
A MultiLingual MultiTask (MLMT) model that integrates multilingual speech generation and recognition tasks into a single large language model, and a data construction strategy that creates code-switched data without relying on high-quality code-switched data.


coremsg

enhancing-multilingual-speech-generation-and-recognition-capabilities-in-large-language-models-using-constructed-code-switched-data


Enhancing Multilingual Speech Generation and Recognition Capabilities in Large Language Models Using Constructed Code-Switched Data


title_rewrite


Phoneme-based models can achieve strong crosslinguistic generalizability to unseen languages for open-vocabulary keyword spotting and zero-shot forced alignment.


multilingual-phoneme-based-speech-processing-towards-open-vocabulary-keyword-spotting-and-forced-alignment-in-any-language


Multilingual Phoneme-based Speech Processing: Towards Open-Vocabulary Keyword Spotting and Forced Alignment in Any Language
