Name disambiguation method and system based on LightGBM classification and representation learning

Abstract

The invention provides a LightGBM classification and representation learning-based name disambiguation method and a LightGBM classification and representation learning-based name disambiguation system for scientific literature data and aiming at author homonymy phenomena in literatures. According to the supervised learning part, meta-information features of papers in a training set and associated information features among the papers are extracted by utilizing feature engineering, a positive example and negative example sample pair data set is constructed through sampling and serves as input of a LightGBM dichotomy model, and model output serves as the probability that the two papers belong to the same author. The representation learning part refers to a word2vec text semantic representation method and a meta-path-based relation network representation method to capture semantic information of papers and relation characteristics between the papers. And finally, based on the output of the supervision model and the representation learning model, cluster division is performed on the to-be-disambiguated paper set by using a hierarchical clustering algorithm to realize homonymy disambiguation. According to the method, high expandability and stability can be achieved on the premise that the accuracy rate and the recall rate are not lost, parallel calculation can be completely achieved, and the execution efficiency is improved.

Description

technical field [0001] The invention belongs to the field of information technology, and in particular relates to a name disambiguation method and system based on LightGBM classification and representation learning. Background technique [0002] Name disambiguation is considered a focal task in the field of scientific literature data. It is mainly used in literature data management, analysis, scholar retrieval, and construction of scholar social network. With the rapid increase in the number of scientific literature in recent years, the number of scholars has also increased, and the rate of duplicate names of scholars is getting higher and higher, which makes the task of disambiguation of the same name face a huge challenge. Previously, many solutions for the name disambiguation task have been proposed at home and abroad. Due to the multi-source data and the complexity of the application scenarios, there is still room for optimization in the disambiguation method. [0003]...

AI Technical Summary

Problems solved by technology

When these methods capture the similarity between papers, the features involved are limited to semantics, co-authorship or co-occurrence of authors, etc. Few methods fully consider other meta-information of papers and use them as feature information to calculate the similarity between papers. perfect space

Images