A Comparative Study of Occurrence of Errors in Machine Translation in a Multilingual Environment

Abstract — Machine translation is an inevitable field of Natural Language Processing, which includes two steps. The first step literally follows the reference method of machine translation, taking advantage of corpus of knowledge already available in the system. The second step involves the post-editing done by human intervention. This paper analyzes the errors, categorizes them, and gives a comparative study of the errors in the first and second steps. Further, this analysis is done in a multilingual environment. Every step is accompanied by the machine training on the corpus of annotations and systematic classifiers