Research map / Probabilistic and scientific modeling / Causal discovery
Causally Denoise Word Embeddings Using Half-Sibling Regression
Zekun Yang, Tianlin Liu
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 |
|---|---|---|---|
| Apply HSR-RR postprocessing | Word2Vec embeddings · STS-2015 | Pearson correlation: 66.35 -> 73.62 | Improved |
| Apply HSR-RR postprocessing | GloVe embeddings · STS-2014 | Pearson correlation: 56.29 -> 69.23 | Improved |
| Apply HSR-RR postprocessing | Paragram embeddings · STS-2015 | Pearson correlation: 58.10 -> 65.04 | Improved |
| Apply HSR-RR postprocessing | Word2Vec embeddings · SICK | Pearson correlation: 72.25 -> 72.02 | Hurt |
| Apply HSR-RR postprocessing | Word2Vec embeddings · SimVerb-3500 | Spearman correlation: 0.3659 -> 0.3978 | Improved |
| Apply HSR-RR postprocessing | GloVe embeddings · SimVerb-3500 | Spearman correlation: 0.2842 -> 0.3980 | Improved |
| Apply HSR-RR postprocessing | Paragram embeddings · SimVerb-3500 | Spearman correlation: 0.5411 -> 0.5518 | Improved |
| Apply HSR-RR postprocessing | Word2Vec embeddings · AR | accuracy: 0.8375 -> 0.8377 | No clear effect |
| Apply HSR-RR postprocessing | GloVe embeddings · CR | accuracy: 0.7829 -> 0.7792 | Hurt |
| Apply HSR-RR postprocessing | Paragram embeddings · IMDB | accuracy: 0.7957 -> 0.7999 | Improved |
| Apply HSR-RR postprocessing | Word2Vec embeddings · STS-B | accuracy: 0.8071 -> 0.8056 | Hurt |
| Postprocess content words only and leave function words unchanged (ablation) | HSR-RR · word vector evaluation tasks | performance | Hurt |
Closest papers by meaning
- A Causal Inference Method for Reducing Gender Bias in Word Embedding Relations (AAAI 2020 · similarity 0.87)
- Distributed Representation of Words in Cause and Effect Spaces (AAAI 2019 · similarity 0.86)
- A General Framework for Implicit and Explicit Debiasing of Distributional Word Vector Spaces (AAAI 2020 · similarity 0.86)
- Enriching Word Embeddings with a Regressor Instead of Labeled Corpora (AAAI 2019 · similarity 0.85)
- Unsupervised Post-Processing of Word Vectors via Conceptor Negation (AAAI 2019 · similarity 0.85)
- A Cat Is A Cat (Not A Dog!): Unraveling Information Mix-ups in Text-to-Image Encoders through Causal Analysis and Embedding Optimization (NeurIPS 2024 · similarity 0.83)
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.
