Multi-Task Learning for Supervised Pretraining of Goal-Oriented Dialogue Policies

Sarah McLeod, Ivana Kruijff-Korbayová, Bernd Kiefer

In: 20th Annual SIGdial Meeting on Discourse and Dialogue - Proceedings of the Conference. Annual SIGdial Meeting on Discourse and Dialogue (SIGdial-2019) 20th September 11-13 Stockholm Sweden Seiten 411-417 ISBN 978-1-950737-61-1 The Association of Computational Linguistics 9/2019.


This paper describes the use of Multi-Task Neural Networks (NNs) for system dialogue act selection. These models leverage the rep-resentations learned by the Natural Language Understanding (NLU) unit to enable robust initialization/bootstrapping of dialogue poli-cies from medium sized initial data sets. We evaluate the models on two goal-oriented di- alogue corpora in the travel booking domain. Results show the proposed models improve over models trained without knowledge of NLU tasks.

Weitere Links

ACL_short_paper_158.pdf (pdf, 304 KB )

Deutsches Forschungszentrum für Künstliche Intelligenz
German Research Center for Artificial Intelligence