【摘要】 中介语语音库的构建面临一个文本设计问题:既要涵盖目的语的基本语音要素,又要严控其大小。因为文本过大,单人收录成本过高,不利于收集更多的话者被试。本文介绍我们面向汉语中介语语音库构建所提出的一套文本方案:在控制文本集大小的前提下,最大限度地涵盖更多的语音要素。除了单音节、音韵平衡双音节文本外,该方案还包含一个覆盖音段、声调、调联三音子、焦点语调且文本难度适中的最小句子集合,该句子集合由计算机算法从大语料库中搜索产生。我们认为基于这个录音文本,不仅能够收集到二语习得研究所关注的各种语音现象,还能够比较方便地获得大量的话者数量,从而更好地服务于语音习得及计算机辅助语音教学研究。更多还原

【Abstract】 Text design is challenging in the work of building a non-native speech corpus,as it is desired to cover as many as possible the essential phonetic elements in a small text size.This paper proposes a text design which covers rich phonetic elements of Chinese Putonghua with a small text size for the purpose of developing non-native speech corpora.The designed text consists of monosyllables,phonetically balanced disyllables,a minimum sentence set,a short passage and a few other sentences.The minimum sentence set was selected by agreedy search algorithm from texts of primary to medium levels of Chinese textbooks,and it has a rich coverage of Chinese segments,lexical tones,tri-tones,prosody phrases and different intonation patterns.The text offers a possibility to collect various kinds of speech phenomena,and simultaneously guarantee the easiness of collecting more speakers’ data.These two advantages are beneficial to studies of Second Language Acquisition and Computer-assisted Speech Teaching technology.更多还原