學齡前兒童中文語彙毗鄰測驗的編製與驗證

Mandarin Lexical Neighborhood Test (M-LNT) for Pre-school Children: Development of Test and Its Validation

楊惠美(Hui-Mei Yang) ； 吳俊良(Jiunn-Liang Wu)

《台灣耳鼻喉頭頸外科雜誌》 40卷1期 (2005/01) Pp. 1-12

https://doi.org/10.6286/2005.40.1.1

摘要

BACKGROUND: Open-set spoken word recognition test is essential for monitoring the benefits of cochlear implants on the implanted children. For developing the open-set spoken word recognition test, two stages of experiment were included in this study. First, we set up the Mandarin language sampling database in order to construct the open-set Mandarin Lexical Neighborhood Test (M-LNT). Second, the testing materials were used to evaluate the speech discrimination of children for the validation of M-LNT.
METHODS: The Mandarin language sampling database, comprising 80 transcripts of preschool children via audio-recordings, was used to generate the test items. Word lists were constructed to calculate word frequencies, lexical neighborhood of the target word, and to identify the median of word frequency and lexical density. According to the Neighborhood Activation Model (NAM), words were categorized by their lexical properties into ”easy” and ”hard” lists, which comprised the M-LNT. Twenty-eight cochlear implanted children, who used the devices for at least one year and demonstrated some open-set spoken word recognition, were enrolled. Word lists of M-LNT were administered to the participants via recorded CD in the auditory-only modality. The participants responded by verbally repeating the words they heard. Their responses were scored as the percent of words and phonemes correctly identified.
RESULTS: The database contained 45,197 utterances and 183,053 word-tokens. The mean word of utterances was 4.062. The word frequency ranged from 1 to 6l30 and the median was 46. Lexical neighborhood density ranged from 0 to 27 and the median was 11. There were 207 and 193 words identified for the lexical ”easy” and ”hard” categories respectively. They were classified into four ”easy” and four ”hard” word lists. Each list consisted of 25 test items. For the LNT easy lists, the average scores of 28 children correctly identified word, 76.82% and phoneme, 88.64%. For the hard lists, the scores were 68.71% and 85.75%. The results of repeated measure analysis of variance revealed that the percent of words and phoneme correctly identified was significantly higher for the easy lists than for the hard lists (F=46.579, p＜.000). The inter-list correlation among lists was statistically significant (p＜.01) and the repeated measure analysis of variance revealed that the performance of speech discrimination of children was not different among lists (p＞.05).
CONCLUSION: Four pairs of lexical easy and hard word lists, theoretically driven by the NAM, were created for preschool children. The new M-LNT is expected to be useful for evaluating spoken word recognition in children with hearing loss. Also, the M-LNT provided reliable information about spoken word recognition abilities of children with cochlear implants. These Mandarin speaking hearing-impaired children with cochlear implants are sensitive to the acoustic-phonetic similarities among words. Thus, they organize words into similarity neighborhoods in long-term memory.

關鍵字

人工耳蝸 ； 口語字詞辨識 ； 學齡前兒童

並列摘要

BACKGROUND: Open-set spoken word recognition test is essential for monitoring the benefits of cochlear implants on the implanted children. For developing the open-set spoken word recognition test, two stages of experiment were included in this study. First, we set up the Mandarin language sampling database in order to construct the open-set Mandarin Lexical Neighborhood Test (M-LNT). Second, the testing materials were used to evaluate the speech discrimination of children for the validation of M-LNT.
METHODS: The Mandarin language sampling database, comprising 80 transcripts of preschool children via audio-recordings, was used to generate the test items. Word lists were constructed to calculate word frequencies, lexical neighborhood of the target word, and to identify the median of word frequency and lexical density. According to the Neighborhood Activation Model (NAM), words were categorized by their lexical properties into ”easy” and ”hard” lists, which comprised the M-LNT. Twenty-eight cochlear implanted children, who used the devices for at least one year and demonstrated some open-set spoken word recognition, were enrolled. Word lists of M-LNT were administered to the participants via recorded CD in the auditory-only modality. The participants responded by verbally repeating the words they heard. Their responses were scored as the percent of words and phonemes correctly identified.
RESULTS: The database contained 45,197 utterances and 183,053 word-tokens. The mean word of utterances was 4.062. The word frequency ranged from 1 to 6l30 and the median was 46. Lexical neighborhood density ranged from 0 to 27 and the median was 11. There were 207 and 193 words identified for the lexical ”easy” and ”hard” categories respectively. They were classified into four ”easy” and four ”hard” word lists. Each list consisted of 25 test items. For the LNT easy lists, the average scores of 28 children correctly identified word, 76.82% and phoneme, 88.64%. For the hard lists, the scores were 68.71% and 85.75%. The results of repeated measure analysis of variance revealed that the percent of words and phoneme correctly identified was significantly higher for the easy lists than for the hard lists (F=46.579, p＜.000). The inter-list correlation among lists was statistically significant (p＜.01) and the repeated measure analysis of variance revealed that the performance of speech discrimination of children was not different among lists (p＞.05).
CONCLUSION: Four pairs of lexical easy and hard word lists, theoretically driven by the NAM, were created for preschool children. The new M-LNT is expected to be useful for evaluating spoken word recognition in children with hearing loss. Also, the M-LNT provided reliable information about spoken word recognition abilities of children with cochlear implants. These Mandarin speaking hearing-impaired children with cochlear implants are sensitive to the acoustic-phonetic similarities among words. Thus, they organize words into similarity neighborhoods in long-term memory.

並列關鍵字

cochlear implant ； spoken word recognition ； pre-school children