Loading...
Paper Number
1134
Paper Type
Short
Description
Question answer (QA) assistants are vital tools to address users’ information needs in healthcare. Knowledge graphs (KGs) and language models (LMs) have shown promise in building QA systems, but face challenges in their integration, and performance. Motivated thus, we take the case of a specific disease, skin eczema, to design a QA system combining KG and LM approaches. We present design iterations for systematically developing the KG, then fine-tuning a LM, and finally carrying out joint reasoning over both. We observe that while KGs are effective for fact finding, fine-tuned LMs perform better at answering complex queries. Initial results suggest that combining KG and LM approaches can improve the performance of the system. Our study contributes by laying out the design steps and developing a QA system that addresses various gaps in the related literature. Our future plan is to refine these techniques towards building a full-fledged healthcare QA assistant.
Recommended Citation
Sukhwal, Prakash Chandra; Kankanhalli, Atreyi; and Rajan, Vaibhav, "Designing a Healthcare QA Assistant: A Knowledge Based Approach" (2022). ICIS 2022 Proceedings. 1.
https://aisel.aisnet.org/icis2022/is_health/is_health/1
Designing a Healthcare QA Assistant: A Knowledge Based Approach
Question answer (QA) assistants are vital tools to address users’ information needs in healthcare. Knowledge graphs (KGs) and language models (LMs) have shown promise in building QA systems, but face challenges in their integration, and performance. Motivated thus, we take the case of a specific disease, skin eczema, to design a QA system combining KG and LM approaches. We present design iterations for systematically developing the KG, then fine-tuning a LM, and finally carrying out joint reasoning over both. We observe that while KGs are effective for fact finding, fine-tuned LMs perform better at answering complex queries. Initial results suggest that combining KG and LM approaches can improve the performance of the system. Our study contributes by laying out the design steps and developing a QA system that addresses various gaps in the related literature. Our future plan is to refine these techniques towards building a full-fledged healthcare QA assistant.
When commenting on articles, please be friendly, welcoming, respectful and abide by the AIS eLibrary Discussion Thread Code of Conduct posted here.
Comments
16-HealthCare