Ghosh Shameek, Bhatia Sammi, Bhatia Abhi
Medius Health, Sydney, Australia.
Stud Health Technol Inform. 2018;252:51-56.
Automated conversational agents built with medical applications in mind, have the potential to reduce healthcare readmissions and improve accessibility to medical knowledge. In this work, we demonstrate the development and evaluation of an automated chatbot for triage and conditions assessment, based on user inputs in natural language. The implemented bot engages patients in conversation about symptoms experienced and provides a personalized pre-synopsis based on their symptoms and profile. Our chatbot system was able to predict user conditions correctly based on two sets of patient test cases with an average precision of 0.82. Our implementation demonstrates that a medical chatbot can help with automatic triage and pre-assessment of patients with simple symptom analysis and a conversational approach without the use of cumbersome form-based data entry.
专为医疗应用构建的自动化对话代理，有潜力减少医疗再入院率并提高获取医学知识的便利性。在这项工作中，我们展示了一个基于自然语言用户输入的用于分诊和病情评估的自动化聊天机器人的开发与评估。所实现的聊天机器人让患者参与有关所经历症状的对话，并根据他们的症状和个人资料提供个性化的预摘要。我们的聊天机器人系统能够基于两组患者测试用例正确预测用户病情，平均精确率为0.82。我们的实现表明，一个医疗聊天机器人可以通过简单的症状分析和对话方式，在不使用繁琐的基于表单的数据输入的情况下，帮助对患者进行自动分诊和预评估。
