Generative dialogue system for auxiliary diagnosis and treatment of digestive medicine
面向消化内科辅助诊疗的生成式对话系统

摘要

随着经济的快速提升，人们的生活和工作压力逐渐增加，受社会环境和自身状况的影响，许多人们的生活规律都发生了较大的变化，尤其是饮食方面，多数人的饮食呈现出不规律和不健康的现象，这种情况下，消化系统疾病的发生率呈现出逐年上升的趋势。许多人发现健康出现异常时，第一反应不是入院就医，而是通过网络寻求答案。就消化内科疾病而言，其具有多样性的特点，仅仅依靠网络搜索，难以获取疾病的相关信息[1]。因此，为了提高患者获取信息的准确性，应该对传统的搜索方式进行改进。在信息检索系统中，对话系统相对较为高级，用户通过该系统将自己的需求输入，对话系统的另一方根据文字描述，将有效信息及时反馈，通过这种方法能够提高用户获取信息的准确性。本研究对消化内科中的生成式对话系统进行探索，利用向量机的学习功能和分类功能，从医学网站中将有关于消化内科的专业问诊语句进行获取，同时通过统计方法将消化内科中的各种知识制作成专业词典，并对分词效果进行改善和提升，在此基础上，将注意力机制加入多层结构和门控循环单元的结合方式中，并对训练方法进行加强，通过以上步骤，促进消化内科生成式对话系统的完善。通过研究表明，生成式对话系统与传统的对比方式，比较分词准确率，传统方式相对较低，由此可见生成式对话系统能够提高回答的准确性。

Abstract

With the rapid economic improvement, people's life and work pressure are gradually increasing. Affected by the social environment and their own conditions, many people's life laws have undergone major changes, especially in terms of diet, most people's diet appear irregular and unhealthy phenomena, in this case, the incidence of digestive system diseases shows a trend of increasing year by year. When many people find abnormal health, the first reaction is not to be admitted to the hospital, but to seek answers through the Internet. As far as gastroenterological diseases are concerned, it has the characteristics of diversity, and it is difficult to obtain disease-related information only by relying on network search [1]. Therefore, in order to improve the accuracy of information obtained by patients, the traditional search method should be improved. In the information retrieval system, the dialogue system is relatively advanced. The user enters his needs through the system. The other party of the dialogue system feeds back the effective information in time according to the text description. This method can improve the accuracy of the user's access to information. This study explores the generative dialogue system in Gastroenterology, using the learning function and classification function of the vector machine to obtain professional questioning statements about Gastroenterology from the medical website, and at the same time statistical methods Various knowledge is made into a professional dictionary, and the effect of word segmentation is improved. On this basis, the attention mechanism is added to the combination of multi-layer structure and gated circulation unit, and the training method is strengthened. Promote the improvement of the generative dialogue system of digestive medicine through the above steps. The research shows that the comparison between the generative dialogue system and the traditional way compares the accuracy of word segmentation. The traditional method is relatively low, which shows that the generative dialogue system can improve the accuracy of the answer.