Efficient lexical retrieval for English text-to-speech synthesis

We present a first version of a filter dictionary for use in a computer-telephony text-to-speech synthesis system. The aim of the filter dictionary was to provide a lexicon that was compact, fast and had broader coverage than the standard dictionary used to create it. Correct phonemic transcriptions and lexical stress assignment were both required for a transcription to be deemed accurate. The approach taken here guarantees 100% accurate coverage of the original dictionary, but also gives 93% accurate transcription of the expected coverage of novel words. Lexical stress and the phonemic transcription were retrieved in one pass, resulting in an extremely fast system. We also allowed userdefinition to retain accuracy for non-standard transcriptions. This algorithm was developed for British English, but could be applied to other languages.