On the articulation between acoustic and semantic uncertainty in speech perception: Investigating the interaction between sources of information in perceptual classification.

Abstract

Listeners processing speech signals have to deal with two main classes of uncertainty occurring in the vicinity of a given speech segment: both acoustic properties of the contextual environment (Ladefoged and Broadbent, 1957; Sjerps and McQueen, 2013) and lexical hypotheses based on word co-occurence probabilities or semantic relations (e.g., Connine, 1987; Gow and Olson, 2015) may affect the interpretation of a given sound. We investigate this issue by independently manipulating (1) semantic relationships between words using word embeddings estimations and (2) acoustic relations between a contextual part and the final word in the sentence. Based on word pairs that contrast on their vowel target only (e.g., french “balle” versus “belle”, pronounced /bal/ vs. /bEl/ -- eng. “ball” versus “beauty”), 3 types of sentences are generated: (1) a sentence that would semantically “prime” the word /bal/ (“Le joueur a dévié la”, eng. “The player deflected the”), (2) a sentence that would favour the word /bEl/ “Le prince a charmé la”, eng. “The prince charmed the”), and (3) a semantically incongruous sentence in both cases “Le journaliste a découvert la”, eng. “The journalist discovered the.” Listeners are presented with fully ambiguous final words (acoustically located between, e.g., /bal/ and /bEl/) in contexts where semantic influence varies (sentence-types 1/2/3) and is balanced with acoustic manipulations of formant frequencies favouring one word or the other. This will provide cues to modelling how both sources of entropy alter speech perception.