Part-of-Speech and Morphological Tagging of Algerian Judeo-Arabic

Part-of-Speech and Morphological Tagging of Algerian Judeo-Arabic Ofra Tirosh-Becker1∗ Michal Kessler2* Oren M. Becker3 Yonatan Belinkov4† 1 The Hebrew University of Jerusalem, Israel. otirosh@mail.huji.ac.il 2 The Hebrew University of Jerusalem, Israel. michalskessler@gmail.com 3 Becker Consulting Ltd., Mevasseret Zion, Israel. becker.oren@gmail.com 4 Technion – Israel Institute of Technology, Israel. belinkov@technion.ac.il Abstract Most linguistic studies of Judeo-Arabic, the ensemble of dialects spoken and written by Jews in Arab lands, are qualitative in nature and rely on laborious manual annotation work, and are therefore limited in scale. In this work, we develop automatic methods for morpho-syntactic tagging of Algerian Judeo-Arabic texts published by Algerian Jews in the 19th–20th centuries, based on a linguistically tagged corpus. First, we describe our semi-automatic approach for preprocessing these texts. Then, we experiment with both an off-the-shelf morphological tagger, several specially designed neural network taggers, and a hybrid human-in-the-loop approach. Finally, we perform a real-world evaluation of new texts that were never tagged before in comparison with human expert annotators. Our experimental results demonstrate that these methods can dramatically speed up and improve the linguistic research pipeline, enabling linguists to study these dialects on a much greater scale. 1 Introduction Application of Natural Language Processing (NLP) to real-world problems has been the field’s goal from its early days. As algorithms advance, the contribu- tion of NLP to real problems has become more evi- dent and more substantial. The present study origi- nates from a real-world challenge faced by linguists of Semitic languages, in this case researchers of the Judeo- Arabic dialects of Algeria (AJA). Their challenge, sim- ply put, is how to scale up linguistic analyses of such dialects. Semitic languages in general, and Arabic in particular, are characterized by a very rich morphol- ogy that uses both templatic and concatenative mor- phemes, combined with the use of a vowelless script (“abjad”). This makes morphological analysis of Arabic very time-consuming even for expert linguists. Because speakers of the AJA dialects are becoming scarce, the attention of linguists in this field has shifted from field- work interviews with native speakers to library-based analysis of texts written in those dialects. Fortunately, vast collections of AJA texts were preserved in printed books, journals and handwritten manuscripts. Analyz- ing this linguistic treasure-trove, however, is proving ∗Equal contribution †Supported by the Viterbi Fellowship in the Center for Computer Engineering at the Technion. to be challenging due to its size. The time-consuming manual annotation does not scale, and requires exper- tise that is hard to find. We aim to scale up the linguistic analysis of this Arabic dialect using NLP tools. In particular, our goal is to develop an NLP tool that will assist AJA linguists in their real-world task, in a way that they will find it useful. Basing our work on the existing linguistically Tagged Algerian Judeo-Arabic (TAJA) corpus (Tirosh- Becker and Becker, 2022), we set out to develop auto- matic methods for morpho-syntactic tagging of such texts. Several specially designed neural network tag- gers and an off-the-shelf morphological tagger were ex- …

Content truncated to 500 words