Research map / Label-efficient and robust learning / Recognition and long-tail
Word translation without parallel data
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 CSLS rather than nearest-neighbor retrieval (ablation) | Procrustes word translation · English-Esperanto | P@1 | Helped |
| Use CSLS rather than nearest-neighbor retrieval (ablation) | Adversarial word translation · Spanish-English | P@1: +8.4% | Helped |
| Refine the adversarial mapping with Procrustes using a synthetic dictionary (ablation) | Adversarial cross-lingual mapping · Word translation retrieval | P@1 | Helped |
| Perform more than one refinement iteration (ablation) | Adversarial cross-lingual mapping · Word translation retrieval | P@1: below 1% | No clear effect |
| Feed rare words to the discriminator (ablation) | Adversarial mapping · Cross-lingual word translation | P@1 | Hurt |
| Sample discriminator inputs by word frequency rather than uniformly (ablation) | Adversarial mapping · Cross-lingual word translation | P@1 | No clear effect |
| Vary CSLS neighborhood size among K=5, 10 and 50 (ablation) | CSLS retrieval · Word translation retrieval | P@1 | No clear effect |
| Use fastText rather than CBOW embeddings trained on the same corpus (ablation) | Cross-lingual word translation · Word translation retrieval | P@1: two percent more accuracy | Helped |
| Use Wikipedia rather than WaCky training corpora (ablation) | Cross-lingual word translation · Word translation retrieval | P@1: 18.8% gain | Helped |
| Use a moment-matching baseline instead of adversarial distribution alignment (ablation) | Unsupervised cross-lingual mapping · Word translation retrieval | P@1 | Hurt |
| Use mutual nearest-neighbor pairs when generating the synthetic dictionary (ablation) | Procrustes refinement · Word translation retrieval | P@1 | Helped |
| Apply CSLS rather than nearest neighbors to dictionary retrieval (ablation) | English-Esperanto word-by-word translation · English-Esperanto | BLEU: 6.1 -> 11.1 | Helped |
Closest papers by meaning
- Unsupervised Machine Translation Using Monolingual Corpora Only (ICLR 2018 · similarity 0.92)
- Bilingual Lexicon Induction from Non-Parallel Data with Minimal Supervision (AAAI 2017 · similarity 0.91)
- Unsupervised Hyper-alignment for Multilingual Word Embeddings (ICLR 2019 · similarity 0.90)
- Normalization of Language Embeddings for Cross-Lingual Alignment (ICLR 2022 · similarity 0.90)
- Enhancing Bilingual Lexicon Induction via Bi-directional Translation Pair Retrieving (AAAI 2024 · similarity 0.86)
- MMTEB: Massive Multilingual Text Embedding Benchmark (ICLR 2025 · 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.