aiqbench analysis
This paper adds a KL-divergence regularizer that pushes a dialogue model's predicted act distributions toward transition statistics measured in the training corpus. Reported gains are 9-42% relative macro-F1 on a 60-class German counselling taxonomy, with a cross-dataset check on HOPE. Readers should note that the evaluation is narrow (one domain family, mostly small-data settings) and that the largest gains appear for weaker baselines, so the benefit over strong pretrained encoders is more modest.
Transition-Matrix Regularization for Next Dialogue Act Prediction in Counselling Conversations
arXiv:2604.18539v2 Announce Type: replace-cross Abstract: 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.
Abstract: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.
Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
arXiv:2604.18539 [cs.CL]
(or arXiv:2604.18539v2 [cs.CL] for this version)
https://doi.org/10.48550/arXiv.2604.18539 Focus to learn more arXiv-issued DOI via DataCite
https://doi.org/10.18653/v1/2026.findings-acl.1271 Focus to learn more DOI(s) linking to related resources
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 213 words