Transition-Matrix Regularization for Next Dialogue Act Prediction in Counselling Conversations
For researchers working on dialogue act prediction in data-sparse domains like counselling, this provides a lightweight method to incorporate discourse-flow priors that consistently boosts performance, especially for weaker models.
The paper introduces a KL regularization term that aligns predicted dialogue act distributions with corpus-derived transition patterns, improving macro-F1 by 9-42% relative on a 60-class German counselling taxonomy and showing cross-dataset transfer to English.
This paper studies how empirical dialogue-flow statistics can be incorporated into Next Dialogue Act Prediction (NDAP). A KL regularization term is proposed that aligns predicted act distributions with corpus-derived transition patterns. Evaluated on a 60-class German counselling taxonomy using 5-fold cross-validation, this improves macro-F1 by 9--42% relative depending on encoder and substantially improves dialogue-flow alignment. Cross-dataset validation on HOPE suggests that improvements transfer across languages and counselling domains. In systematic ablations across pretrained encoders and architectures, the findings indicate that transition regularization provides consistent gains and disproportionately benefits weaker baseline models. The results suggest that lightweight discourse-flow priors complement pretrained encoders, especially in fine-grained, data-sparse dialogue tasks.