Augmenting Knowledge Graphs for Better Link Prediction
Authors: Jiang Wang, Filip Ilievski, Pedro Szekely, Ke-Thia Yao
IJCAI 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experiments on legacy benchmarks and a new large benchmark, DWD, show that augmenting the knowledge graph with quantities and years is beneficial for predicting both entities and numbers, as KGA outperforms the vanilla models and other relevant baselines. Our ablation studies confirm that both quantities and years contribute to KGA s performance, and that its performance depends on the discretization and binning settings. |
| Researcher Affiliation | Academia | Jiang Wang , Filip Ilievski , Pedro Szekely and Ke-Thia Yao USC Information Sciences Institute {jiangwan, ilievski, pszekely, kyao}@isi.edu |
| Pseudocode | No | The paper describes the KGA method and its steps verbally and with diagrams (Figure 1), but it does not include a formal pseudocode block or algorithm listing. |
| Open Source Code | Yes | We make the code, models, and the DWD benchmark publicly available to facilitate reproducibility and future research.1...Public release of our code, resulting models, and the DWD benchmark: https://github.com/Otamio/KGA/. |
| Open Datasets | Yes | We use benchmarks based on three KGs to evaluate LP performance...FB15K-237 [Toutanova and Chen, 2015] is a subset of Freebase [Bollacker et al., 2008]...YAGO15K [Liu et al., 2019] is a subset of YAGO [Suchanek et al., 2007]...We introduce DARPA Wikidata (DWD), a large subset of Wikidata [Vrandeˇci c and Kr otzsch, 2014]...Public release of our code, resulting models, and the DWD benchmark: https://github.com/Otamio/KGA/. |
| Dataset Splits | Yes | We split the DWD benchmark at a 98-1-1 ratio, for both entity and numeric link prediction...We use the MRR on the validation set to select the best model checkpoint. |
| Hardware Specification | No | The paper mentions '34 GB GPU memory, which is challenging for most GPU devices today.' but does not specify any particular GPU models (e.g., NVIDIA A100, V100), CPU types, or other hardware components used for running the experiments. |
| Software Dependencies | No | The paper mentions various embedding models and baselines (e.g., Trans E, Dist Mult, Compl Ex, Conv E, Rotat E, Tuck ER, KBLN, Literal E, NAP++, Mr AP) and cites 'Pytorch-biggraph' but does not specify version numbers for these software components or underlying libraries like Python, PyTorch, or CUDA. |
| Experiment Setup | Yes | We run KGA with b [2, 4, 8, 16, 32] and show the best result for each model...We use 32-bin KGA with QOC (quantile, overlapping, and chaining) discretization...We study different variants of discretization (Fixed and Quantile-based), bin levels (Single, Overlapping, and Hierarchy), and bin sizes (2, 4, 8, 16, and 32) on the FB15K-237 benchmark...For entity LP on FB15K-237 and YAGO15K, we preserve a checkpoint of the model every 5 epochs for Dist Mult, Compl Ex, every 10 epochs for Conv E, Tuck ER, and every 50 steps for Trans E, Rotat E. |