
|
|
- transformers |
|
|
- context-aware |
|
|
datasets: |
|
|
- squad |
|
|
metrics: |
|
|
- exact_match |
|
|
- f1_score |
|
|
model-index: |
|
|
- name: Harpertoken ConvAI |
|
|
results: |
|
|
- task: |
|
|
type: question-answering |
|
|
dataset: |
|
|
name: squad |
|
|
type: question-answering |
|
|
metrics: |
|
|
- type: exact_match |
|
|
value: 0.75 |
|
|
- type: f1_score |
|
|
value: 0.85 |
|
|
--- |
|
|
|
|
|
# Harpertoken ConvAI |
|
|
|
|
|
## Model Overview |
|
|
|
|
|
A context-aware conversational AI model based on DistilBERT for natural language understanding and generation. |
|
|
|
|
|
### Key Features |
|
|
- **Advanced Response Generation** |
|
|
- Multi-strategy response mechanisms |
|
|
- Context-aware conversation tracking |
|
|
- Intelligent fallback responses |
|
|
|
|
|
- **Flexible Architecture** |
|
|
- Built on DistilBERT base model |
|
|
- Supports TensorFlow and PyTorch |
|
|
- Lightweight and efficient |
|
|
|
|
|
- **Robust Processing** |
|
|
- 512-token context window |
|
|
- Dynamic model loading |
|
|
- Error handling and recovery |
|
|
|
|
|
## Quick Start |
|
|
|
|
|
### Installation |
|
|
```bash |
|
|
pip install transformers torch |
|
|
``` |
|
|
|
|
|
### Usage Example |
|
|
```python |
|
|
from transformers import AutoModelForQuestionAnswering, AutoTokenizer |
|
|
|
|
|
# Load model and tokenizer |
|
|
model = AutoModelForQuestionAnswering.from_pretrained('harpertoken/harpertokenConvAI') |
|
|
tokenizer = AutoTokenizer.from_pretrained('harpertoken/harpertokenConvAI') |
|
|
``` |
|
|
|
|
|
## Model Capabilities |
|
|
- Semantic understanding of context and questions |
|
|
- Ability to extract precise answers |
|
|
- Multiple response generation strategies |
|
|
- Fallback mechanisms for complex queries |
|
|
|
|
|
## Performance |
|
|
- Trained on Stanford Question Answering Dataset (SQuAD) |
|
|
- Exact Match: 75% |
|
|
- F1 Score: 85% |
|
|
|
|
|
## Limitations |
|
|
- Primarily trained on English text |
|
|
- Requires domain-specific fine-tuning |
|
|
- Performance varies by use case |
|
|
|
|
|
## Technical Details |
|
|
- **Base Model:** DistilBERT |
|
|
- **Variant:** Distilled for question-answering |
|
|
- **Maximum Sequence Length:** 512 tokens |
|
|
- **Supported Backends:** TensorFlow, PyTorch |
|
|
|
|
|
## Citation |
|
|
```bibtex |
|
|
@misc{harpertoken-convai, |
|
|
title={Harpertoken ConvAI}, |
|
|
author={Niladri Das}, |
|
|
year={2025}, |
|
|
url={https://huggingface.co/harpertoken/harpertokenConvAI} |
|
|
} |
