Model Adaptation · Data artifact
ASR Word Boost List
Data artifactModel AdaptationModelsarc:ASRWordBoostList
A configuration listing domain-specific terms and boost scores that add a decoding bonus to transcript hypotheses containing those terms.
Responsibility. Biases speech decoding toward expected domain vocabulary without retraining.
Also known as: Word boosting list, Boosted vocabulary
Relationships
configures structural
- Automatic Speech Recognizer abstract Ch7.5
Design guidance
- SHOULD be used as the simple first customization; it cannot overcome strong acoustic mismatch and is not a substitute for acoustic adaptation in highly specialized domains.
Quantitative guidance
As stated by the sources; verify before use.
- About 30 minutes setup; improves domain-term accuracy 5-15% when acoustics are ambiguous; example boosted_score 20.0 (Ch7.5).
Classification
- Patterns
- Additive beam-search score bonus
- Technologies
- NVIDIA Riva ASR
- Quality attributes
- Functional suitability: correctness and validity (ISO/IEC 25010 | NIST AI RMF: valid)Maintainability (ISO/IEC 25010)
- Risks mitigated
- Homophone errors on domain terms (e.g., 'pain' vs. 'pane')
Sources
- Ch7.5: T. Nguyen, "NeMo Curator, Riva Speech AI & Multimodal Integration," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 7.5. ISBN: 9798244538229.