Year: 2026
Type: conference-paper
Abstract: Judeo-Arabic refers to Arabic variants historically spoken by Jewish communities across the Arab world, primarily during the Middle Ages.Unlike standard Arabic, it is written in Hebrew script by Jewish writers and for Jewish audiences.Transliterating Judeo-Arabic into Arabic script is challenging due to ambiguous letter mappings, inconsistent orthographic conventions, and frequent code-switching into Hebrew.In this paper, we introduce a two-step approach to automatically transliterate Judeo-Arabic into Arabic script: simple characterlevel mapping followed by post-correction to address grammatical and orthographic errors.We also present the first benchmark evaluation of LLMs on this task.Finally, we show that transliteration enables Arabic NLP tools to perform morphosyntactic tagging and machine translation, which would have not been feasible on the original texts.We make our code and data publicly available.1 Mitelman et al. (2024) 58.5 14.8 23.6 36.7 70.0 58.4 14.7 23.5 36.7 70.1 ⇒ Seq2Seq 75.0 26.6 39.3 55.0 82.9 74.9 26.7 39.3 55.0 82.8 ⇒ Seq2Seq+MG 75.6 26.9 39.7 55.5 82.9 75.5 27.4 40.2 55.9 82.9 ⇒ SWEET 76.9 23.5 36.0 52.9 83.9 77.0 23.5 36.0 52.9 84.1 ⇒ GPT-3.5-turbo74.0 20.8 32.5 49.0 78.1 74.1 21.6 33.4 49.8 79.3 ⇒ GPT-4o 77.9 23.8 36.4 53.5 86.1 77.7 23.5 36.1 53.2 87.4 ⇒ Fanar 66.4 9.7 17.0 30.7 61.0 66.8 11.5 19.6 34.0 61.7 ⇒ Jais-13B-Chat 54.0 3.8 7.0 14.7 24.9 53.6 4.2 7.8 16.0 24.7 CharMapper 100 0 0 0 53.0 75.5 14.1 23.8 40.4 64.9 ⇒ Seq2Seq 72.8 27.2 39.6 54.5 74.5 81.2 30.3 44.1 60.8 85.2 ⇒ Seq2Seq+MG 73.4 29.0 41.6 56.2 76.3 82.5 30.8 44.9 61.8 86.9 ⇒ SWEET 71.8 28.5 40.8 55.1 76.5 80.1 32.9 46.7 62.3 86.0 ⇒ GPT-3.5-turbo72.1 23.9 35.9 51.4 75.6 78.7 27.5 40.8 57.4 82.6 ⇒ GPT-4o 82.0 28.4 42.2 59.5 86.4 85.9 30.7 45.2 63.1 90.4 ⇒ Fanar 65.5 13.9 23.0 37.7 67.6 71.9 18.9 29.9 46.0 74.0 ⇒ Jais-13B-Chat 48.0 3.4 6.4 13.3 31.4 53.2 4.2 7.8 16.0 31.1 Table 4: Post-correction results for various systems on transliterated Arabic text derived from either dotless or dotted Judeo-Arabic Hebrew script.
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7140136559