The effect of pruning and compression on graphical representations of the output of a speech recognizer

Abstract

Large vocabulary continuous speech recognition can benefit from an efficient data structure for representing a large number of acoustic hypotheses compactly. Word graphs or lattices have been chosen as such an efficient interface between acoustic recognition engines and subsequent language processing modules. This paper first investigates the effect of pruning during acoustic decoding on the quality of word lattices and shows that by combining different pruning options (at the model level and word level), we can obtain word lattices with comparable accuracy to the original lattices and a manageable size. In order to use the word lattices as the input for a post-processing language module, they should preserve the target hypotheses and their scores while being as small as possible. In this paper, we introduce a word graph compression algorithm that significantly reduces the number of words in the graphical representation without eliminating utterance hypotheses or distorting their acoustic scores. We compare this word graph compression algorithm with several other lattice size-reducing approaches and demonstrate the relative strength of the new word graph compression algorithm for decreasing the number of words in the representation. Experiments are conducted across corpora and vocabulary sizes to determine the consistency of the pruning and compression results.

A Simplification Algorithm for Visualizing the Structure of Complex Graphs

“…Additional simplification algorithms have been proposed to assist in robot path planning [17], classifying the topology of surfaces [3], speech recognition [13], and improving the computational complexity and memory requirements of dense graph processing algorithms [12,18,6].…”

The effect of lattice pruning on MMIE training

“…If no pruning is done, the lattice can be highly accurate but also exorbitantly large. One of the most effective lattice pruning methods is the beam pruning during decoding [5]. Another very common method is to further compress the lattice after it has been generated, where a word lattice is usually converted to a word graph and all the time alignment information is discarded, such as the finite state automata determination and minimization [6] and the confusion network approach [7].…”