[arXiv]score: 0.20

Transition-Matrix Regularization Boosts Counselling Dialogue Act Prediction

November 3, 2026

Applying KL regularization to align predicted dialogue act distributions with corpus-derived transition patterns improves macro-F1 by 9–42%. The method demonstrates cross-language and cross-domain transferability when tested on the HOPE dataset.

HOW THIS AFFECTS YOU

●

builderIntegrating lightweight discourse-flow priors can significantly improve performance in data-sparse, fine-grained dialogue tasks.

●

researcherThe results suggest that statistical transition matrices are a highly effective regularizer for encoder-based dialogue models.

read original ↗arxiv.org

DAILY DIGEST

you don't check 9 sources — we do. one email every morning, read in 2 min. free. unsubscribe anytime. privacy

← back to feed