Revisit and Outstrip Entity Alignment: A Perspective of Generative Models
Authors: Lingbing Guo, Zhuo Chen, Jiaoyan Chen, Yin Fang, Wen Zhang, Huajun Chen
ICLR 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We conduct experiments to validate the performance of GEEA, where it achieves state-of-the-art performance in entity alignment and generates high-quality new entities in entity synthesis. The source code and datasets are available at github.com/zjukg/GEEA. |
| Researcher Affiliation | Collaboration | 1College of Computer Science and Technology, Zhejiang University 2Zhejiang University Ant Group Joint Laboratory of Knowledge Graph 3Alibaba-Zhejiang University Joint Reseach Institute of Frontier Technologies 4Department of Computer Science, The University of Manchester 5School of Software Technology, Zhejiang University |
| Pseudocode | Yes | Algorithm 1 Generative Embedding-based Entity Alignment |
| Open Source Code | Yes | The source code and datasets are available at github.com/zjukg/GEEA. |
| Open Datasets | Yes | We used the multi-modal EEA benchmarks (DBP15K (Sun et al., 2017), FB15K-DB15K and FB15K-YAGO15K (Chen et al., 2020)) as datasets, excluding surface information (i.e., the textual label information) to prevent data leakage (Sun et al., 2020b; Chen et al., 2022b). |
| Dataset Splits | Yes | An EEA model M uses a small number of aligned entity pairs S (a.k.a., seed alignment set) as training data to infer the remaining alignment pairs T in the testing set. ... until the performance on the valid dataset does not improve. |
| Hardware Specification | Yes | We implement GEEA with Py Torch and run the main experiments on a RTX 4090. |
| Software Dependencies | No | We implement GEEA with Py Torch and run the main experiments on a RTX 4090. The paper mentions 'Py Torch' but does not specify a version number or other software dependencies with version numbers. |
| Experiment Setup | Yes | The overall hyper-parameter settings in the main experiments are presented in Table 6. |