Speech synthesis from english text: A progress report

Recent improvements in a system for speech synthesis from ordinary English (Coker, Umeda & Browman, Au-21, 1973, 293-298) include: (i) more accurate timing rules, based on extensive measurements of natural speech (see Umeda); (2) extended rules for allophonic variation as a function of word- and other boundaries; (3) extensive low-level (usually compensatory) adjustments in speech production --- assimilation, flaps, changes within consonant cluster, syllabic consonants, and context-dependent vowel glides; (4) a low sampling-rate model of vocal-cord and noise excitation. We summarize these, paying particular attention to items (3) and (4).