Efficient Trie-based Biasing using K-step Prediction for Rare Word Recognition
This work addresses efficiency and accuracy issues in rare word recognition for ASR systems, representing an incremental improvement over existing Trie-based biasing methods.
The paper tackles the computational expense and limitations of Trie-based biasing for rare word recognition in ASR by proposing a method that adapts models to predict multiple steps ahead, avoiding revocation steps. By fine-tuning Whisper with 10 hours of synthetic data, it reduces word error rate on the NSC Part 2 test set from 30.86% to 12.19%.
Contextual biasing improves rare word recognition of ASR models by prioritizing the output of rare words during decoding. A common approach is Trie-based biasing, which gives "bonus scores" to partial hypothesis (e.g. "Bon") that may lead to the generation of the rare word (e.g. "Bonham"). If the full word ("Bonham") isn't ultimately recognized, the system revokes those earlier bonuses. This revocation is limited to beam search and is computationally expensive, particularly for models with large decoders. To overcome these limitations, we propose adapting ASR models to look ahead and predict multiple steps at once. This avoids the revocation step entirely by better estimating whether a partial hypothesis will lead to the generation of the full rare word. By fine-tuning Whisper with only 10 hours of synthetic data, our method reduces the word error rate on the NSC Part 2 test set from 30.86% to 12.19%.