Massive Open-Vocabulary Keyword Spotting

2026-06-09

AI Summary

This work addresses the limitations of current automatic speech recognition systems in accurately recognizing rare domain-specific terms and the poor scalability of conventional open-vocabulary keyword spotting methods to large-scale lexicons. The authors propose an efficient open-vocabulary keyword detection framework that integrates contextual biasing with compressed feature storage, significantly enhancing recognition performance for rare terms without requiring fine-tuning of the underlying speech recognition model. By introducing an innovative feature representation and fusion mechanism, the method reduces memory consumption by up to 128× while enabling real-time detection over large-scale terminology banks for the first time. Moreover, it maintains entity recall rates on unseen languages comparable to those of uncompressed approaches, substantially improving the system’s scalability and practical applicability.

Abstract

Automatic speech recognition systems have been shown to under-perform when it comes to transcribing words rarely seen in the training data, namely specialized terminology. Open-vocabulary keyword spotting, combined with contextual biasing, has been shown to mitigate this issue. However, existing systems can only handle glossaries of a few hundred terms without becoming an infeasible bottleneck. We propose a system that stores features with a memory footprint up to 128 times smaller than a comparable baseline and allows users to process massive databases while remaining open-vocabulary. Without fine-tuning the speech recognition model, our system achieves a comparable entity recall as uncompressed solutions, even in languages not seen during training.

Problem

• open-vocabulary keyword spotting
• automatic speech recognition
• specialized terminology
• massive glossaries
• contextual biasing

Innovation

• open-vocabulary keyword spotting
• contextual biasing
• feature compression
• massive glossary
• zero-shot language transfer