As a preliminary study for developing an automatic English composition evaluation system, this study presented the results of an analysis of the establishment of a Korean college student English composition error corpus and the correlation between English composition error types and errors. The results of this study used a relatively small amount of English composition data, and the distribution of analysis subjects was not balanced by English composition score, so it is considered somewhat unreasonable to interpret it as a Meaningful result. To establish a quality set for the automatic English composition evaluator in the future, it is necessary to analyze errors in more balanced English composition and research how to extract qualities highly related to error types and English composition scores.