BLISS: Behaviour based Language Interactive Speaking Systems
Talking to the computer to feel better!
Language scientists Helmer Strik and Iris Hendrickx from Radboud University are working on an innovative Dutch-language chatbot. After the research, their chatbot should be able to talk to people about their personal well-being and happiness.
The idea is to sift through large amounts of clients' personal data (texts, interviews, dialogues) to extract information about their health and well-being and to find out what makes them happy. To make this possible, technology is being developed that can extract relevant information from Dutch texts and audio recordings. This information can then be used to design computer programmes that can communicate with people in low-threshold, spoken Dutch to support them in their daily actions and managing their health and well-being.
Spoken data
Here, the information from the large amounts of personal data can be used to personalise the system for individual clients so that the system takes into account personal needs and background.
Until now, research with Big Data was mainly based on written data, while this project also uses spoken interviews, as these are now collected much more than written questionnaires. The advantage is also that spoken data contains much more information than texts (such as intonation, emphasis, hesitations, etc.).
We spoke to Helmer and Iris in early September 2022.

Fine cooperation
The project involves collaboration with colleagues at the University of Twente with their specialism in human-computer interaction. Two companies are also involved; Readspeaker for the development of the computer voice and the company Games for Health. Games for Health deploys games & apps to influence the behaviour of people and organisations in an impactful way. They do this by developing serious games & gamified apps with clients from different fields, with a penchant for health and happiness. Helmer says: "By adding applications with game elements, we make training more fun and interesting and encourage participants to drop out less quickly." Helmer emphasises that the collaboration in the project gives a lot of energy "we have biweekly online meetings where everyone is present. Everyone is active and together we look for creative solutions for new experiments within the project. The knowledge we gain within BLISS can then be used again in other projects such as ST CART and is also extremely useful for the participating companies." A large number of Bachelor and Master students are also working on the project. "It is a subject that appeals, that is scientifically interesting as well as socially very relevant. People find it a very enjoyable topic."
Experiments adapted
To arrive at such a complex computer programme requires a lot of data: answers from people telling how they are doing. Here, unfortunately, Covid threw a spanner in the works. "We too suffered from corona," Helmer explains. "We were originally going to conduct a large survey with people in (home) care. In the care sector, however, it was 'all hands on deck' and so taking part in our study was not a top priority. We therefore did not manage to find enough participants and had to adjust our plan. We switched to a number of small-scale experiments, which certainly also yielded a lot of useful information. "Iris adds "originally, software had been developed for on the laptop, the idea was to go to care homes with the laptop and start the conversation with participants on the spot. After we prepared everything for this experiment, we had to switch to an online interface so that participants could also engage on other devices. This switch brought new challenges and took a lot of extra time. But in the end we succeeded." Other side steps included an experiment around 'automatically summarising dialogues' and a study on whether people like an empathetic voice better than a neutral one.
'Death' sub-experiment
Other choices were also made around the themes as the research progressed. "We deliberately did not pick up the theme of loneliness as a sub-experiment due to the lack of the right expertise in the team," Iris explains. "Another sensitive topic - death - was chosen in collaboration with one of the colleagues at Radboud University. We looked at how films featuring death (eudamonic entertainment) could help people talk and think about death and whether a chatbot could have these conversations."
Understanding well-being
The aim of BLISS is to gain insight into your own well-being; we don't give any handles in doing so, Iris says. "We are language scientists; giving health advice is outside our expertise. We focus on positive health and instead let people talk about the things that make them happy. Positive Health is a broad view of health, elaborated in six pillars. With this broad approach, you contribute to people's ability to cope with life's physical, emotional and social challenges. With our BLISS speech programme, we focus on the two pillars 'daily functioning' and 'participating/social participation'."
Upcoming experiment with caregivers
A major challenge in the entire project is finding and engaging enough participants and then being able to have multiple conversations with them. Iris: "Next experiment we will focus on the care providers. We think it's nice to also be able to talk about yourself instead of just someone else's care. We really want to have a series of several conversations then. This is really necessary to take steps towards personalisation."
Results
This research will yield several results. First, important Dutch-language technology to analyse large amounts of texts and recordings about and from clients and extract relevant information about their health and well-being. This technology can also be used for other purposes, such as improving a speech recognition system. Furthermore, this project also provides knowledge about an interactive system that can speak in Dutch, understand their answers and itself has important information about the clients that they can use to improve their wellbeing, thus keeping them in control. This system can also potentially be used for other applications in care. Important outcomes are, of course, also the knowledge and insights that will be gained about what works best for whom.
- The project will be completed in September 2023, for more information see https://bliss.ruhosting.nl/.
- Get involved? Engage with the online Blissbot: https://forms.office.com/r/tqg...
Gerelateerde projects
Bekijk hieronder de Projecten gerelateerd aan het thema BLISS: Behaviour based Language Interactive Speaking Systems
BLISS: Behaviour based language interactive speaking systems
The availability of large amounts of personal data (Big Data) and the significant recent advancements in language and speech technology offer new opportunities for self-management and empowerment of clients in healthcare. This research proposal adopts a broad definition of health and well-being as happiness and aims to develop an interactive, Dutch-speaking system that helps people enhance their well-being and happiness.