Use of boundary magnitude information based on syntactic condition in the development of pitch control

Abstract

Under an assumption that boundary forms a continuum in magnitude rather than a dichotomy (presence or absence), a perception experiment in which listeners detected a boundary signal in running speech was conducted. The result indicated that there was a positive correlation between the number of listeners perceiving a boundary signal and the duration of the word prior to the boundary. A detailed data analysis showed that syntactic condition was one of the dominant factors determining the magnitude of a boundary between two words. A variety of syntactic conditions were classified into five groups according to boundary magnitude. Group I (the highest): between clauses, between nouns and their appositions, etc. Group 2: between nouns and modifying clauses headed by “that” or by a preposition plus a relative pronoun, between introductory prepositional phrases and the subject. etc. Group 3: between the subject‐noun and the predicate‐verb, between nouns and modifying prepositional phrases, etc. Group 4: between adjectives and nouns in noun phrases, between verbs and prepositional phrases, etc. This boundary magnitude information leads to a pitch control with different degrees of phrase termination which produces a desirable variety in intonation for longer sentences and paragraphs in synthetic speech.