Contrastive Representation Learning for Self-Supervised Taxonomy Completion
Authors: Yuhang Niu, Hongyuan Xu, Ciyi Liu, Yanlong Wen, Xiaojie Yuan
IJCAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experimental results on three datasets verify the effectiveness of our approach. |
| Researcher Affiliation | Academia | Yuhang Niu , Hongyuan Xu , Ciyi Liu , Yanlong Wen and Xiaojie Yuan College of Computer Science, Nankai University, Tianjin, China {niuyuhang, xuhongyuan, liuciyi}@dbis.nankai.edu.cn, {wenyl, yuanxj }@nankai.edu.cn |
| Pseudocode | No | The paper describes the proposed method in text and figures but does not include any pseudocode or algorithm blocks. |
| Open Source Code | Yes | Our code is available at https://github.com/nyh-a/Co STC. |
| Open Datasets | Yes | We conduct experiments on three taxonomy completion datasets: Sem Eval-Food, Me SH and Word Net-Verb. These are hierarchy taxonomies from different domains. The statistics of these datasets are shown in Table 1. Sem Eval-Food: This dataset contains a food domain taxonomy, derived from Sem Eval-2015 Task 17 [Bordea et al., 2015]. Me SH: This dataset contains a widely used clinical domain taxonomy. It is a subgraph of the Medical Subject Headings(Me SH) [Lipscomb, 2000]. Word Net-Verb: This dataset contains a verb taxonomy, derived from Sem Eval-2016 Task 14 [Jurgens and Pilehvar, 2016]. |
| Dataset Splits | Yes | For Sem Eval-Food and Me SH, we split them by 8:1:1 into train, validation and test sets. For Word Net-Verb, we randomly sample 1000 nodes respectively for validation and testing. |
| Hardware Specification | Yes | All experiments are accelerated using NVIDIA RTX 4090 GPUs. |
| Software Dependencies | No | Co STC is implemented with Py Torch. For fair comparison, we use BERT as the backbone model. The paper mentions PyTorch and BERT but does not provide specific version numbers for the libraries used (e.g., PyTorch 1.x.x). |
| Experiment Setup | Yes | For intra-view contrastive pertaining, the batch size is set to 64, the training epoch to 5, and the learning rate to 5e-5. For inter-view contrastive matching, the batch size is set to 4, the negative size to 40, the training epoch to 50, and the learning rate to 5e-5. ... Initial hyperparameters r and m for our sampling method are 0.25 and 0.4, respectively. |