上声变调在语音识别中的实证研究  

An Empirical Study on Tone Change of Third Tone Sandhi in Speech Recognition

作　　者：杨静[1] Yang Jing(School of Literature,Central China of Normal University,Wuhan 430079 Hubei)

机构地区：[1]华中师范大学文学院,湖北武汉430079

出　　处：《科学与信息化》2021年第15期30-31,共2页Technology and Information

基　　金：2020年华中师范大学大学生创新创业训练项目,项目编号:CCNU202010511186。

摘　　要：旨在测试语音识别软件对上声变调的识别情况,分析识别错误类型、成因,提出对策。选取《现代汉语词典》(第七版)中两个上声相连与三个上声音节相连的词,让一名普通话等级“一乙”者朗读,获取科大讯飞输入法识别结果与录音,并利用Praat进行语图分析。分析得知影响识别正确率的因素有语流音变、音节形式、使用频率与范围,并就此提出相关对策。This paper wants to test the recognition state about tone change of third tone sandhi by speech recognition application,analyze the types and causes of recognition errors,and propose measures.Select some disyllabic and trisyllabic words all consisting of third tone sandhi in Modern Chinese Dictionary(7th edition),let a person with Mandarin level one grade B read aloud to get the recognition result of iFlytek Voice Input Method and record,use Praat to analyse.The analysis lead to a conclusion that the factors affecting the accuracy of recognition are tone change,form of syllable,frequency and range of use,and related measures are proposed.

关 键 词：上声 语流音变 语音识别 实证研究 

分 类 号：H17[语言文字—汉语]