Word pair approximation for more efficient decoding with high-order language models

The search effort in LVCSR depends on the order of the language model (LM); search hypotheses are only recombined once the LM allows for it. In this work we show how the LM dependence can be partially eliminated by exploiting the well-known word pair approximation. We enforce preemptive unigramor bigram-like LM recombination at word boundaries. We capture the recombination in a lattice, and later expand the lattice using LM rescoring. LM rescoring unfolds the same search space which would have been encountered without the preemptive recombination, but the overall efficiency is improved, because the amount of redundant HMM expansion in different LM contexts is reduced. Additionally, we show how to expand the recombined hypotheses on-the-fly, omitting the intermediate lattice form. Our new approach allows using the full n-gram LM for decoding, but based on a compact unigramor bigram search space. We show that our approach works better than common lattice rescoring pipelines, where a pruned lowerorder LM is used to generate lattices; such pipelines suffer from the weak lower-order LM, which guides the pruning suboptimally. Our new decoding approach improves the runtime efficiency by up to 40% at equal precision when using a large vocabulary and high-order LM.