Machine Translation for Historical Research: A Case Study of Aramaic-Ancient Hebrew Translations

In this article, by the ability to translate Aramaic to another spoken languages, we investigated machine translation in a cultural heritage domain for two primary purposes: evaluating the quality of ancient translations and preserving Aramaic (an endangered language). First, we detailed the construction of a publicly available Biblical parallel Aramaic-Hebrew corpus based on two ancient (early 2 nd to late 4 th century) Hebrew-Aramaic translations: Targum Onkelus and Targum Jonathan. Then using the statistical machine translation approach, which in our use case significantly outperforms neural machine translation, we validated the excepted high quality of the translations. The trained model failed to translate Aramaic texts of other dialects. However, when we trained the same statistical machine translation model on another Aramaic-Hebrew corpus of a different dialect (Zohar, 13 th century), a very high translation score was achieved. We examined an additional important cultural heritage source of Aramaic texts, the Babylonian Talmud (early 3 rd to late 5 th century). Since we do not have a parallel Aramaic-Hebrew corpus of the Talmud, we used the model trained on the Bible corpus for translation. We performed an analysis of the results and suggest some potential promising future research.