Research map / Label-efficient and robust learning / Recognition and long-tail
Learning Multi-Level Dependencies for Robust Word Recognition
Zhiwei Wang, Hui Liu, Jiliang Tang, Songfan Yang, Gale Yan Huang, Zitao Liu
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 MUDE's character-level encoder and word-level contextual model instead of scRNN | Word recognition · PER, DEL, INS, SUB | accuracy (%): 98.23 -> 98.81; 91.55 -> 95.86; 95.95 -> 97.16; 87.09 -> 90.52 | Improved |
| Use MUDE's character-level encoder and word-level contextual model instead of scRNN | Word recognition · W-PER, W-DEL, W-INS, W-SUB | accuracy (%): 97.36 -> 98.75; 89.99 -> 93.29; 95.96 -> 97.10; 81.12 -> 85.17 | Improved |
| Add character-level sequence-to-sequence training (beta 1 rather than beta 0) (ablation) | MUDE · W-PER | prediction loss | Improved |
| Add character-level sequence-to-sequence training (beta 1 rather than beta 0) (ablation) | MUDE · W-PER | sequence-to-sequence loss | Improved |
| Train with W-ALL noise augmentation | MUDE · Deletion and substitution noise, including whole-word variants | accuracy (%) | Improved |
| Train with W-ALL noise augmentation | MUDE · Permutation and insertion noise | accuracy (%) | Hurt |
| Train on one noise type and test on a different, dissimilar type | MUDE · Penn Treebank noise generalization | accuracy (%) | Hurt |
Closest papers by meaning
- Word-Level Speech Recognition With a Letter to Word Encoder (ICML 2020 · similarity 0.86)
- Training and Evaluating Improved Dependency-Based Word Embeddings (AAAI 2018 · similarity 0.85)
- Tying Word Vectors and Word Classifiers: A Loss Framework for Language Modeling (ICLR 2017 · similarity 0.84)
- Sequence Modeling via Segmentations (ICML 2017 · similarity 0.84)
- Hierarchical Autoregressive Transformers: Combining Byte- and Word-Level Processing for Robust, Adaptable Language Models (ICLR 2025 · similarity 0.83)
- Training Neural Networks as Recognizers of Formal Languages (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.
