Acoustic data-driven lexicon learning based on a greedy pronunciation selection framework
This addresses the need for efficient lexicon learning in ASR for irregularly-spelled languages, though it is incremental as it builds on existing G2P and pruning methods.
The paper tackles the problem of automatically learning pronunciations for words in irregularly-spelled languages like English, where hand-written lexicons are typically required, by integrating letter sequences and acoustic evidence to prune entries and improve ASR performance. Experiments show the learned lexicon performs close to a full expert lexicon in WER, outperforming G2P-based or probability-pruned lexicons.
Speech recognition systems for irregularly-spelled languages like English normally require hand-written pronunciations. In this paper, we describe a system for automatically obtaining pronunciations of words for which pronunciations are not available, but for which transcribed data exists. Our method integrates information from the letter sequence and from the acoustic evidence. The novel aspect of the problem that we address is the problem of how to prune entries from such a lexicon (since, empirically, lexicons with too many entries do not tend to be good for ASR performance). Experiments on various ASR tasks show that, with the proposed framework, starting with an initial lexicon of several thousand words, we are able to learn a lexicon which performs close to a full expert lexicon in terms of WER performance on test data, and is better than lexicons built using G2P alone or with a pruning criterion based on pronunciation probability.