German Voice Training with a Speech Recognition Application
オンセイ ニンシキ アプリ ヲ カツヨウ シタ ドイツゴ ハツオン トレーニング
岩居, 弘樹 ,
Iwai, Hirokiイワイ, ヒロキ
15 , 2015-03-31 , 大阪大学全学教育推進機構 , オオサカ ダイガク ゼンガク キョウイク スイシン キコウ
This paper shows the potential of pronunciation practice using a voice recognition system in German class for beginners. The study involves the analysis of the output texts produced by 96 (Summer Semester) / 93 (Winter Semester) students who practiced reading 5 to 10 German sentences into Speech to Text (STT) apps such as “Dragon Dictation” for about 30 minutes of each lesson. Analysis of the output from STT revealed a number of findings. The students tried to have their voice recognized for an average of 77 to 87 times within 30 minutes. This means that they read more than 300 words aloud while checking and modifying the position of their mouth and tongue carefully. There is some evidence in the output texts of the exercise that shows that some of the students corrected their pronunciation successfully without help from their instructor. This exercise also shows that there are some sentences Japanese students correctly pronounce without any training. On other hand, it clearly showed that rounded vowels such as “u” are one of the weakest points of Japanese students. From the patterns of failures the students made in the exercise, the instructor can recognize what part of the articulatory organ students should pay attention to, and can give them advice on how they should change its form. The results of the exercise suggest that “speak-aloud-practice” using an STT app gives learners more motivation and concentration than traditional pronunciation exercises that don’t give students any realtime feedback on the sounds they produce.