Research map / Architectures and efficiency / RNNs and normalization
A Multilayer Convolutional Encoder-Decoder Neural Network for Grammatical Error Correction
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 |
|---|---|---|---|
| Ensemble four independently initialized convolutional models | MLConv · CoNLL-2014 test | F0.5: 45.36 -> 48.05 | Helped |
| Add edit-operation rescoring features | MLConv (4 ens.) · CoNLL-2014 test | F0.5: 48.05 -> 49.78 | Helped |
| Initialize embeddings with fastText trained on Wikipedia | MLConv (4 ens.) +EO · CoNLL-2014 test | F0.5: 49.78 -> 50.70 | Helped |
| Add a web-scale language-model feature for rescoring | MLConv embed (4 ens.) +EO · CoNLL-2014 test | F0.5: 50.70 -> 54.13 | Helped |
| Add the spell-checking component | MLConv embed (4 ens.) +EO +LM · CoNLL-2014 test | F0.5: 54.13 -> 54.79 | Helped |
| Add edit-operation rescoring features | MLConv embed (4 ens.) · JFLEG Test | GLEU: 51.06 -> 53.38 | Helped |
| Add a language-model feature for rescoring | MLConv embed (4 ens.) +EO · JFLEG Test | GLEU: 53.38 -> 55.99 | Helped |
| Add the spell-checking component | MLConv embed (4 ens.) +EO +LM · JFLEG Test | GLEU: 55.99 -> 57.47 | Helped |
| Use a multilayer convolutional encoder-decoder instead of a bidirectional LSTM (ablation) | GEC encoder-decoder · CoNLL-2013 test | F0.5: 29.84 -> 31.96 | Helped |
| Use multiple convolutional layers instead of a single convolutional layer (ablation) | GEC convolutional encoder-decoder · CoNLL-2013 test | F0.5: 26.39 -> 31.96 | Helped |
| Initialize embeddings with fastText character N-gram vectors rather than random vectors (ablation) | MLConv · CoNLL-2013 test | F0.5: 31.96 -> 32.97 | Helped |
| Initialize embeddings with Word2vec rather than random vectors (ablation) | MLConv · CoNLL-2013 test | F0.5: 31.96 -> 32.49 | Helped |
Closest papers by meaning
- ScholarGEC: Enhancing Controllability of Large Language Model for Chinese Academic Grammatical Error Correction (AAAI 2025 · similarity 0.85)
- ALRMR-GEC: Adjusting Learning Rate Based on Memory Rate to Optimize the Edit Scorer for Grammatical Error Correction (AAAI 2025 · similarity 0.84)
- Learning from others' mistakes: Finetuning machine translation models with span-level error annotations (ICML 2025 · similarity 0.84)
- DSGram: Dynamic Weighting Sub-Metrics for Grammatical Error Correction in the Era of Large Language Models (AAAI 2025 · similarity 0.84)
- Learning from Mistakes: Self-correct Adversarial Training for Chinese Unnatural Text Correction (AAAI 2025 · similarity 0.84)
- A Paradigm Shift in Machine Translation: Boosting Translation Performance of Large Language Models (ICLR 2024 · similarity 0.84)
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.