An effective knowledge graph entity alignment model based on multiple information.

Area of Science:

• Artificial Intelligence
• Data Science
• Knowledge Representation

Background:

• Knowledge graphs often lack sufficient structural information, hindering effective entity alignment.
• Knowledge graph heterogeneity presents challenges for accurate entity matching.
• Existing methods underutilize semantic and string information for entity alignment.

Purpose of the Study:

• To propose a novel entity alignment model, EAMI, that leverages multiple information sources.
• To enhance entity alignment by integrating structural, semantic, and string information.
• To address limitations of sparse and heterogeneous knowledge graphs in entity alignment tasks.

Main Methods:

• Utilizing multi-layer graph convolutional networks for learning structural representations.
• Incorporating attribute semantic representations with structural representations for improved entity vectors.
• Employing entity name string similarity, calculated without training, to further enhance alignment.

Main Results:

• The proposed EAMI model demonstrates effectiveness on cross-lingual and cross-resource datasets.
• Integration of multiple information sources significantly improves entity alignment accuracy.
• The model successfully mitigates issues arising from sparse and heterogeneous knowledge graphs.

Conclusions:

• The EAMI model offers a robust solution for entity alignment by effectively combining diverse information types.
• Leveraging semantic and string information alongside structural data is crucial for advanced knowledge graph alignment.
• The findings highlight the potential of multi-information fusion for improving the quality and utility of knowledge graphs.