Research map / Label-efficient and robust learning / Recognition and long-tail
Extending Multi-Sense Word Embedding to Phrases and Sentences for Unsupervised Semantic Applications
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 K=10 multi-facet codebook embeddings instead of a single embedding for sentence similarity (ablation) | Our c SC · STS-B | Pearson correlation (%): 47.6 -> 52.6 | Helped |
| Apply multi-facet attention weighting to GloVe Prob_WMD | GloVe Prob_WMD · STS-B | Pearson correlation (%): 63.1 -> 66.1 | Helped |
| Use K=10 attention facets rather than K=1 (ablation) | Our a Prob_WMD · STS-B | Pearson correlation (%): 62.9 -> 66.1 | Helped |
| Replace NNSC loss with k-means loss (ablation) | Our a K10 · STS-B | Pearson correlation (%): 66.1 -> 60.3 | Hurt |
| Replace Transformer encoder and decoder with LSTM (ablation) | Our a K10 SIF · STS-B | Pearson correlation (%): 64.3 -> 64.6 | No clear effect |
| Use K=100 sentence facets instead of K=3 (ablation) | Our c extractive summarization · CNN/Daily Mail | ROUGE-1 F1: 32.2 -> 35.0 | Helped |
| Use a multi-facet phrase embedding instead of a single facet (ablation) | Ours Emb · SemEval 2013 | AUC (%): 87.8 -> 85.6 | Hurt |
| Use a multi-facet phrase embedding instead of a single facet (ablation) | Ours Emb · Turney (5) | Accuracy (%): 50.3 -> 49.4 | Hurt |
| Use a multi-facet phrase embedding instead of a single facet (ablation) | Ours Emb · Turney (10) | Accuracy (%): 32.5 -> 31.8 | Hurt |
| Add encoder-decoder attention to the phrase model (ablation) | Phrase representation model · Phrase similarity | similarity performance | No clear effect |
| Remove decoder attention to contextualized input-word embeddings (ablation) | Sentence representation model · Sentence validation | validation loss | Hurt |
| Apply multi-facet attention weighting to sentence similarity scoring | GloVe Prob_avg · STS-B | Pearson correlation (%): 59.2 -> 61.4 | Helped |
Closest papers by meaning
- A Simple but Tough-to-Beat Baseline for Sentence Embeddings (ICLR 2017 · similarity 0.90)
- Learning Context-Specific Word/Character Embeddings (AAAI 2017 · similarity 0.88)
- SensEmBERT: Context-Enhanced Sense Embeddings for Multilingual Word Sense Disambiguation (AAAI 2020 · similarity 0.88)
- Learning Semantic Representations for Novel Words: Leveraging Both Form and Context (AAAI 2019 · similarity 0.88)
- Repurposing Language Models into Embedding Models: Finding the Compute-Optimal Recipe (NeurIPS 2024 · similarity 0.87)
- Narrowing the Gap between Supervised and Unsupervised Sentence Representation Learning with Large Language Model (AAAI 2024 · similarity 0.87)
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.