Medical consultation information aggregation analysis method

Abstract

The invention discloses a medical consultation information aggregation analysis method, and the method comprises the following steps: 1, employing the web crawler technology to obtain medical information in the form of 'question-answer-doctor' triple on an existing medical health consultation website; 2, converting the medical information in text format into entity-based structured data; 3, calculating the similarity between structured questions, merging the problems with high similarity to expand the answer set of the questions; 4, calculating the credibility of the answers through an iterative method, getting the problem-answers-credibility in the form of medical knowledge. The method achieves the calculation of the medical knowledge level of a doctor and the reliability of the answers in an unsupervised manner, saves a large amount of money and time, and obtains the precise medical knowledge.

Description

technical field [0001] The invention relates to the field of medical consultation, in particular to a method for aggregation and analysis of medical consultation information. Background technique [0002] With the development of society and economy, people's living standards are getting higher and higher, and people pay more attention to their health. The demand for authoritative health information is growing urgently. In recent years, with the rapid development of Internet, mobile Internet and other technologies, people are deeply involved in network knowledge production and online social interaction, and a large number of online question-answer consulting application systems have emerged. Among them, the medical crowdsourcing question-and-answer website is welcomed by the majority of users because it can conveniently provide timely first-diagnosis to patients. In foreign countries, such as medhelp.org in the United States, there are xywy.com and Baidu Thumb Doctor (muzhi....

AI Technical Summary

Problems solved by technology

First, the truth discovery method. In order to calculate the credibility of the answer and the knowledge level of the doctor, one possible way is to mark the quality of the answer and then learn a classification or regression model. However, this is a relatively professional field. , ordinary workers cannot effectively label samples, and workers with certain medical knowledge are required to label samples, which increases the cost of labeling. Realization; second, the purification of noise input, the medical questions and answers on the medical consultation website are all unstructured and noisy text data, which brings inconvenience to the reliability of the calculation and the doctor's knowledge level, in order to obtain more For good results, you need to better express the questions and answers; third, the long tail phenomenon, most questions only received two or three or even only one answer, which brings trouble to the screening of answers, because there are too few answers to screen It is meaningless, and these few answers are not even reasonable, so we need to find a way to expand the answer set; fourth, the correlation between the answers to the same question, the answer to the medical consultation question is not uniquely determined Yes, a question may have multiple reasonable and reliable answers, and these answers may have a relatively high similarity, which can also be an important basis for calculating the credibility of the answer

[0005] At present, most of the text mining of medical question-answering information focuses on the identification of medical named entities and the discovery of the relationship between entities, and eventually the potential medical knowledge in medical question-answer pairs can be obtained, but the quality of medical question-answering is not considered.

In reality, there are indeed some low-level doctors on medical question-and-answer websites, and even some doctors do not seriously answer patients' questions just to make money, so the quality of medical knowledge obtained is difficult to guarantee

Images