Mining Semantic Tags in a Content Analysis System for a Letter Database of Ethnic Koreans Living in China

摘要

In this paper, we present a content analysis system for the letter database for ethnic Koreans living in China using several machine learning techniques. First, letters from ethnic Koreans living in China to Korea Broadcasting System (KBS) to find separated families were digitized and constructed as a letter database. Second, to help retrieval of digitized letters, semantic tags were annotated to each letter manually. Third, to find the contents of the digitized letters, machine learning techniques were applied to those semantic tags. Tag-based topic modeling was performed and a tag network was constructed. Over 150,000 letters were scanned as image files and have been constructed as a letter database. Among them, 3,500 letters are randomly selected and analyzed. Our approach shows that the semantic tags play an important role in the digitized letter retrieval. We are currently enhancing algorithms for tag analysis, and the developed system will broaden the horizon of digital humanities.