Record Deduplication for Entity Distribution Modeling in ASR Transcripts
This work improves contextual biasing for voice assistants by handling ASR misrecognitions, offering an incremental enhancement to entity distribution modeling.
The paper tackles the problem of modeling true entity distributions from ASR transcripts for contextual biasing in voice assistants, using record deduplication to address misrecognitions, achieving 95% retrieval of misrecognized entities and a 5% relative word error rate reduction.
Voice digital assistants must keep up with trending search queries. We rely on a speech recognition model using contextual biasing with a rapidly updated set of entities, instead of frequent model retraining, to keep up with trends. There are several challenges with this approach: (1) the entity set must be frequently reconstructed, (2) the entity set is of limited size due to latency and accuracy trade-offs, and (3) finding the true entity distribution for biasing is complicated by ASR misrecognition. We address these challenges and define an entity set by modeling customers true requested entity distribution from ASR output in production using record deduplication, a technique from the field of entity resolution. Record deduplication resolves or deduplicates coreferences, including misrecognitions, of the same latent entity. Our method successfully retrieves 95% of misrecognized entities and when used for contextual biasing shows an estimated 5% relative word error rate reduction.