A quantum search decoder for natural language processing

Abstract Probabilistic language models, e.g. those based on recurrent neural networks such as long short-term memory models (LSTMs), often face the problem of finding a high probability prediction from sequence random variables over set tokens. This is commonly addressed using form greedy decoding beam search, where limited number highest-likelihood paths (the width) decoder are kept, and at end maximum-likelihood path chosen. In this work, we construct quantum algorithm to find globally optimal parse (i.e. for infinite with constant success probability. When input follows power law exponent k > 0, our has runtime R n f ( , ) alphabet size, length; here < 1/2, $f\rightarrow 0$ f→0 exponentially fast increasing hence making always more than quadratically faster its classical counterpart. We further modify procedure recover finite width variant, which enables an even stronger empirical speedup while still retaining higher accuracy possible classically. Finally, apply search Mozilla’s implementation Baidu’s DeepSpeech net, show exhibit word rank frequency.