Papers with entity alignment method
DeepAlignment: Unsupervised Ontology Matching with Refined Word Vectors (N18-1)
Copied to clipboard
| Challenge: | Ontologies compartmentalize types and relations in a domain and require a process to establish alignments between entities to unify and extend existing knowledge. |
| Approach: | They propose a method which refines pre-trained word vectors to derivate ontological entity descriptions tailored to the ontology matching task. |
| Outcome: | The proposed method improves ontology matching performance over the current state-of-the-art. |
NALA: an Effective and Interpretable Entity Alignment Method (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Existing embedding-based EA methods encode entities as embeddables and learn to align embeddibles. |
| Approach: | They propose to capture three types of logical inference paths with Non-Axiomatic Logic to iteratively align entities and relations by integrating the conclusions of the inference path. |
| Outcome: | The proposed method outperforms state-of-the-art methods in terms of Hits@1 on all three datasets of DBP15K with both supervised and unsupervised settings. |
Entity Profile Generation and Reasoning with LLMs for Entity Alignment (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Entity alignment is a process of identifying and linking equivalent entities across knowledge graphs . only a small fraction of these entities are aligned . |
| Approach: | They propose a method that combines large language models with entity embeddings to align entities. |
| Outcome: | ProLEA is a method that combines large language models with entity embeddings to improve alignment accuracy, robustness, and explainability. |