Correction of Errors in a Modality Corpus Used for Machine Translation Using Machine-learning

We performed corpus correction on an annotated corpus for machine translation using machine-learning methods such as the maximum-entropy method. We thus constructed a high-quality annotated corpus based on corpus correction. We compared several di erent methods of corpus correction in our experiments and developed a suitable method for correction. Recently, corpus-based machine translation has been investigated. Since corpus-based machine translation uses corpora, the corpus correction we discuss in this paper should prove to be signi cant.