Can You Traducir This? Machine Translation for Code-Switched Input

TL;DR

This paper addresses the challenge of translating code-switched texts by generating artificial training data, resulting in machine translation systems that outperform multilingual models on such texts and aiding L2 writing assistance.

Contribution

It introduces a novel training strategy using artificially generated data to improve machine translation of code-switched language, a previously underexplored area.

Findings

Artificial training data improves MT performance on CSW texts

Proposed method surpasses existing multilingual systems for code-switched translation

Effective in assisting L2 writing with contextual translations

Abstract

Code-Switching (CSW) is a common phenomenon that occurs in multilingual geographic or social contexts, which raises challenging problems for natural language processing tools. We focus here on Machine Translation (MT) of CSW texts, where we aim to simultaneously disentangle and translate the two mixed languages. Due to the lack of actual translated CSW data, we generate artificial training data from regular parallel texts. Experiments show this training strategy yields MT systems that surpass multilingual systems for code-switched texts. These results are confirmed in an alternative task aimed at providing contextual translations for a L2 writing assistant.