Impact of Corpora Quality on Neural Machine Translation

TL;DR

This paper examines how corrupted data in large web-sourced corpora impacts neural machine translation quality and proposes scripts to identify and remove problematic sentences to improve system performance.

Contribution

It introduces practical methods and scripts for detecting and filtering corrupted data in corpora used for neural machine translation.

Findings

Corrupted data negatively affects translation quality.

Scripts can effectively identify problematic sentences.

Filtering improves translation system performance.

Abstract

Large parallel corpora that are automatically obtained from the web, documents or elsewhere often exhibit many corrupted parts that are bound to negatively affect the quality of the systems and models that learn from these corpora. This paper describes frequent problems found in data and such data affects neural machine translation systems, as well as how to identify and deal with them. The solutions are summarised in a set of scripts that remove problematic sentences from input corpora.