Fused Acoustic and Text Encoding for Multimodal Bilingual Pretraining and Speech Translation

Recently text and speech representation learning has successfully improved many language related tasks. However, all existing methods only learn from one input modality, while a unified acoustic and text representation is desired by many speech-related tasks such as speech translation. We propose a Fused Acoustic and Text Masked Language Model (FAT-MLM) which jointly learns a unified representation for both acoustic and text in-put. Within this cross modal representation learning framework, we further present an end-to-end model for Fused Acoustic and Text Speech Translation (FAT-ST). Experiments on three translation directions show that our proposed speech translation models fine-tuned from FAT-MLM substantially improve translation quality (+5.90 BLEU).