구글 음성인식기를 활용한 한국 초등학생들의 영어 단어 발음인식감도 조사

Analysis of Google Voice Actions’ Recognition of Korean Young Learners’ English Word Pronunciation

The present study aims to investigate how well Google Voice Actions, an automatic speech recognition system, recognizes Korean young English learners’ pronunciation of English words. To achieve this aim, the current study arranged for 18 Korean elementary school students to pronounce 219 English words and recorded their pronunciation. Then, the intelligibility of their pronunciation was measured using Google Voice Actions. The current study analyzed the measured intelligibility of Google Voice Actions in terms of the phonemic difficulty and familiarity the learners have with the words. The phonemic difficulty of each word was labelled as Group 1 to 5, depending on the number of difficult phonemic elements it contains. The familiarity of each word was also measured through a questionnaire. The findings revealed that the accuracy in Google Voice Actions’ recognition was closely related to the students’the phonemic difficulties and familiarity of the words. Interestingly, the pronunciation of words in Group 5 gained the highest recognition scores and that of Group 1 the second highest scores. Those of Group 2, 3, and 4 recorded lower recognition scores than the two groups. These results suggest that the more phonemic information the pronunciation of a word provides, the more relevant clues would be available for Google Voice Actions. This increases the possibility of successful search from the speech database.