Dimensions of Segmental Variability: Interaction of Prosody and Surprisal in Six Languages

Abstract

Contextual predictability variation affects phonological and phonetic structure. Reduction and expansion of acoustic-phonetic features is also characteristic of prosodic variability. In this study, we assess the impact of surprisal and prosodic structure on phonetic encoding, both independently of each other and in interaction. We model segmental duration, vowel space size and spectral characteristics of vowels and consonants as a function of surprisal as well as of syllable prominence, phrase boundary, and speech rate. Correlates of phonetic encoding density are extracted from a subset of the BonnTempo corpus for six languages: American English, Czech, Finnish, French, German, and Polish. Surprisal is estimated from segmental n-gram language models trained on large text corpora. Our findings are generally compatible with a weak version of Aylett and Turk's Smooth Signal Redundancy hypothesis, suggesting that prosodic structure mediates between the requirements of efficient communication and the speech signal. However, this mediation is not perfect, as we found evidence for additional, direct effects of changes in surprisal on the phonetic structure of utterances. These effects appear to be stable across different speech rates.

Different languages, similar encoding efficiency: Comparable information rates across the human communicative niche

“…The RMarkdown script continuing all the R code needed to reproduce the results and plots reported here. References (50)(51)(52)(53)(54)(55)(56)(57)(58)(59) program (ANR-11-IDEX-0007) operated by the National Research Agency (ANR). Author contributions: C.C., Y.O., and F.P.…”

Modeling the predictive potential of extralinguistic context with script knowledge: The case of fragments

“…The tendency to reduce predictable expressions has been observed on different levels of linguistic analysis, ranging from phonetics [1,[38][39][40][41][42][43][44][45][46][47] and morphology [48] to the omission of predictable words [4,[6][7][8][9]49]. If this principle applies to fragments as well, we expect that fragments are more strongly preferred over the corresponding full sentence if the omission of words that are predictable in a specific context results in a well-formed fragment.…”

Speech Synthesis Evaluation — State-of-the-Art Assessment and Suggestion for a Novel Research Program

“…We contend that just as human speech production is highly variable and comes in many different "styles", which are continuously adapted by speakers given dynamically changing social (tutoring, chatting, arguing, counseling...), individual (hearing problems, attitude, level of distraction, motivation, familiarity), linguistic (frequency, predictability, suprisal, importance) or environmental settings (external noise, mutual visibility, ...) [11,12,13,14,15,16,17,18]. Due to this inherent contextual embedding, human speech production can never be "neutral" or "perfectly natural", and no speaking style therefore qualifies as a reference signal that a speech event of inherently less quality, e.g.…”