Availability evaluation method for medical text structuring algorithm

Abstract

The invention provides a novel availability evaluation method for a medical text structuring algorithm. For any medical text structuralization algorithm, the availability is used for evaluating the medical text structuralization algorithm, and the numerical value proves that the algorithm is suitable for practical application scenes to a large extent and comprises three dimensions: generalizationability on corpora of different hospitals and different departments; the anti-interference capability of the algorithm is improved; and the tolerance of algorithm errors is improved. Based on the definition of availability, a specific available evaluation method is constructed in combination with the similarity relationship between medical texts, the type of algorithm errors, the type of data interference and the like. Indexes and methods investigate whether medical text-oriented machine learning algorithm theoretical research is suitable for practical application scenarios or not, and can also be used in other fields.

Description

technical field [0001] The invention relates to an evaluation technology of an algorithm, in particular to an evaluation method for the availability of a medical text structured algorithm. Background technique [0002] Since the existing text recognition evaluation methods are all based on the basic assumption of machine learning, that is, the independent and identical distribution characteristics of the learning corpus and the real corpus. In algorithm execution and evaluation, the same data is split into training set, validation set and test set. However, datasets used in real medical applications may not fully meet the above assumptions. [0003] In addition, in the deep learning algorithm, there is a situation where the non-related context changes, resulting in a complete change in the algorithm result. The robustness of the algorithm is relatively poor, but the traditional evaluation method cannot solve the above problems. Therefore, a large number of algorithms that ...

AI Technical Summary

Problems solved by technology

However, the data sets used in the actual medical field may not fully meet the above assumptions

[0003] In addition, in the deep learning algorithm, there is a situation where the non-related context changes, resulting in a complete change in the algorithm result. The robustness of the algorithm is relatively poor, but the traditional evaluation method cannot solve the above problems

Therefore, a large number of algorithms that perform well in the laboratory perform poorly in real scenarios

Images