Towards Building a Spoken Dialogue System for Argument Exploration

Proceedings of the 13th Conference on Language Resources and Evaluation (LREC 2022), pages 1234–1241 Marseille, 20-25 June 2022 © European Language Resources Association (ELRA), licensed under CC-BY-NC-4.0 1234 Towards Building a Spoken Dialogue System for Argument Exploration Annalena Aicher1, Nadine Gerstenlauer1, Isabel Feustel1, Wolfgang Minker1, Stefan Ultes2 1Ulm University, Albert-Einstein-Allee 43, 89075 Ulm 2Mercedes Benz AG, Stuttgart, Germany annalena.aicher@uni-ulm,de Abstract Speech interfaces for argumentative dialogue systems (ADS) are rather scarce. The complex task they pursue hinders the application of common natural language understanding (NLU) approaches in this domain. To address this issue we include an adaption of a recently introduced NLU framework tailored to argumentative tasks into a complete ADS. We evaluate the likeability and motivation of users to interact with the new system in a user study. Therefore, we compare it to a solid baseline utilizing a drop-down menu. The results indicate that the integration of a flexible NLU framework enables a far more natural and satisfying interaction with human users in real-time. Even though the drop-down menu convinces regarding its robustness, the willingness to use the new system is significantly higher. Hence, the featured NLU framework provides a sound basis to build an intuitive interface which can be extended to adapt its behavior to the individual user. Keywords: Natural Language Understanding (NLU), Argumentative Dialogue Systems (ADS), Conversational Systems, Speech Interaction/Recognition, User Usability/Satisfaction, HCI, Preference Modelling 1. Introduction Building dialogue systems that can converse with hu- mans via natural language is a challenging yet intrigu- ing problem of artificial intelligence. Most of the pop- ular and accessible virtual agents (VA) are adequately trained to handle simple conversations, e.g. inquiry for hotels, restaurants etc. (Saha et al., 2020). However, such VAs are inept in managing complex dialogues es- pecially with regard to demanding conversations, like discussing a controversial topic and providing logically consistent arguments. For example, if the user requests for a supporting argument on a certain topic, they might more likely engage in a discussion if they are able to ask “Could you please be more specific on this argu- ment. I do not see why...” instead of simple, short commands like “Why?” or “More information!”. Fur- thermore, a natural language system response like “An argument in favor of the previously discussed statement is, that... which is underpinned...” may be more appeal- ing than a written system response. Such complex tasks demand for a flexible natural language understanding (NLU), an intuitive dialogue structure and the integration of commonsense knowl- edge. These components are combined in the speech- driven argumentative dialogue system (ADS) BEA (’Building Engaging Argumentation’ (Aicher et al., 2021)) we introduce in this paper. To the best of our knowledge this is the first ADS which tries to cooper- atively engage the user to explore arguments in natural language. Therefore, we investigate to what extent a spoken interaction increases the willingness to explore arguments with the aid of this system in a user study. In order to examine the perception of the system with re- gard to dimensions …

Content truncated to 500 words