Entity alignment method and system applied to knowledge graph

Abstract

The invention discloses an entity alignment method and system applied to a knowledge graph, and relates to the technical field of computers. The method comprises the steps: carrying out the entity alignment of an author name and a stored author name, and enabling matched second author attribute data to serve as a candidate entity; performing entity alignment scoring on the first author attribute data and the second author attribute data according to a first entity alignment model to obtain a first entity alignment score, wherein the first entity alignment score is smaller than a first entity alignment threshold and the knowledge graph data is a college thesis, and performing entity alignment scoring on the first author attribute data and the second author attribute data according to a second entity alignment model to obtain a second entity alignment score; taking a summation result of the first entity alignment score and the second entity alignment score as a total entity alignment score, and adding knowledge graph data into an author entity library when the total entity alignment score is smaller than a second entity alignment threshold; otherwise performing knowledge fusion. According to the method, the entity alignment accuracy and recall rate are effectively improved.

Description

technical field [0001] The present invention relates to the field of computer technology, in particular to an entity alignment method and system applied to knowledge graphs. Background technique [0002] In today's society, everyone pays more and more attention to scientific research, and a large number of academic papers emerge. Scientific researchers use these scientific research papers to publish their latest research results and research progress. On the one hand, they can be recognized by everyone, and on the other hand, they can Know and cooperate with researchers in other related fields, so the thesis has an important impact on scientific researchers, and the dissertation has the same effect on students. For paper management platforms, such as HowNet, Wanfang, etc., they obtain a large number of papers from colleges, journals and conferences to provide services for researchers, making it convenient for researchers to consult and download papers in fields of interest, ...

AI Technical Summary

Problems solved by technology

[0004] The current entity alignment combines methods such as rules, machine learning, and deep learning. Most of these methods tend to be machine learning and deep learning, but these two methods require a large number of training samples, which will increase significantly with the increase of attributes. The amount of data required to label data, a large number of labeled samples also limits machine learning and deep learning

The regular method generally has a high recall rate, but the accuracy rate is not high, so manual participation is required to improve the accuracy of the results

Images