2021/04/09 by W. Ronny Huang, Huang, W. Ronny, Tara N. Sainath +9 · 1 citation
Computer Science · Engineering · #Audio and Speech Processing (eess.AS) #Computation and Language (cs.CL) #FOS: Computer and information sciences #FOS: Electrical engineering #Sound (cs.SD) #cs.CL #cs.SD #eess.AS #electronic engineering #information engineering
paper · pdf · doi:10.48550/arxiv.2104.04552
Presented as conference paper at Interspeech 2021
arxiv created 2021/06/07 · arxiv updated 2021/06/08
We introduce Lookup-Table Language Models (LookupLM), a method for scaling up the size of RNN language models with only a constant increase in the floating point operations, by increasing the expressivity of the embedding table. In particular, we instantiate an (additional) embedding table which embeds the previous n-gram token sequence, rather than a single token. This allows the embedding table to be scaled up arbitrarily -- with a commensurate increase in performance -- without changing the token vocabulary. Since embeddings are sparsely retrieved from the table via a lookup; increasing the size of the table adds neither extra operations to each forward pass nor extra parameters that need to be stored on limited GPU/TPU memory. We explore scaling n-gram embedding tables up to nearly a billion parameters. When trained on a 3-billion sentence corpus, we find that LookupLM improves long tail log perplexity by 2.44 and long tail WER by 23.4% on a downstream speech recognition task over a standard RNN language model baseline, an improvement comparable to a scaling up the baseline by 6.2x the number of floating point operations.