Dataset Information

Unsupervised learning of vowel categories from infant-directed speech.

ABSTRACT: Infants rapidly learn the sound categories of their native language, even though they do not receive explicit or focused training. Recent research suggests that this learning is due to infants' sensitivity to the distribution of speech sounds and that infant-directed speech contains the distributional information needed to form native-language vowel categories. An algorithm, based on Expectation-Maximization, is presented here for learning the categories from a sequence of vowel tokens without (i) receiving any category information with each vowel token, (ii) knowing in advance the number of categories to learn, or (iii) having access to the entire data ensemble. When exposed to vowel tokens drawn from either English or Japanese infant-directed speech, the algorithm successfully discovered the language-specific vowel categories (/I, i, epsilon, e/ for English, /I, i, e, e/ for Japanese). A nonparametric version of the algorithm, closely related to neural network models based on topographic representation and competitive Hebbian learning, also was able to discover the vowel categories, albeit somewhat less reliably. These results reinforce the proposal that native-language speech categories are acquired through distributional learning and that such learning may be instantiated in a biologically plausible manner.

SUBMITTER: Vallabha GK

PROVIDER: S-EPMC1934922 | biostudies-literature | 2007 Aug

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Unsupervised learning of vowel categories from infant-directed speech.

Vallabha Gautam K GK McClelland James L JL Pons Ferran F Werker Janet F JF Amano Shigeaki S

Proceedings of the National Academy of Sciences of the United States of America 20070730 33

Infants rapidly learn the sound categories of their native language, even though they do not receive explicit or focused training. Recent research suggests that this learning is due to infants' sensitivity to the distribution of speech sounds and that infant-directed speech contains the distributional information needed to form native-language vowel categories. An algorithm, based on Expectation-Maximization, is presented here for learning the categories from a sequence of vowel tokens without ( ...[more]

PMID: 17664424

Similar Datasets

Project description:ObjectivesRelatively little is known about how young children with hearing impairment (HI) learn novel words in infant- and adult-directed speech (ADS). Infant-directed speech (IDS) supports word learning in typically developing infants relative to ADS. This study examined how children with normal hearing (NH) and children with HI learn novel words in IDS and ADS. It was predicted that IDS would support novel word learning in both groups of children. In addition, children with HI were expected to be less proficient word learners as compared with their NH peers.DesignA looking-while-listening paradigm was used to measure novel word learning in 16 children with sensorineural HI (age range 23.2 to 42.1 months) who wore either bilateral hearing aids (n = 10) or bilateral cochlear implants (n = 6) and 16 children with NH (age range 23.1 to 42.1 months) who were matched for gender, chronological age, and maternal education level. Two measures of word learning were assessed (accuracy and reaction time). Each child participated in two experiments approximately 1 week apart, one in IDS and one in ADS.ResultsBoth groups successfully learned the novel words in both speech type conditions, as evidenced by children looking at the correct picture significantly above chance. As a group, children with NH outperformed children with HI in the novel word learning task; however, there were no significant differences between performance on IDS versus ADS. More fine-grained time course analyses revealed that children with HI, and particularly children who use hearing aids, had more difficulty learning novel words in ADS, compared with children with NH.ConclusionsThe pattern of results observed in the children with HI suggests that they may need extended support from clinicians and caregivers, through the use of IDS, during novel word learning. Future research should continue to focus on understanding the factors (e.g., device type and use, age of intervention, audibility, acoustic characteristics of input, etc.) that may influence word learning in children with HI in both IDS and ADS.

Dataset Information

Unsupervised learning of vowel categories from infant-directed speech.

Publications

Unsupervised learning of vowel categories from infant-directed speech.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets