Investigating Code-Mixed Modern Standard Arabic-Egyptian to English Machine Translation

TL;DR

This paper explores neural machine translation for code-mixed Modern Standard Arabic and Egyptian Arabic to English, demonstrating effective models and achieving top results in a shared task.

Contribution

It introduces models for translating code-mixed Arabic dialects to English, utilizing both training from scratch and pre-trained language models, and achieves state-of-the-art performance.

Findings

Pre-trained language models improve translation quality.

Reasonable performance achieved with limited parallel data.

First place in the shared task evaluation.

Abstract

Recent progress in neural machine translation (NMT) has made it possible to translate successfully between monolingual language pairs where large parallel data exist, with pre-trained models improving performance even further. Although there exists work on translating in code-mixed settings (where one of the pairs includes text from two or more languages), it is still unclear what recent success in NMT and language modeling exactly means for translating code-mixed text. We investigate one such context, namely MT from code-mixed Modern Standard Arabic and Egyptian Arabic (MSAEA) into English. We develop models under different conditions, employing both (i) standard end-to-end sequence-to-sequence (S2S) Transformers trained from scratch and (ii) pre-trained S2S language models (LMs). We are able to acquire reasonable performance using only MSA-EN parallel data with S2S models trained from scratch. We also find LMs fine-tuned on data from various Arabic dialects to help the MSAEA-EN task. Our work is in the context of the Shared Task on Machine Translation in Code-Switching. Our best model achieves $\bf25.72$ BLEU, placing us first on the official shared task evaluation for MSAEA-EN.