This paper proposes a hierarchical word domain assignment algorithm to automatically build domain dictionaries from Machine-Readable Dictionary(MRD).The process for word domain assignment can be divided into three ste...This paper proposes a hierarchical word domain assignment algorithm to automatically build domain dictionaries from Machine-Readable Dictionary(MRD).The process for word domain assignment can be divided into three steps:1) Hierarchical structure constructing;2) Classifier training;3) Word domain assigning.Compared with the traditional methods,the hierarchical word domain assignment algorithm enhances the accuracy of word domain assignment while reducing human efforts on collecting corpus.Experiments on WordNet 2.0 show that 62.53% of the first domain labels are matched with the WordNet Domains 3.0 by using gloss-based word domain assignment,and the performance can be further improved by utilizing the hierarchical relationships among the domain sets.展开更多
Integrating letters and sounds are essential for successful reading in alphabetic languages. It remains unclear if native speakers of non-alphabetic languages integrate letters and sounds in reading an alphabetic lan-...Integrating letters and sounds are essential for successful reading in alphabetic languages. It remains unclear if native speakers of non-alphabetic languages integrate letters and sounds in reading an alphabetic lan- guage in the same way as native alphabetic readers do. Chinese is a morpho-syllabic system （each character cor- responds to one syllable） and contrasts sharply with alphabetic languages such as English. Several fMRI studies have shown that native Chinese speakers apply their native language system to read English words. By using the cross- modal mismatch negativity （MMN） paradigm, we directly investigated letter-sound integration for reading in English among native Chinese speakers. To investigate the effect of native language background on letter-sound integration in second language reading, a group of native Korean English learners served as a comparison group. We compared MMN responses between an auditory only condition （only vowels presented） and two audiovisual conditions （AV0, vowel presented synchronously with the corresponding letter; AV200, the letter presented 200 ms before the corresponding vowel） for both native Chinese and native Korean speakers. Native Chinese speakers demonstrated significantly attenuated MMN amplitudes in audiovisual conditions compared with the auditory only condition, regardless of their phonological decoding speed. In con- trast, native Korean speakers showed amplified amplitude MMN in AV200 compared with that in the auditory only condition. The results suggest that native language may shape the brain responses of second language learners to reading a second language in the early stages. Native non- alphabetic language speakers may be unable to use visual information to facilitate their phonological processing in the early stage while naT：lye alphabetic language speakers are capable of integrating letter sounds automatically.展开更多
基金supported by the BIT Technology Innovation Program "cloud computing-oriented intelligent processing theory and method of massive language information"underGrant No.3070012231102the BIT Fundamental Research Projects under Grant No.3070012210917
文摘This paper proposes a hierarchical word domain assignment algorithm to automatically build domain dictionaries from Machine-Readable Dictionary(MRD).The process for word domain assignment can be divided into three steps:1) Hierarchical structure constructing;2) Classifier training;3) Word domain assigning.Compared with the traditional methods,the hierarchical word domain assignment algorithm enhances the accuracy of word domain assignment while reducing human efforts on collecting corpus.Experiments on WordNet 2.0 show that 62.53% of the first domain labels are matched with the WordNet Domains 3.0 by using gloss-based word domain assignment,and the performance can be further improved by utilizing the hierarchical relationships among the domain sets.
基金supported by the National Natural Science Foundation of China(31221003)National Basic Research Program of China(2014CB846103)
文摘Integrating letters and sounds are essential for successful reading in alphabetic languages. It remains unclear if native speakers of non-alphabetic languages integrate letters and sounds in reading an alphabetic lan- guage in the same way as native alphabetic readers do. Chinese is a morpho-syllabic system （each character cor- responds to one syllable） and contrasts sharply with alphabetic languages such as English. Several fMRI studies have shown that native Chinese speakers apply their native language system to read English words. By using the cross- modal mismatch negativity （MMN） paradigm, we directly investigated letter-sound integration for reading in English among native Chinese speakers. To investigate the effect of native language background on letter-sound integration in second language reading, a group of native Korean English learners served as a comparison group. We compared MMN responses between an auditory only condition （only vowels presented） and two audiovisual conditions （AV0, vowel presented synchronously with the corresponding letter; AV200, the letter presented 200 ms before the corresponding vowel） for both native Chinese and native Korean speakers. Native Chinese speakers demonstrated significantly attenuated MMN amplitudes in audiovisual conditions compared with the auditory only condition, regardless of their phonological decoding speed. In con- trast, native Korean speakers showed amplified amplitude MMN in AV200 compared with that in the auditory only condition. The results suggest that native language may shape the brain responses of second language learners to reading a second language in the early stages. Native non- alphabetic language speakers may be unable to use visual information to facilitate their phonological processing in the early stage while naT：lye alphabetic language speakers are capable of integrating letter sounds automatically.
