GLoHBCD: A Naturalistic German Dataset for Language of Health Behaviour Change on Online Support Forums

Proceedings of the 13th Conference on Language Resources and Evaluation (LREC 2022), pages 2226–2235 Marseille, 20-25 June 2022 © European Language Resources Association (ELRA), licensed under CC-BY-NC-4.0 2226 GLoHBCD: A Naturalistic German Dataset for Language of Health Behaviour Change on Online Support Forums Selina Meyer, David Elsweiler Chair for Information Science, University of Regensburg Universitätsstr. 31, 93051 Regensburg {selina.meyer, david.elsweiler}@ur.de Abstract Health behaviour change is a difficult and prolonged process that requires sustained motivation and determination. Conversa- tional agents have shown promise in supporting the change process in the past. One therapy approach that facilitates change and has been used as a framework for conversational agents is motivational interviewing. However, existing implementations of this therapy approach lack the deep understanding of user utterances that is essential to the spirit of motivational interviewing. To address this lack of understanding, we introduce the GLoHBCD, a German dataset of naturalistic language around health behaviour change. Data was sourced from a popular German weight loss forum and annotated using theoretically grounded motivational interviewing categories. We describe the process of dataset construction and present evaluation results. Initial experiments suggest a potential for broad applicability of the data and the resulting classifiers across different behaviour change domains. We make code to replicate the dataset and experiments available on Github. Keywords: Conversational Agents, Motivational Interviewing, Behaviour Change, Language Resources 1. Introduction Over the second half of the past century, illnesses resulting from poor health decisions, such as smok- ing, alcoholism and behaviours leading to obesity have emerged as a leading cause of death (Keeney, 2008; Johnson, N. B., Hayes, L. D., Brown, K., Hoo, E. C., Ethier, K. A., & Centers for Disease Control and Pre- vention (CDC), 2014). Despite their benefits and of- tentimes necessity, health behaviour changes are diffi- cult to put into practice and sustain (Kelly and Barker, 2016). A therapy approach developed to facilitate behaviour change is motivational interviewing (MI) (Miller and Rollnick, 2002). The automated delivery of MI by conversational agents (CA) has shown promise in the past (da Silva et al., 2018; Friederichs et al., 2015; Olafsson et al., 2019) and potentially offers mul- tiple benefits such as constant availability and higher cost effectiveness when compared to a qualified coun- sellor. This would result in a lower entry barrier for first contact (Lisetti et al., 2015). Past studies have, however, largely disregarded the motivational state and utterances of the user, as the CAs were usually based on a rigid action framework and often limited user in- teraction to multiple choice entries with only few free text inputs. This is problematic since an important part of MI is tailoring the conversation to the client’s needs (Miller and Rollnick, 2002; Hall et al., 2012). MI is a very client-centered therapy approach and revolves around making a person aware of their own personal reasons for behaviour change by using open questions, reflec- tions and affirmations. One of the central goals of MI is to elicit change talk, meaning language in favour of a behaviour …

Content truncated to 500 words