Knowledge learning method and system based on binary relations

Abstract

The invention discloses a knowledge learning method and system based on binary relations. According to the method, phrases and / or sentences in a training database and a material database are parsed through a natural-language processing method to obtain word trees; and label information of word nodes and label information of the binary word relations in word trees of training data are acquired, andthe word trees and the corresponding label information are used as templates to be stored to a template library. Therefore, the template library is established through the small amount of training data on the basis of the binary relations, learning skills are obtained from the template library through machine learning, and are applied to to-be-learned material data, learned knowledge is stored ifapplication is successful, otherwise, new templates are obtained and are added to the template library, the template library can be expanded at any time, thus a large number of repeated labels on thesame or similar learning materials are not needed, manual workloads are greatly reduced, and reusability is improved. The knowledge learning method and system based on the binary relations of the invention can be widely applied to the field of artificial intelligence.

Description

technical field [0001] The invention relates to the field of artificial intelligence, in particular to a binary relation-based knowledge learning method and system. Background technique [0002] One field of artificial intelligence is to let the machine directly learn the material learned by the human being, that is, the material expressed in natural language, and use the learned knowledge to guide its work. One way of learning is to annotate a large amount of materials, and let the machine learn the format of these materials from the annotations, so that the machine can learn knowledge from other materials in the same or similar format. [0003] The limitation of the above approach is that it needs to label a large amount of materials, so generally it can only be targeted at a specific field. For example, in a paper, in order to realize machine-aided medical diagnosis, it is necessary to label thousands or even tens of thousands of cases. In order to understand the cases i...

AI Technical Summary

Problems solved by technology

[0003] The limitation of the above approach is that it needs to label a large amount of materials, so generally it can only be targeted at a specific field. For example, in a paper, in order to realize machine-aided medical diagnosis, it is necessary to label thousands or even tens of thousands of cases. In order to understand the cases in a similar format and learn from them, if another type of material is changed, such as B-ultrasound results, or a new format is changed, such as cases from another hospital, a large amount of material needs to be re-documented Labeling, high work repetition and low efficiency

Images