llm-semantic-router/dissat-detector

ModernBERT SAT/DISSAT Binary Classifier

A fine-tuned ModernBERT model that classifies user follow-up messages as satisfied (SAT) or dissatisfied (DISSAT) based on their response to a system's answer.

Model Description

This model detects user dissatisfaction in conversational AI interactions. It's designed to identify when users:

• Reject an answer ("No, that's wrong")
• Request alternatives ("Show me other options")
• Express frustration ("That doesn't help")
• Resend/rephrase their query
• Add constraints implying the answer was insufficient

Citation

If you use this model, please cite:

@misc{modernbert-sat-dissat-2024,
  title={ModernBERT SAT/DISSAT Classifier for Conversational AI},
  author={Rootfs},
  year={2024},
  publisher={Hugging Face}
}

License

Apache 2.0