Research map / Label-efficient and robust learning / Self-supervised and contrastive
Fake it Till You Make it: Self-Supervised Semantic Shifts for Monolingual Word Embedding Tasks
Maurício Gruppi, Pin-Yu Chen, Sibel Adali
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 S4-D with top-frequency landmark alignment instead of cosine-distance classification | British-American English semantic change detection · British-American English | F1: $0.15/0.00$ -> $0.82 \pm 0.01$ | Improved |
| Use S4-D instead of Noisy-Pairs | British-American English semantic change detection · British-American English | F1: $0.06$ -> $0.82 \pm 0.01$ | Improved |
| Use S4-A landmark selection instead of frequency-based or global alignment | SemEval-2020 lexical semantic change detection · SemEval-2020 English and German | accuracy: English: 0.68 -> 0.70; German: 0.79 -> 0.81 | Improved |
| Use S4-A rather than the strongest frequency-based alignment | SemEval-2020 lexical semantic change detection · SemEval-2020 Latin and Swedish | accuracy: Latin: 0.70 -> 0.68; Swedish: 0.81 -> 0.77 | Hurt |
| Replace Word2Vec with FastText embeddings (ablation) | S4-D · British-American English | F1: 0.82 $\pm 0.01$ -> 0.67 $\pm 0.02$ | Hurt |
| Replace Word2Vec with GloVe embeddings (ablation) | S4-A · SemEval-2020 English | accuracy: 0.70 -> 0.68 | No clear effect |
| Replace Word2Vec with FastText embeddings (ablation) | S4-A · SemEval-2020 English | accuracy: 0.70 -> 0.65 | Hurt |
Closest papers by meaning
- The Global Anchor Method for Quantifying Linguistic Shifts and Domain Adaptation (NeurIPS 2018 · similarity 0.89)
- VAST: The Valence-Assessing Semantics Test for Contextualizing Language Models (AAAI 2022 · similarity 0.87)
- Minimally-Constrained Multilingual Embeddings via Artificial Code-Switching (AAAI 2016 · similarity 0.86)
- SensEmBERT: Context-Enhanced Sense Embeddings for Multilingual Word Sense Disambiguation (AAAI 2020 · similarity 0.86)
- Unseen Word Representation by Aligning Heterogeneous Lexical Semantic Spaces (AAAI 2019 · similarity 0.86)
- Dynamic Word Embeddings (ICML 2017 · similarity 0.86)
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.
