Research map / Architectures and efficiency / RNNs and normalization
How recurrent networks implement contextual processing in sentiment analysis
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 convolutional modifier effects with learned modifier-specific weights to Bag of Words | Bag of words · Yelp 2015 test | accuracy: 93.57% -> 95.40% | Helped |
| Add convolutional modifier effects with learned modifier-specific weights to Bag of Words | Bag of words · IMDB test | accuracy: 89.47% -> 90.68% | Helped |
| Add learned beginning- and end-of-document effects to the modifier-augmented model (ablation) | CoMW + β_mod · Yelp 2015 test | accuracy: 95.40% -> 95.63% | Helped |
| Add learned beginning- and end-of-document effects to the modifier-augmented model (ablation) | CoMW + β_mod · IMDB test | accuracy: 90.68% -> 90.75% | Helped |
| Add convolutional modifier words with shared regression weights (ablation) | Bag of words · Yelp 2015 test | accuracy: 93.57% -> 94.76% | Helped |
| Add convolutional modifier words with shared regression weights (ablation) | Bag of words · IMDB test | accuracy: 89.47% -> 89.90% | Helped |
| Add BOD and EOD tokens with shared regression weights (ablation) | Bag of words · Yelp 2015 test | accuracy: 93.57% -> 93.99% | Helped |
| Add BOD and EOD tokens with shared regression weights (ablation) | Bag of words · IMDB test | accuracy: 89.47% -> 89.57% | Helped |
| Project hidden states out of the modifier subspace at each step (ablation) | GRU · Yelp 2015 test | accuracy | Hurt |
| Project hidden states out of a random subspace at each step (ablation) | GRU · Yelp 2015 test | accuracy | No clear effect |
| Project the learned initial state out of the modifier subspace (ablation) | GRU · Yelp 2015 test | accuracy: 0.17% | Hurt |
| Project the learned initial state out of the modifier subspace (ablation) | GRU · Yelp 2015 train | accuracy: 0.23% | Hurt |
Closest papers by meaning
- Reverse engineering recurrent networks for sentiment classification reveals line attractor dynamics (NeurIPS 2019 · similarity 0.90)
- TopicRNN: A Recurrent Neural Network with Long-Range Semantic Dependency (ICLR 2017 · similarity 0.86)
- Beyond Word Importance: Contextual Decomposition to Extract Interactions from LSTMs (ICLR 2018 · similarity 0.86)
- CA-RNN: Using Context-Aligned Recurrent Neural Networks for Modeling Sentence Similarity (AAAI 2018 · similarity 0.86)
- Context-Sensitive Twitter Sentiment Classification Using Neural Network (AAAI 2016 · similarity 0.85)
- RNNs are not Transformers (Yet): The Key Bottleneck on In-Context Retrieval (ICLR 2025 · similarity 0.85)
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.