An Expert System For The Production Of Phoneme Strings From Unmarked English Text Using Machine-Induced Rules

The speech synthesis group at the ComputerBased Education Research Laboratory (CERL) of the University of Illinois at Urbana-Champalgn is developing a diphone speech synthesis system based on pltch-adaptive short-tlme Fourier transforms. This system accepts the phonemic specification of an utterance along with pitch, time, and amplitude warping functions in order to produce high quality speech output from stored dlphone templates. This paper describes the operation of a program which operates as a front end for the dlphone speech synthesis system. The UTTER (for "Unmarked Text Transcription by Expert Rule") system maps English text onto a phoneme string, which is then used as an input to the dlphone speech synthesis system. The program is a twotiered Expert System which operates first on the word level and then on the (vowel or consonant) cluster level. The system's knowledge about pronunciation is organized in two decision trees automatically generated by an induction algorithm on a dynamically specified "training set" of examples. in that they are often unable to cope with a letter pattern that maps onto more than one phoneme pattern. Extreme cases are those words which, although differing in pronunciation, share orthographic representations (an analogous problem exists in speech recognition, where words which share phonemic representations differ in orthographic representation, and therefore possibly in semantic interpretation). A notable exception is the MIT speech synthesis system fAllen81] which is llngulstlcally-based, but not solely phoneme-based. A desirable feature in any rule-based system is the ability to automatically acquire or modify its own rules. Previous work [Oakey81] applies this automatic inference process to the text-tophoneme transcription problem. Unfortunately, Onkey's system is strlctly letter-based and suffers from the same deficiencies as other nonilnguistlcally-based systems. The UTTER system is an attempt to provide a llngulstlcally-based transcription system which has the ability to automatically acquire its own rule base.