Technical development of environments for learning and teaching foreign languages by using structural representation of pronunciation

为了提高外语发音分析的性能，将语音结构表征引入到语音评分、语音错误检测和语音聚类任务中。结构表示法允许我们只根据音对音的关系来表示发音。利用传统的关系特征和绝对声音特征，提高了评分和错误检测的性能。学生聚类具有较高的效度。

To improve the performance of analyzing foreign language pro-nunciations, structural representation of speech was introduced to the task of scoring the pronunciation, detecting pronunciation errors, and clustering students' pronunciations. The structural representation allows us only to represent the pronunciation only based on sound-to-sound relations. By using the relations as well as absolute sound features, which have been conventionally used, the performance of scoring and error detection was im-proved. High validity of students' clustering was also shown.