Crowdsourcing (CS) is now the default means for acquiring high-quality training data for AI. However, the process itself can be challenging even for technically savvy researchers. This project sets to provide the de facto method for the ethical crowdsourcing of domain-specific and clinically relevant text corpora for insights on arbitrary medical conditions, thus addressing the dearth of high-quality training data for NLP and consequently rapidly unlocking advances in digital health. We conduct two major case studies on low back pain and mental wellbeing, with diverse samples drawn across the US and Europe. Ultimately, we help decrease the global burden of disease and improve people's quality of life in a scalable and accessible way. Additionally, we also release highly user-friendly software tools for non-technical scientists to replicate the data collection, thus ensuring the broader impact and external validity of our work.

2022

2027