Interactive Healthcare Robot Using Attention-Based Question-Answer Retrieval and Medical Entity Extraction Models

Area of Science:

• Health Informatics
• Artificial Intelligence in Healthcare
• Natural Language Processing

Background:

• Healthcare professionals face increasing demands, limiting time for patient inquiries.
• Accurate health information is vital for improving patient knowledge, attitudes, and behaviors.
• Existing systems may not adequately address the growing need for accessible healthcare information.

Purpose of the Study:

• To develop an interactive healthcare question-answering (QA) system.
• To leverage attention-based models for accurate and real-time responses to patient health questions.
• To enhance user experience through interactive features like question recommendations and web search integration.

Main Methods:

• Utilized attention-based transformer models to encode semantic meanings and extract medical entities from user queries.
• Developed a fusion module to integrate encoded query features with a pre-collected healthcare knowledge set.
• Implemented a recommendation module and an online web search module to improve system interactivity and provide out-of-scope answers.

Main Results:

• The proposed system demonstrated the ability to accurately retrieve answers from healthcare-specific frequently asked question (FAQ) pairs.
• Experimental results validated the effectiveness of the attention-based models and fusion module for question-answer retrieval.
• The system successfully provided accurate responses in real-time.

Conclusions:

• The developed interactive healthcare QA system effectively addresses the need for accessible and accurate health information.
• The integration of attention-based models and interactive modules enhances the system's utility for patients and healthcare providers.
• This application holds significant potential to benefit human beings by improving health literacy and patient engagement.