한국어 학습자 말뭉치의 개인 정보 처리 분석 - 국립국어원 한국어 학습자 말뭉치를 중심으로 -

This study analyzed the pattern of personal information exposure in the corpus of Korean learners. The current status of personal information exposure of the corpus of Korean learners was identified, and the transcription direction for the corpus to be built was presented. This study is noteworthy because it dealt with the protection of learners’ human rights and privacy, which was not covered in previous studies. First, this study examined the function and role of the learner corpus. In order to efficiently utilize the learner corpus, basic information regarding the learners is required. This information should be provided to the extent agreed by the learner, and the anonymity of the corpus provider should be guaranteed. However, it is not easy to process large-scale personal information data on a consistent basis. Second, how the Korean language learner corpus processes the learners’ personal information was confirmed as actual corpus data. It was found that the extent of personal information to be processed as “<name>” should be clarified in the transcription guidelines, and it is also necessary to subdivide the form according to the type of name. There are two types of personal information exposure in the corpus of Korean learners. The first is an example of failure to follow transcription guidelines. The personal information presented includes learner names, peer names, learning institutions, phone numbers, and e-mail IDs. The second is the exposure of personal information, caused by the combination of information in the corpus content even though there is no violation of transcription guidelines. Finally, the direction for the personal information processing of the Korean language learner corpus to be built was divided into three categories. First, there should be sufficient education to prevent personal information exposure caused by violations of transcription guidelines. Second, the scope of items to be processed for personal information should be explicitly defined. Third, since personal identification is possible by combining information in the text, personal information should be processed while considering the entire content.