Research map / Label-efficient and robust learning / Recognition and long-tail
Kernelized Hashcode Representations for Relation Extraction
Sahil Garg, Aram Galstyan, Greg Ver Steeg, Irina Rish, Guillermo Cecchi, Shuyang Gao
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use KLSH-RF instead of the strongest reported baseline | Biomedical relation extraction · AIMed → BioInfer | F1: 0.57 -> 0.57 | No clear effect |
| Use KLSH-RF instead of RNN or KLSH-kNN | Biomedical relation extraction · BioInfer → AIMed | F1: 0.51 -> 0.54 | Improved |
| Use KLSH-RF instead of CNN | Biomedical relation extraction · PubMed45 | F1: 0.51 ± 0.28 -> 0.57 ± 0.25 | Improved |
| Use KLSH-RF instead of Bi-LSTM | Biomedical relation extraction · PubMed45-ERN | F1: 0.37 ± 0.15 -> 0.45 ± 0.22 | Improved |
| Use KLSH-RF instead of CNN or LSTM-CNN | Biomedical relation extraction · BioNLP | F1: 0.60 -> 0.63 | Improved |
| Optimize the reference set instead of selecting it randomly (ablation) | KLSH-RF · PubMed45 | recall: 7-13 pts | Improved |
| Optimize the reference set instead of selecting it randomly (ablation) | KLSH-RF · PubMed45 | precision: 2-3 pts | Mixed |
| Optimize nonstationary kernel parameters (ablation) | KLSH-RF with PK · PubMed45 | recall: 7 pts | Improved |
| Optimize nonstationary kernel parameters (ablation) | KLSH-RF with PK · PubMed45 | precision: 1 pt | Hurt |
| Use fewer hashcode bits (H=15 or 30) (ablation) | KLSH-RF · PubMed45 | F1 | Hurt |
| Increase the number of decision trees beyond a minimal threshold (ablation) | KLSH-RF · PubMed45 | accuracy | No clear effect |
| Use Kulis hashing with higher alpha (ablation) | KLSH-RF · PubMed45 | accuracy | Hurt |
Closest papers by meaning
- Binary Embedding with Additive Homogeneous Kernels (AAAI 2017 · similarity 0.84)
- Locality-Sensitive Hashing for f-Divergences: Mutual Information Loss and Beyond (NeurIPS 2019 · similarity 0.84)
- Prototypical Representation Learning for Relation Extraction (ICLR 2021 · similarity 0.83)
- Efficient Online Learning for Mapping Kernels on Linguistic Structures (AAAI 2019 · similarity 0.83)
- Local Density Estimation in High Dimensions (ICML 2018 · similarity 0.83)
- Prototypical Hash Encoding for On-the-Fly Fine-Grained Category Discovery (NeurIPS 2024 · similarity 0.81)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
