| Presentation | 2014-05-30 Phonological learning model based on phoneme transition statistics Masashi GUNJI, Naokazu GODA, Miyuki G. KAMACHI, |
|---|---|
| PDF Download Page |  PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | Developmental studies have shown that infants are sensitive to statistical regularities in phoneme sequence and use them to learn their mother language. Little is known, however, about what statistical features are specifically important in the language acquisition. Here we explored how structures of syllables, words, or bunsetsu can be acquired based on statistical regularities in the phoneme sequence, by using a model that mimics statistical learning in infancy. Based on so-called N-gram processing, this model learns transition statistics from sequences of phonetic alphabets representing natural language and uses them to segment the sequences into syllables, words, or bunsetsu. We found that, in Japanese language, the transitional regularities between the short arrays (N=1,2) of phonetic alphabets were quite useful for segmentation of syllables, whereas those between the longer arrays (N=2,3) were informative for segmentation of words and bunsetsu. Our results suggest that transitional regularities between phonemes in different scales would provide important cues for learning structure of syllables and words/bunsetsu in infancy. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | language acquisition / phonetic alphabet / transition probability / conditional entropy |
| Paper # | HCS2014-30,HIP2014-30 |
| Date of Issue |
| Conference Information | |
| Committee | HCS |
|---|---|
| Conference Date | 2014/5/22(1days) |
| Place (in Japanese) | (See Japanese page) |
| Place (in English) | |
| Topics (in Japanese) | (See Japanese page) |
| Topics (in English) | |
| Chair | |
| Vice Chair | |
| Secretary | |
| Assistant | |
| Paper Information | |
| Registration To | Human Communication Science (HCS) |
|---|---|
| Language | JPN |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | Phonological learning model based on phoneme transition statistics |
| Sub Title (in English) | |
| Keyword(1) | language acquisition |
| Keyword(2) | phonetic alphabet |
| Keyword(3) | transition probability |
| Keyword(4) | conditional entropy |
| 1st Author's Name | Masashi GUNJI |
| 1st Author's Affiliation | Kogakuin University, Graduate School of Engineering() |
| 2nd Author's Name | Naokazu GODA |
| 2nd Author's Affiliation | National Institute for Physiological Sciences |
| 3rd Author's Name | Miyuki G. KAMACHI |
| 3rd Author's Affiliation | Kogakuin University, Faculty of Information Studies |
| Date | 2014-05-30 |
| Paper # | HCS2014-30,HIP2014-30 |
| Volume (vol) | vol.114 |
| Number (no) | 67 |
| Page | pp.pp.- |
| #Pages | 4 |
| Date of Issue | |