Machine translations(MT) such as Google Translate and Naver Papago are widely used in Korea as translation tools. Although both translators apply the data-based Neural based MT (NMT) method, the translation results are not yet satisfactory compared to skilled human translators. In this paper, the Spanish-Korean translation errors made by these machines translators were analyzed qualitatively based on the Skopos theory, which is a functionalist translation theory. Some macroscopic errors are observed in word recognition, punctuation, spacing, phrase, sentence recognition, and sentential ending style. Microscopic errors in lexical, syntactic, and pragmatic aspects were analyzed by applying eight oblique translation procedures. The only procedure applied well in both translators is transposition and the majority of oblique procedures such as modulation, adaptation, amplification, explicitation, omission, and compensation are not properly applied. We expect that this study will be used as basic data for reference when applying a machine translator in the educational field or when a developer wants to improve the performance of both translators.