Adapting Neural Link Predictors for Data-Efficient Complex Query Answering
Authors: Erik Arakelyan, Pasquale Minervini, Daniel Daza, Michael Cochez, Isabelle Augenstein
NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In our experiments, CQDA produces significantly more accurate results than current state-of-the-art methods, improving from 34.4 to 35.1 Mean Reciprocal Rank values averaged across all datasets and query types while using 30% of the available training query types. |
| Researcher Affiliation | Collaboration | 1University of Copenhagen 2University of Edinburgh 3Vrije Universiteit Amsterdam 4University of Amsterdam 5Discovery Lab, Elsevier, The Netherlands |
| Pseudocode | No | The paper describes the method and steps in natural language but does not contain a formally labeled pseudocode or algorithm block. |
| Open Source Code | Yes | Source code and datasets are available at https://github.com/Edinburgh NLP/adaptive-cqd. |
| Open Datasets | Yes | Datasets To evaluate the complex query answering capabilities of our method, we use a benchmark comprising of 3 KGs: FB15K [Bordes et al., 2013], FB15K-237 [Toutanova and Chen, 2015] and NELL995 [Xiong et al., 2017]. |
| Dataset Splits | Yes | Valid 1p 59,078 20,094 16,910 Others 8,000 5,000 4,000 |
| Hardware Specification | No | The paper mentions 'GPU donations' in the acknowledgements but does not provide specific hardware details (e.g., specific GPU models, CPU types, or memory) used for running the experiments. |
| Software Dependencies | No | The paper mentions Compl Ex-N3 as the link prediction model and Adagrad as the optimizer, but it does not specify any software dependencies with version numbers (e.g., Python version, PyTorch version). |
| Experiment Setup | Yes | We train for 50, 000 steps using Adagrad as an optimiser and 0.1 as the learning rate. The beam-size hyper-parameter k was selected in k {512, 1024, . . . , 8192}, and the loss was selected across 1-vs-all [Lacroix et al., 2018] and binary cross-entropy with one negative sample. |