In this study, we propose a system for detecting subject ellipsis. Subject ellipsis is known to make sentence comprehension difficult for non-native speakers. This system first extracts a noun which may be the subject and the predicate from an input sentence by using a dependency parser. Next, the system makes two types of sentences with the extracted noun and predicate: one with the particle ga and the other with the particle wo. Then it compares the frequency of each sentence in Web texts using a search engine to judge whether the extracted noun is the subject of the predicate. Based on the results, it can be determined whether the noun is the subject or the subject has been deleted.