A Comparison of Techniques for Language Model Integration in Encoder-Decoder Speech Recognition

TL;DR

This paper compares various methods for integrating unpaired text data into encoder-decoder speech recognition models, revealing that simple shallow fusion often performs best initially, while more complex methods like cold fusion excel after rescoring on large datasets.

Contribution

It provides a comprehensive comparison of existing and new techniques for leveraging unpaired text in encoder-decoder speech recognition, filling a gap in prior research.

Findings

Shallow fusion performs best for first-pass decoding across datasets.

Cold fusion reduces oracle error rate and outperforms others after rescoring on Google data.

Using unpaired text improves speech recognition performance across methods and datasets.

Abstract

Attention-based recurrent neural encoder-decoder models present an elegant solution to the automatic speech recognition problem. This approach folds the acoustic model, pronunciation model, and language model into a single network and requires only a parallel corpus of speech and text for training. However, unlike in conventional approaches that combine separate acoustic and language models, it is not clear how to use additional (unpaired) text. While there has been previous work on methods addressing this problem, a thorough comparison among methods is still lacking. In this paper, we compare a suite of past methods and some of our own proposed methods for using unpaired text data to improve encoder-decoder models. For evaluation, we use the medium-sized Switchboard data set and the large-scale Google voice search and dictation data sets. Our results confirm the benefits of using unpaired text across a range of methods and data sets. Surprisingly, for first-pass decoding, the rather simple approach of shallow fusion performs best across data sets. However, for Google data sets we find that cold fusion has a lower oracle error rate and outperforms other approaches after second-pass rescoring on the Google voice search data set.