This paper analyzed the error patterns that appear when machine translators translate Hindi.
With the development of computer technology, methods for converting and analyzing language into data are required in various fields. One of the things developed to meet this demand is Google Translate. Google Translate is an artificial intelligence translator. The artificial intelligence translator derives more accurate results through a self-learning process. Therefore, machine translation can help you understand the approximate contents in a short time. However, since machine translation is not yet perfect, it shows many errors. This study aims to find out what kind of errors occur along with the characteristics of these machine translators. There are several artificial intelligence translators, but Google Translator was chosen as the target because it seemed to show more accurate results than other translators. As for the text to be translated, Hindi novels were chosen because the translation of a grammar book will be too good. Among them, the study was centered on Premchand's short stories, because Premchand's novels were the stories of the poor and neglected, so it was expected that there would be similarities in style and words used in each novel.
Each sentence was divided into 1-5 grades according to the translation level, and the framework for sentence grade analysis was used as a scale. Besides, the types of errors were analyzed by dividing them into eight categories.
As a result of analyzing the three novels, more than half of them were translated well. In terms of errors, lexical errors appeared the most. Especially in the case of polysemy, there were many cases in which the vocabulary was not selected correctly.
Machine translation has not yet developed to the level of replacing humans. However, if machine translation is used well, it is possible to improve the efficiency of Hindi learners.