Non-native speech corpora for STiLL at SiTEC

Abstract

In this paper we introduce non-native English and Korean speech corpora that SITEC (Speech Information Technology & Industry Promotion Center) has created for research into speech technology in language learning. Non-native speakers' English speech corpora are divided largely into three types of Korean, Chinese and Japanese speakers' English. Koreans' English speech corpora are composed of three corpus: K-AESOP(200 speakers), K-SEC (342 speakers) and Koreans' English speech corpus for fluency evaluation(200 speakers). Chinese people's English speech corpus and Japanese people's English speech corpus comprise 200 speakers individually. Non-native speakers' Korean speech corpora are two types as F-Korean01(180 speakers) and non-native speakers' Korean speech corpus(200 speakers).