Talk2UZH («talk to us»)

The Talk2UZH Lab aims to create a collaborative network to explore the application of speech recognition in a secure setting. The basis for this is a UZH-based, multilingual speech recognition software, which is integrated into a survey platform and will be made more easily accessible for UZH researchers. As part of an interdisciplinary pilot project, the usefulness of speech recognition for observational studies, as well as ethical and data protection challenges, will be examined.

«Siri», «Alexa», «OK Google» ... Many devices and operating systems already offer sophisticated speech recognition programs, but their use in scientific research faces various hurdles. On the one hand, the majority of these programs are not well adapted to subject-specific terminology or local languages (e.g., dialects). On the other hand, privacy and data protection concerns remain salient because the data is stored and processed in cloud servers abroad.

Likely owing to these hurdles, speech recognition is still relatively rarely used in observational studies. Yet the use of speech recognition would open up new avenues for person-centered research, for example in surveys of chronically ill people. Spoken language contains additional layers of information. Compared to structured surveys, spoken language provides a more detailed picture of life circumstances or attitudes and views of the respondents, thus accommodating the wish of many chronically ill people to convey their personal stories to researchers. At the same time, speech tempo, tonality, or pauses carry information that can be indicative of mental health problems, for example.

The Talk2UZH Lab aims to facilitate the integration of speech recognition in observational research. As a first step, a pilot software for speech recognition will be optimized to make it more easily accessible for research and teaching via programming interfaces and an easy-to-use survey, hosted on secure UZH servers. The optimization steps will be centered around the principles of data protection, privacy, and scalability. In the next step, the software will be practice-tested in an interdisciplinary, community-led project. The project results and experiences will provide insights into new application areas for speech recognition, as well as clues about remaining ethical or privacy challenges.

The Talk2UZH Lab is a collaboration between the Digital & Mobile Health Group , the Population Research Center , the Linguistic Research Infrastructure Network, and the DIPEx project at the Institute of Biomedical Ethics. The software, as well as results and experiences from the application project, will be shared with the research community through workshops and consulting.

Here you can find a short interview about the project.

Project duration: 01.09.2023 – 28.02.2025

Contact: Prof. Dr. Viktor von Wyl

Project Team

Prof. Dr. Viktor von Wyl (Projektleitung)
Digital and Mobile Health Group

Prof. Dr. Volker Dellwo
Linguistic Research Infrastructure

Prof. Dr. Andrea Glässel
Institute for Biomedical Ethics & ZHAW

Prof. Dr. Milo Puhan
Population Research Center