Motif discovery in infant- and adult-directed speech

摘要

Infant-directed speech (IDS) is thought to play a key role in determining infant language acquisition. It is thus important to describe how computational models of infant language acquisition behave when given an input of IDS, as compared to adult-directed speech (ADS). In this paper, we explore how an acoustic motif discovery algorithm fares when presented with speech from both registers. Results show small but significant differences in performance, with lower recall and lower cluster collocation in IDS than ADS, but a higher cluster purity in IDS. Overall, these results are inconsistent with a view suggesting that IDS is acoustically clearer than ADS in a way that systematically facilitates lexical recognition. Similarities and differences with human infants' word segmentation are discussed.