Translating Between Morphologically Rich Languages: An Arabic-to-Turkish Machine Translation System

Abstract

This paper introduces the work on building a machine translation system for Arabic-to-Turkish in the news domain. Our work includes collecting parallel datasets in several ways for a new and low-resource language pair, building baseline systems with state-ofthe-art architectures and developing language specific algorithms for better translation. Parallel datasets are mainly collected three different ways; i) translating Arabic texts into Turkish by professional translators, ii) exploiting the web for open-source Arabic-Turkish parallel texts, iii) using back-translation. We performed preliminary experiments for Arabicto-Turkish machine translation with neural (Marian) machine translation tools with a novel morphologically motivated vocabulary reduction method.

Resources for Turkish Natural Language Processing: A critical survey

“…All numbers are based on the corpora as available from OPUS parallel corpora collection http://opus.nlpl.eu/. on MT between Turkish and mainly English (e.g., Durgar El-Kahlout and Oflazer, 2010;Oflazer et al, 2018;Durgar El-Kahlout et al, 2019). Similarly, shared tasks which included Turkish as one of the languages, such as two IWLST shared tasks (Paul et al, 2010;Cettolo et al, 2013), and WMT shared tasks between 2016 and 2018 (Bojar et al, 2016), also provided data for use during the shared tasks.…”