UML: A Universal Monolingual Output Layer for Multilingual ASR
This addresses efficiency and scalability issues in multilingual ASR systems, though it appears incremental as it builds on existing word-piece models.
The paper tackled the problem of overly large output layers in multilingual automatic speech recognition (ASR) by proposing a universal monolingual output layer (UML) that shares nodes across languages, resulting in a smaller and more efficient model. Experimental results on an 11-language task demonstrated high-quality and high-efficiency streaming ASR.
Word-piece models (WPMs) are commonly used subword units in state-of-the-art end-to-end automatic speech recognition (ASR) systems. For multilingual ASR, due to the differences in written scripts across languages, multilingual WPMs bring the challenges of having overly large output layers and scaling to more languages. In this work, we propose a universal monolingual output layer (UML) to address such problems. Instead of one output node for only one WPM, UML re-associates each output node with multiple WPMs, one for each language, and results in a smaller monolingual output layer shared across languages. Consequently, the UML enables to switch in the interpretation of each output node depending on the language of the input speech. Experimental results on an 11-language voice search task demonstrated the feasibility of using UML for high-quality and high-efficiency multilingual streaming ASR.