Yun Zhu(Google), Parisa Haghani(Google), Anshuman Tripathi(Google), Bhuvana Ramabhadran(Google), Brian Farris(Google), Hainan Xu(Google), Han Lu(Google), Hasim Sak(Google), Isabel Leal(Google), Neeraj Gaur(Google), Pedro Moreno(google inc.) and Qian Zhang(Google)
Multilingual automatic speech recognition systems can transcribe utterances from different languages. These systems are attractive from different perspectives: they can provide quality improvements, specially for lower resource languages, and simplify the training and deployment procedure. End-to-end speech recognition has further simplified multilingual modeling as one model, instead of several components of a classical system, have to be unified. In this paper, we investigate a stream- able end-to-end multilingual system based on the Transformer Transducer. We propose several techniques for adapting the self-attention architecture based on the language id. We analyze the trade-offs of each method with regards to quality gains and number of additional parameters introduced. We conduct experiments in a real-world task consisting of five languages. Our experimental results demonstrate 8% to 20% relative gain over the baseline multilingual model.