Research map / Label-efficient and robust learning / Recognition and long-tail
Robsut Wrod Reocginiton via Semi-Character Recurrent Neural Network
Keisuke Sakaguchi, Kevin Duh, Matt Post, Benjamin Van Durme
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 semi-character inputs that preserve first and last characters and bag internal characters | scRNN · Penn Treebank test, jumble | accuracy (%): 57.15 -> 98.96 | Improved |
| use semi-character inputs that preserve first and last characters and bag internal characters | scRNN · Penn Treebank test, delete | accuracy (%): 37.01 -> 85.74 | Improved |
| use semi-character inputs that preserve first and last characters and bag internal characters | scRNN · Penn Treebank test, insert | accuracy (%): 88.54 -> 96.70 | Improved |
| increase BPTT from β=1 to β=3 (ablation) | scRNN · Penn Treebank test, jumbled words | accuracy (%): 98.69 -> 98.96 | No clear effect |
| increase BPTT from β=3 to β=5 (ablation) | scRNN · Penn Treebank test, jumbled words | accuracy (%): 98.96 -> 98.91 | No clear effect |
| increase BPTT from β=3 to β=10 (ablation) | scRNN · Penn Treebank test, jumbled words | accuracy (%): 98.96 -> 98.95 | No clear effect |
| reduce LSTM hidden units to 50 (ablation) | scRNN · Penn Treebank test, jumbled words | accuracy (%): 98.96 -> 98.48 | No clear effect |
| reduce LSTM hidden units to 40 (ablation) | scRNN · Penn Treebank test, jumbled words | accuracy (%): 98.96 -> 96.90 | Hurt |
| reduce LSTM hidden units to 30 (ablation) | scRNN · Penn Treebank test, jumbled words | accuracy (%): 98.96 -> 94.15 | Hurt |
| reduce LSTM hidden units to 20 (ablation) | scRNN · Penn Treebank test, jumbled words | accuracy (%): 98.96 -> 84.82 | Hurt |
| reduce LSTM hidden units to 15 (ablation) | scRNN · Penn Treebank test, jumbled words | accuracy (%): 98.96 -> 73.32 | Hurt |
| reduce LSTM hidden units to 10 (ablation) | scRNN · Penn Treebank test, jumbled words | accuracy (%): 98.96 -> 48.43 | Hurt |
Closest papers by meaning
- Spell Once, Summon Anywhere: A Two-Level Open-Vocabulary Language Model (AAAI 2019 · similarity 0.84)
- LightRNN: Memory and Computation-Efficient Recurrent Neural Networks (NeurIPS 2016 · similarity 0.84)
- Learning Multi-Level Dependencies for Robust Word Recognition (AAAI 2020 · similarity 0.84)
- Sparsing Law: Towards Large Language Models with Greater Activation Sparsity (ICML 2025 · similarity 0.83)
- Training Neural Networks as Recognizers of Formal Languages (ICLR 2025 · similarity 0.83)
- Hierarchical Autoregressive Transformers: Combining Byte- and Word-Level Processing for Robust, Adaptable Language Models (ICLR 2025 · similarity 0.82)
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.
