HBot: A Chatbot for Healthcare Applications in Traditional Chinese Medicine Based on Human Body 3D Visualization (2408.00481v1)

Abstract: The unique diagnosis and treatment techniques and remarkable clinical efficacy of traditional Chinese medicine (TCM) make it play an important role in the field of elderly care and healthcare, especially in the rehabilitation of some common chronic diseases of the elderly. Therefore, building a TCM chatbot for healthcare application will help users obtain consultation services in a direct and natural way. However, concepts such as acupuncture points (acupoints) and meridians involved in TCM always appear in the consultation, which cannot be displayed intuitively. To this end, we develop a \textbf{h}ealthcare chat\textbf{bot} (HBot) based on a human body model in 3D and knowledge graph, which provides conversational services such as knowledge Q&A, prescription recommendation, moxibustion therapy recommendation, and acupoint search. When specific acupoints are involved in the conversations between user and HBot, the 3D body will jump to the corresponding acupoints and highlight them. Moreover, Hbot can also be used in training scenarios to accelerate the teaching process of TCM by intuitively displaying acupuncture points and knowledge cards. The demonstration video is available at https://www.youtube.com/watch?v=UhQhutSKkTU . Our code and dataset are publicly available at Gitee: https://gitee.com/plabrolin/interactive-3d-acup.git

• The paper introduces HBot, a TCM healthcare chatbot that leverages an interactive 3D human body model for precise visualization of acupuncture points.
• It employs a multi-phase intent detection and entity extraction approach, integrating a TCM knowledge graph for accurate, context-aware responses.
• Extensive testing shows robust performance with F1 scores of 79.47 for entity extraction and 81.56 for relation extraction, demonstrating effective system validation.

Overview of “HBot: A Chatbot for Healthcare Applications in Traditional Chinese Medicine Based on Human Body 3D Visualization”

The paper "HBot: A Chatbot for Healthcare Applications in Traditional Chinese Medicine Based on Human Body 3D Visualization" introduces a sophisticated chatbot system (HBot) designed for healthcare applications in the domain of Traditional Chinese Medicine (TCM). The key innovation of HBot lies in its use of a 3D human body model, which facilitates the intuitive location and visualization of acupuncture points and meridians during interaction. This paper delineates the system's architecture, functionality, and evaluation, offering valuable insights into the development and capabilities of TCM-oriented dialogue systems.

Core Contributions

The authors enumerate several principal contributions:

1. Interactive 3D Human Body Visualization:
   • HBot implements a fully interactive 3D model of the human body, which visually represents meridians and acupuncture points according to the State Standard of the Location of Acupoints. This model can rotate, zoom, and highlight specific acupuncture points to enhance user comprehension.
2. Knowledge Graph Construction:
   • The system incorporates a TCM-specific Knowledge Graph (TCM-KG), which includes a structured ontology and instances derived from authoritative TCM literature. This knowledge graph facilitates accurate knowledge retrieval and enhances the chatbot's response capability.
3. Sophisticated Intent Detection and Entity Extraction:
   • HBot utilizes a three-phase intent detection strategy composed of rule-based matching, model-based understanding using Conco-ERNIE, and similarity-based query searching with SBERT. For entity extraction, a BERT-CRF model is employed to identify TCM-related entities, augmented by a custom relation extraction model for constructing TCM-KG.
4. Robust System Architecture:
   • The system is structurally divided into six key modules, including User Intent Detection, Entity-Relation Extraction, LLM Handler, and Wrapper, among others. This modular approach ensures the chatbot's operational robustness and facilitates multi-turn dialogue management.

System Architecture and Functionality

HBot's architecture features comprehensive interactive components and sophisticated backend processing capabilities:

• Interactive 3D Body Model:
  • Implemented using Three.js and Blender, the 3D model renders a realistic representation of the human body, enabling users to interact with and visualize specific acupuncture points effectively.
• User Intent Detection:
  • The dual-phase strategy enhances accuracy and adaptability, allowing the system to manage predefined intents and handle unexpected user queries with a high degree of precision.
• Entity-Relation Extraction:
  • Utilizing advanced natural language processing techniques, the system accurately identifies and categorizes entities enhancing the interaction quality and the system's overall understanding of user queries.
• LLM Handler and Wrapper:
  • These modules ensure that the chatbot can generate coherent and contextually relevant responses by drawing upon a combination of text-based documents and graph-based knowledge.

Evaluation and Human Testing

The authors conducted extensive $\alpha$ and $\beta$ testing to evaluate the system's performance rigorously:

• Alpha Testing:
  • The developmental team executed 100 test cases to identify and rectify 21 bugs, focusing on model operation, intent understanding, and system robustness.
• Beta Testing:
  • Conducted by non-developers, this assessment gauged user satisfaction across 120 distinct instructions. Despite variations in the testing approach, the system demonstrated notable improvements post-alpha testing and offered satisfactory performance in the majority of cases.

Numerical Results and Implications

The paper provides robust numerical outcomes from the entity and relation extraction tasks, showcasing the efficacy of their models:

• Entity Extraction:
  • The BERT-CRF model achieved an F1 score of 79.47, demonstrating substantial accuracy in identifying TCM-related entities.
• Relation Extraction:
  • The combined ERNIE+GRU model reported an F1 score of 81.56, indicating effective relation extraction capabilities integral to the knowledge graph construction.

Implications and Future Directions

The practical implications of HBot are considerable, particularly for TCM practitioners and patients requiring remote consultations and educational training. The system's integration of 3D visualization and interactive dialogue signifies a meaningful advancement in TCM healthcare technologies.

Theoretically, the methodologies for intent detection, entity extraction, and knowledge graph construction presented in this work could be adapted to other domains requiring specialized terminology and complex relational data.

Future developments may explore integrating more advanced LLMs, enhancing the knowledge graph's completeness, and refining user interaction interfaces to improve accessibility and user experience.

In summary, the paper presents a well-structured and technically robust framework for a TCM-specific healthcare chatbot, offering significant contributions to the field of intelligent dialogue systems and healthcare informatics. The integration of 3D visualization and extensive NLP techniques sets a precedent for future research and applications in this multidisciplinary domain.