Research map / Label-efficient and robust learning / Recognition and long-tail
Learning Context-Specific Word/Character Embeddings
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 multi-prototype senses (CSV-M over CSV-S) (ablation) | CSV · POS tagging | accuracy (PWA): 96.26 -> 97.15 | Helped |
| add multi-prototype senses (CSV-M over CSV-S) (ablation) | CSV · Chunking | F1: 91.49 -> 92.12 | Helped |
| add multi-prototype senses (CSV-M over CSV-S) (ablation) | CSV · Chinese word segmentation | F1: 93.26 -> 94.03 | Helped |
| add multi-prototype senses (CSV-M over CSV-S) (ablation) | CSV · Chinese NER (MSRA) | F1: 85.31 -> 85.59 | Helped |
| use CSV-S pretrained embeddings instead of SKIP | downstream tagger · Chunking | F1: 89.77 -> 91.49 | Helped |
| use convolutional context representation (CSV-S) instead of averaged-context embeddings | word embeddings · SimLex-999 | Spearman's rho | Helped |
| use CSV-S instead of SKIP on relatedness-oriented data | word embeddings · WordSim-353 | Spearman's rho: 0.713 vs CSV-S about 0.03 lower | No clear effect |
| initialize downstream tagger with pretrained embeddings vs random | Collobert network · Chunking | F1: 89.10 -> 91.49 | Helped |
| select a single sense by context rather than weighting similarity | Neelakantan et al (2014) · SCWS | Spearman's rho: 0.691 -> 0.598 | Hurt |
Closest papers by meaning
- Learning Semantic Representations for Novel Words: Leveraging Both Form and Context (AAAI 2019 · similarity 0.92)
- Improving Word Embeddings with Convolutional Feature Learning and Subword Information (AAAI 2017 · similarity 0.90)
- All Word Embeddings from One Embedding (NeurIPS 2020 · similarity 0.90)
- A Simple but Tough-to-Beat Baseline for Sentence Embeddings (ICLR 2017 · similarity 0.90)
- Geometric Relationship between Word and Context Representations (AAAI 2018 · similarity 0.89)
- Contextual Document Embeddings (ICLR 2025 · similarity 0.89)
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.