Cairo Student Code-Switch (CSCS) Corpus: An Annotated Egyptian Arabic-English Corpus

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 3973–3977 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 3973 Cairo Student Code-Switch (CSCS) Corpus: An Annotated Egyptian Arabic-English Corpus Mohamed Balabel1,2, Injy Hamed2,3, Slim Abdennadher3, Ngoc Thang Vu2, ¨Ozlem C¸ etino˘glu2 1IBM Germany Research and Development, B¨oblingen, Germany 2Institute for Natural Language Processing, University of Stuttgart, Stuttgart, Germany, 3Computer Science Department, The German University in Cairo, Cairo, Egypt mohamed.balabel1@ibm.com, {hamediy,thang.vu,ozlem.cetinoglu}@ims.uni-stuttgart.de, slim.abdennadher@guc.edu.eg Abstract Code-switching has become a prevalent phenomenon across many communities. It poses a challenge to NLP researchers, mainly due to the lack of available data needed for training and testing applications. In this paper, we introduce a new resource: a corpus of Egyptian- Arabic code-switch speech data that is fully tokenized, lemmatized and annotated for part-of-speech tags. Beside the corpus itself, we provide annotation guidelines to address the unique challenges of annotating code-switch data. Another challenge that we address is the fact that Egyptian Arabic orthography and grammar are not standardized. 1. Introduction Code-switching (CS) is the act of using more than one lan- guage in text, or more commonly, in speech. According to Poplack (1980), there are several types of language alterna- tions, which include: • Intra-sentential CS: defined as using multiple lan- guages within the same sentence. For example: “I do not think ������� ����� �� ���� student any more.” (I don’t think I want to be a student anymore.) • Inter-sentential CS: defined as switching languages from one sentence to another. For example: “It was really nice. ������ �������� .” (It was really nice. I learnt a lot.) CS is a phenomenon commonly observed in the Arabic- speaking world. Arabic itself encompasses many varieties, most prominently modern standard Arabic (MSA) and di- alectal Arabic (DA). The former is used in formal contexts such as formal writings and speeches, while the latter is the language used in everyday conversations as well as in- formal writings such as in social media. DA itself encom- passes a number of dialects, of which the most widely used (by number of speakers) is Egyptian Arabic. Arabic speak- ers typically mix MSA and DA in the same utterances (El- fardy and Diab, 2012), sometimes in addition to English, French or both. Colonization and international business and education have played a major role in introducing the En- glish and French languages into everyday conversations. Despite Arabic being one of the most widely-used lan- guages, there is still a huge gap in the available re- sources and Natural Language Processing (NLP) applica- tions. Most of the work has focused on MSA, with com- paratively less work on DA, and even less on code-switched DA. In this paper, we aim at providing a multi-lingual cor- pus1 for code-switched Egyptian Arabic-English that can 1The corpus is publicly available and can be obtained from the authors. be used in several NLP tasks. We build on the speech data in Hamed et al. (2018) by tokenizing, lemmatizing and an- …

Content truncated to 500 words