The role of knowledge-derived and signal-derived knowledge in speech segmentation

old_uid2277
titleThe role of knowledge-derived and signal-derived knowledge in speech segmentation
start_date2007/02/19
schedule11h-12h30
onlineno
summaryResearch shows that speech segmentation arises from mechanisms involving both lexical/sentential knowledge and sublexical cues (e.g., stress, allophony, phonotactics). However, few studies have indicated how these mechanisms operate in a complex, multi-cue environment. Our research delineates (a) the default hierarchy of weights among segmentation mechanisms/cues, (b) how the hierarchy changes as a function of some interpretive conditions (e.g., signal degradation, lack of lexical or contextual information), (c) the extent to which lexical and sentential knowledge modifies our perception of low-level cues for word boundaries.
responsiblesInformation non disponible