Skip to content

Read-only version: For security reasons, the Research Information System (FIS) has been temporarily replaced by this read-only version. All public content and full texts remain available; login, editing, submission and statistics are not available. Questions: fis@uni-bamberg.de

Home › Publications › On the Controllability of Large Language Models for Dialogue Interaction

Publication

On the Controllability of Large Language Models for Dialogue Interaction

Wagner, Nicolas; Ultes, Stefan (2024): On the Controllability of Large Language Models for Dialogue Interaction, in: Tatsuya Kawahara, Vera Demberg, Stefan Ultes, u. a. (Hrsg.), Proceedings of the 25th Annual Meeting of the Special Interest Group on Discourse and Dialogue, Kyoto: Association for Computational Linguistics, S. 216–221, doi: 10.18653/v1/2024.sigdial-1.19.

Details

Faculty/Chair
Faculty/Professorship
Author
Wagner, Nicolas
Title of the compilation
Proceedings of the 25th Annual Meeting of the Special Interest Group on Discourse and Dialogue
Editors
Kawahara, Tatsuya
Demberg, Vera
Inoue, Koji
Mehri, Shikib
Howcroft, David
Komatani, Kazunori
Conference
25th Annual Meeting of the Special Interest Group on Discourse and Dialogue, September 18 - 20, 2024, Kyoto
Publisher Information
Association for Computational Linguistics : Kyoto
Year of publication
2024
Language
English
DOI
Abstract

This paper investigates the enhancement of Dialogue Systems by integrating the creative capabilities of Large Language Models. While traditional Dialogue Systems focus on understanding user input and selecting appropriate system actions, Language Models excel at generating natural language text based on prompts. Therefore, we propose to improve controllability and coherence of interactions by guiding a Language Model with control signals that enable explicit control over the system behaviour. To address this, we tested and evaluated our concept in 815 conversations with over 3600 dialogue exchanges on a dataset. Our experiment examined the quality of generated system responses using two strategies: An unguided strategy where task data was provided to the models, and a controlled strategy in which a simulated Dialogue Controller provided appropriate system actions. The results show that the average BLEU score and the classifcation of dialogue acts improved in the controlled Natural Language Generation.

Keywords
-
Peer Reviewed
yes
International Distribution
yes
Type
Conferenceobject
URI
Activation date
2025-09-16

Permalink: https://fis.uni-bamberg.de/handle/uniba/110386 · Data as JSON · Cite: BibTeX · RIS

Data as of: 2026-11-15

Restricted read-only access: login, editing and submission are temporarily disabled.