Timezone: »

 
Poster
Imputer: Sequence Modelling via Imputation and Dynamic Programming
William Chan · Chitwan Saharia · Geoffrey Hinton · Mohammad Norouzi · Navdeep Jaitly

Wed Jul 15 08:00 AM -- 08:45 AM & Wed Jul 15 07:00 PM -- 07:45 PM (PDT) @ None #None
in Poster Session 22 »

This paper presents the Imputer, a neural sequence model that generates output sequences iteratively via imputations. The Imputer is an iterative generation model, requiring only a constant number of generation steps independent of the number of input or output tokens. The Imputer can be trained to approximately marginalize over all possible alignments between the input and output sequences, and all possible generation orders. We present a tractable dynamic programming training algorithm, which yields a lower bound on the log marginal likelihood. When applied to end-to-end speech recognition, the Imputer outperforms prior non-autoregressive models and achieves competitive results to autoregressive models. On LibriSpeech test-other, the Imputer achieves 11.1 WER, outperforming CTC at 13.0 WER and seq2seq at 12.5 WER.

Keywords: Deep Sequence Models · Speech Processing · Natural Language Processing / Dialogue · Sequential, Network, and Time-Series Modeling

Author Information

William Chan (Google Brain)
Chitwan Saharia (Google)
Geoffrey Hinton (Google)
Mohammad Norouzi (Google Research, Brain Team)
Navdeep Jaitly (D. E. Shaw)

More from the Same Authors