We present a novel application of hypothesis ranking (HR) for the task of domain detection in a multi-domain, multi-turn dialog system. Alternate, domain dependent, semantic frames from a spoken language understanding (SLU) analysis are ranked using a gradient boosted decision trees (GBDT) ranker to determine the most likely domain. The ranker, trained using Lambda Rank, makes use of a range of signals derived from the SLU and previous turn context to improve domain detection. On a multi-turn corpus we show that this approach offers accuracy improvements of 3.2% absolute (25.6% relative) compared to relying solely on upfront non-contextual SLU domain models and 2.9% (24.5% relative) improvement even with contextual SLU domain models. We also show that HR can be trained to be robust to changes in the SLU.
Cite as: Robichaud, J.-P., Crook, P.A., Xu, P., Khan, O.Z., Sarikaya, R. (2014) Hypotheses ranking for robust domain classification and tracking in dialogue systems. Proc. Interspeech 2014, 145-149, doi: 10.21437/Interspeech.2014-41
@inproceedings{robichaud14_interspeech, author={Jean-Philippe Robichaud and Paul A. Crook and Puyang Xu and Omar Zia Khan and Ruhi Sarikaya}, title={{Hypotheses ranking for robust domain classification and tracking in dialogue systems}}, year=2014, booktitle={Proc. Interspeech 2014}, pages={145--149}, doi={10.21437/Interspeech.2014-41} }