Research map / Label-efficient and robust learning / Recognition and long-tail
Exploiting Cross-Lingual Subword Similarities in Low-Resource Document Classification
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 |
|---|---|---|---|
| add word translation (dictionary of 100 pairs) auxiliary task (ablation) | CACO · RCV2 avg | accuracy: 50.0 -> 55.7 | Helped |
| add mimick pretrained-embedding task (ablation) | CACO · RCV2 avg | accuracy: 50.0 -> 51.5 | No clear effect |
| add both dict and mimick over dict only (ablation) | CACO · RCV2 avg | accuracy: 55.7 -> 54.7 | No clear effect |
| add knowledge distillation from English reference model (ablation) | CACO · LORELEI am->ti | accuracy: 55.5 -> 57.0 (SRC) | Helped |
| add pretrained CLWE features to CACO | CACO · RCV2 avg | accuracy: 50.0 -> 64.5 | Helped |
| use two source languages instead of one (ablation) | CACO · RCV2 Romance avg (DICT) | accuracy: 55.7 -> 59.6 (SRC 50.0 -> 54.6 on different pair set) | Helped |
| transfer between distant language families (North Germanic to Romance) vs related | CACO · RCV2 | accuracy: CLWE 49.5 vs CACO DICT 38.2 | Hurt |
| Romance to North Germanic transfer | CACO · RCV2 | accuracy: 48.6 vs 48.5 (CLWE) | No clear effect |
| CACO SRC vs CLWE on word translation | CACO · MUSE P@1 | accuracy: 37.7 -> 33.1 | Hurt |
Closest papers by meaning
- Cross-lingual Alignment vs Joint Training: A Comparative Study and A Simple Unified Framework (ICLR 2020 · similarity 0.87)
- Normalization of Language Embeddings for Cross-Lingual Alignment (ICLR 2022 · similarity 0.87)
- Enhanced Meta-Learning for Cross-Lingual Named Entity Recognition with Minimal Resources (AAAI 2020 · similarity 0.87)
- Word translation without parallel data (ICLR 2018 · similarity 0.86)
- MCL-NER: Cross-Lingual Named Entity Recognition via Multi-View Contrastive Learning (AAAI 2024 · similarity 0.85)
- Repurposing Language Models into Embedding Models: Finding the Compute-Optimal Recipe (NeurIPS 2024 · similarity 0.85)
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.