Research map / Architectures and efficiency / RNNs and normalization
Definition Modeling: Learning to Define Word Embeddings in Natural Language
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 gated update to the Seed model (ablation) | Definition model · Dictionary test set | perplexity: 56.350 -> 50.949 | Helped |
| Add gated update to the Seed model (ablation) | Definition model · Definition generation: Avg BLEU | BLEU: 24.58 -> 27.01 | Helped |
| Provide the target-word embedding as a constant input at every time step (ablation) | Seed definition model · Dictionary test set | perplexity: 56.350 -> 57.372 | Hurt |
| Update the hidden representation with the target-word embedding without gates (ablation) | Seed definition model · Dictionary test set | perplexity: 56.350 -> 58.147 | Hurt |
| Add a character-level CNN to the gated model (ablation) | S+G definition model · Dictionary test set | perplexity: 50.949 -> 48.566 | Helped |
| Add a character-level CNN to the gated model (ablation) | S+G definition model · Definition generation: Avg BLEU | BLEU: 27.01 -> 27.62 | Helped |
| Add hypernym embeddings to the gated model with character features (ablation) | S+G+CH definition model · Dictionary test set | perplexity: 48.566 -> 48.168 | Helped |
| Add hypernym embeddings to the gated model with character features (ablation) | S+G+CH definition model · Definition generation: Avg BLEU | BLEU: 27.62 -> 27.46 | No clear effect |
| Replace learned Word2Vec embeddings with randomized embeddings (ablation) | S+G+CH definition model · Dictionary test set | perplexity: 48.566 -> 100.43 | Hurt |
| Remove repeated phrases and output only the most likely definition (ablation) | Seed generation · Definition generation: Avg BLEU | BLEU: 24.58 -> 25.01 | Helped |
| Remove repeated phrases and output only the most likely definition (ablation) | S+G+CH+HE generation · Definition generation: Avg BLEU | BLEU: 27.46 -> 29.65 | Helped |
| Add gated update and character-level CNN to Seed (ablation) | Definition model · Reverse dictionary | accuracy@1: 0.175 -> 0.307 | Helped |
Closest papers by meaning
- Learning Semantic Representations for Novel Words: Leveraging Both Form and Context (AAAI 2019 · similarity 0.90)
- All Word Embeddings from One Embedding (NeurIPS 2020 · similarity 0.89)
- Learning Context-Specific Word/Character Embeddings (AAAI 2017 · similarity 0.89)
- Understanding the Semantic Content of Sparse Word Embeddings Using a Commonsense Knowledge Base (AAAI 2020 · similarity 0.88)
- By Tying Embeddings You Are Assuming the Distributional Hypothesis (ICML 2024 · similarity 0.87)
- Closed-Form Training Dynamics Reveal Learned Features and Linear Structure in Word2Vec-like Models (NeurIPS 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.