Research map / Architectures and efficiency / Attention and transformers
Semantic Sentence Matching with Densely-Connected Recurrent and Co-Attentive Information
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 the autoencoder bottleneck (ablation) | DRCN · SNLI test | accuracy: 88.7 -> 88.9 | Helped |
| remove trainable word embeddings (ablation) | DRCN · SNLI dev | accuracy | Hurt |
| remove fixed word embeddings (ablation) | DRCN · SNLI dev | accuracy | Hurt |
| add trainable rather than fixed word embeddings (ablation) | DRCN · SNLI dev | accuracy | Mixed |
| remove dense connections over co-attentive features (ablation) | DRCN · SNLI dev | accuracy | Hurt |
| remove dense connections over recurrent features (ablation) | DRCN · SNLI dev | accuracy | Hurt |
| remove dense connections over both co-attentive and recurrent features (ablation) | DRCN · SNLI dev | accuracy: 88.5 | Hurt |
| replace dense recurrent and co-attentive connections with residual connections (ablation) | DRCN · SNLI dev | accuracy | Hurt |
| remove dense connections over word-embedding features from the residual variant (ablation) | DRCN residual variant · SNLI dev | accuracy | Hurt |
| add inter-layer connections to a 5-layer RNN (ablation) | DRCN · SNLI dev | accuracy | Helped |
| add attention (ablation) | DRCN · SNLI dev | accuracy | Helped |
| increase the number of recurrent layers in models without inter-layer connections (ablation) | basic RNN with attention and basic RNN without attention · SNLI dev | accuracy | Hurt |
Closest papers by meaning
- DRr-Net: Dynamic Re-Read Network for Sentence Semantic Matching (AAAI 2019 · similarity 0.89)
- A Deep Architecture for Semantic Matching with Multiple Positional Sentence Representations (AAAI 2016 · similarity 0.88)
- Scalable Attentive Sentence Pair Modeling via Distilled Sentence Embedding (AAAI 2020 · similarity 0.87)
- Siamese Recurrent Architectures for Learning Sentence Similarity (AAAI 2016 · similarity 0.86)
- An Unsupervised Model With Attention Autoencoders for Question Retrieval (AAAI 2018 · similarity 0.86)
- Making the Relation Matters: Relation of Relation Learning Network for Sentence Semantic Matching (AAAI 2021 · 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.