Research map / Label-efficient and robust learning / Recognition and long-tail
Leveraging Lexical Substitutes for Unsupervised Word Sense Induction
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 |
|---|---|---|---|
| replace context-embedding similarity with predicted lexical-substitute similarity (ablation) | WSI with affinity propagation · SemEval-2010 WSI | SR (80/20): .693 -> .722 | Helped |
| combine context and substitute similarities (late fusion) (ablation) | AUTOLS · SemEval-2010 WSI | SR (80/20): .722 -> .750 | Helped |
| substitutes instead of context only (ablation) | CTX · SemEval-2010 WSI | UAVG: .210 -> .242 (V-measure .225 -> .268) | Helped |
| substitutes instead of context only, with WordNet prior (ablation) | CTX (+WN prior) · SemEval-2010 WSI | UAVG: .243 -> .279 | Helped |
| add WordNet prior on number of clusters (ablation) | AUTOLS+CTX · SemEval-2010 WSI | SR (80/20): .750 -> .693 | Hurt |
| add WordNet prior on number of clusters (ablation) | AUTOLS+CTX · SemEval-2010 WSI | UAVG: .250 -> .284 | Helped |
| use human-produced substitutes instead of predicted substitutes (ablation) | AUTOLS (+WN prior) · SemEval-2010 20-word sample | UAVG: .352 -> .410 | Helped |
| use human-produced substitutes instead of predicted ones (ablation) | AUTOLS (+WN prior) · SemEval-2010 20-word sample | SR (80/20): .221 -> .227 | No clear effect |
| increase ratio of human substitutes mixed with system substitutes (ablation) | AUTOLS+CTX (+WN prior) · SemEval-2010 20-word sample | unsupervised WSI scores | Helped |
Closest papers by meaning
- AutoSense Model for Word Sense Induction (AAAI 2019 · similarity 0.88)
- SensEmBERT: Context-Enhanced Sense Embeddings for Multilingual Word Sense Disambiguation (AAAI 2020 · similarity 0.87)
- Knowledge-based Word Sense Disambiguation using Topic Models (AAAI 2018 · similarity 0.87)
- Joint Word Representation Learning Using a Corpus and a Semantic Lexicon (AAAI 2016 · similarity 0.86)
- VAST: The Valence-Assessing Semantics Test for Contextualizing Language Models (AAAI 2022 · similarity 0.86)
- Geometry of Polysemy (ICLR 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.