Third International Conference on Spoken Language Processing (ICSLP 94)
The first stage of text-to-speech (tts) conversion involves analyzing the text to determine the correct pronunciation of the individual words or word groups. Building upon a previously devised modular architecture for text processing within the context of a TTS system, we describe recent development in each of the areas "Word Pronunciation" and "Text Analysis", and describe the different modules involved. We describe how the tokeniser interacts with the lexicon in handling word groups (collocations), abbreviations and typing case. We also describe and give further references to on-going work, where the output of a probabilistic part-of-speech tagger will be used as input to a prosodic phrasing algorithm.
Bibliographic reference. Lindstrom, Anders / Ljungqvist, Mats (1994): "Text processing within a speech synthesis system", In ICSLP-1994, 1683-1686.