Expert Work Automation in Healthcare: the Case of a Retrieval-Based Medical Chatbot
Aunimo, Lili; Kauttonen, Janne; Alamäki, Ari (2022)
Aunimo, Lili
Kauttonen, Janne
Alamäki, Ari
Haaga-Helia ammattikorkeakoulu
2022
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi-fe2023030329620
https://urn.fi/URN:NBN:fi-fe2023030329620
Tiivistelmä
This paper presents how medical expert work may be partially automated and made more interesting as input for routine conversation is handled by a software bot. However, the responsibility for treating the patient in the right way always stays with the medical doctor. The researchers describe how a retrieval-based one-to-one medical chatbot can be implemented for the Finnish language using neural networks based deep learning. The chatbot is evaluated using separate test data. The results show that a Top1 precision score of about 80% can be reached when the context size is 20. The Top1 precision score tells how often the chatbot ranks the correct answer as 1st among 10 candidates, where 1 answer is correct and 9 are wrong. The qualitative evaluation with healthcare services management shows that the healthcare industry shows great interest in medical chatbot systems. This is because they would both enhance the user experience and interestingness of work perceived by the medical doctors and at the same time make their work more productive. However, there is demand for practical systems integration and user interface development as well as for the development of task specific medical dialogue systems before medical chatbots become mainstream.