한국어 일반 개체명과 개인정보 특화 개체명 비교 연구

초록

As information technology and artificial intelligence (AI) methods develop, apprehensions over the leaking of personal information during extensive data processing are increasing. To differentiate Personally Identifiable Information (PII) throughout extensive datasets, it is essential to precisely identify and categorize PII, distinct from General Named Entities (GNE). This study examines the prevalence of both categories in authentic interactions by creating a conversation-based dataset annotated with GNE and PII. The analysis indicates that contextual information is crucial for recognizing things associated with 'place,' 'organization,' and 'academic field.' Furthermore, constraints in categorizing things such as 'date' and 'culture' underline the necessity for continual enhancements of currently functioning systems. It also emphasizes the importance for proactive innovations in personal information detection technologies.

키워드