Newspapers ordinarily reflect the meaning of lexical items in two Korean language societies and the changes in vocabulary in the times. This study proposes a methodology that automatically probes for semantic variations in which inter-Korean vocabulary differs from large-scale newspaper data. As a theoretical background, we look at the concepts of the distributional semantics and semantic variations. Next, using deep learning’s word embedding skills, we implement a system to probe semantic variations in an automatic way. This study is significant in the following respects. First, this study systematically concerns the difference in the meaning of inter-Korean vocabulary on a comprehensive scale. Second, this study draws a list of inter-Korean semantic variations by means of the deep learning techniques. Third, this study demonstrates that use of word embedding models facilitate automatic extraction of Korean semantic variations.