Research map / Architectures and efficiency / RNNs and normalization
Recurrent Poisson Process Unit for Speech Recognition
Hengguan Huang, Hao Wang, Brian Mak
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 |
|---|---|---|---|
| Replace SRU/LSTM/quasi-RNN with RPPU | hybrid RNN-HMM acoustic model · CHiME-2 test | WER (%): 26.1 -> 24.4 | Improved |
| Replace LSTM with RPPU | hybrid RNN-HMM · WSJ0 eval92-5k | WER (%): 2.8 -> 2.5 | Improved |
| Replace LSTM with RPPU | hybrid RNN-HMM · WSJ0&1 dev93-20k | WER (%): 7.4 -> 6.9 | Improved |
| Replace LSTM with RPPU | hybrid RNN-HMM · WSJ0&1 eval93-20k | WER (%): 6.8 -> 6.2 | Improved |
| Replace LSTM with RPPU at low SNR | hybrid RNN-HMM · CHiME-2 test, -6 dB | WER (%): 42.4 -> 39.9 | Improved |
| Replace LSTM with RPPU at high SNR | hybrid RNN-HMM · CHiME-2 test, 9 dB | WER (%): 15.3 -> 13.2 | Improved |
| Use RPP-estimated time points to re-align frame-level targets | phoneme-level alignment · CHiME-2 dev | alignment similarity (%): 55.5 -> 64.7 | Improved |
| Add regularization term on intensity (weight gamma around 0.08) (ablation) | RPPU · CHiME-2 dev | WER (%) | Improved |
| Increase LSTM depth beyond three layers (ablation) | LSTM · CHiME-2 | WER (%) | No clear effect |
| Replace LSTM with SRU (12 layers) | hybrid RNN-HMM · CHiME-2 test | WER (%): 26.1 -> 26.2 | No clear effect |
| Replace LSTM with SRU for training speed | hybrid RNN-HMM · CHiME-2 | training time per epoch (hr): 0.71 -> 0.32 | Improved |
| Replace SRU with RPPU | SRU · CHiME-2 | training time per epoch (hr): 0.32 -> 0.37 | Hurt |
Closest papers by meaning
- Fully Neural Network Based Speech Recognition on Mobile and Embedded Devices (NeurIPS 2018 · similarity 0.86)
- Z-Forcing: Training Stochastic Recurrent Networks (NeurIPS 2017 · similarity 0.85)
- Recurrent Hidden Semi-Markov Model (ICLR 2017 · similarity 0.85)
- The Statistical Recurrent Unit (ICML 2017 · similarity 0.85)
- Variable Computation in Recurrent Neural Networks (ICLR 2017 · similarity 0.85)
- Mechanistic Interpretability of RNNs emulating Hidden Markov Models (NeurIPS 2025 · similarity 0.83)
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.
