Enhancing Cross-Lingual Entity Alignment in Knowledge Graphs through Structure Similarity Rearrangement.

Area of Science:

• Artificial Intelligence
• Data Science
• Graph Theory

Background:

• Cross-lingual entity alignment is vital for knowledge fusion.
• Current models use embeddings and nearest neighbors but neglect local structure.
• Poorly represented nodes may not benefit from contextual information.

Purpose of the Study:

• To propose a novel alignment model, SSR, that incorporates local structural features.
• To enhance the accuracy of cross-lingual entity alignment in knowledge graphs.
• To improve upon existing alignment methods by considering node context.

Main Methods:

• Leveraging graph node embedding algorithms for candidate entity selection.
• Rearranging candidate entities based on local structural similarity.
• Evaluating the SSR model on the DBP15k dataset.

Main Results:

• The SSR model demonstrates improved performance over existing cross-lingual entity alignment approaches.
• The method effectively utilizes local structural information for better matching accuracy.
• The approach is compatible with existing alignment models.

Conclusions:

• The proposed SSR model offers a more accurate and efficient solution for cross-lingual entity alignment.
• Incorporating local structural similarity is key to improving alignment performance.
• SSR provides a valuable advancement in knowledge fusion techniques.