Datasets:

QCRI
/

escwa

license: cc-by-nc-4.0
extra_gated_heading: QCRI-ALT Corpus License Agreement
extra_gated_prompt: >-
  By requesting access, you agree to the QCRI-ALT Creative commons Attribution
  Non Commercial 4.0 (cc-by-nc-4.0) license.
extra_gated_button_content: Agree & Request Access
extra_gated_fields:
  Full Name: text
  Institution: text
  Purpose of Use: text
  Country: country
language:
  - ar
  - en
tags:
  - Arabic
  - English
  - Code-switching
pretty_name: ESCWA-CS Corpus

🗣️ ESCWA-CS Corpus

The ESCWA-CS Corpus was collected over two days of meetings of the United Nations Economic and Social Commission for Western Asia (ESCWA) held in 2019.
It contains intra-sentential code-switching between Arabic and English, with some speakers—particularly from Algeria, Tunisia, and Morocco—alternating between Arabic and French.

The dataset spans approximately 2.8 hours of speech, featuring dialectal Arabic and a Code Mixing Index (CMI) of around 28%.
It serves as a valuable resource for studying bilingual and multilingual speech, dialectal variation, and code-mixing phenomena in spontaneous meeting conversations.

📘 Dataset Overview

Feature	Description
Type	Speech Corpus
Languages	Arabic, English, French (code-mixed)
Collection Context	ESCWA meetings, 2019
Duration	~2.8 hours
Phenomena	Code-switching, Dialectal Arabic
Code Mixing Index (CMI)	~28%

📄 Citation

If you use this dataset in your research, please cite the following paper:

@inproceedings{chowdhury2018towards,
    title={Towards Bangla Named Entity Recognition},
    author={Chowdhury, Shammur Absar and Alam, Firoj and Khan, Naira},
    booktitle={Proc. of ICCIT},
    
    year={2018},
    organization={IEEE}
}

@inproceedings{ali_arabic_2021,
    title = {Arabic {Code}-{Switching} {Speech} {Recognition} using {Monolingual} {Data}},
    booktitle = {{{{Proc. of INTERSPEECH}}}},
    author = {Ali, Ahmed and Chowdhury, Shammur and Hussein, Amir and Yasser, Hifny},
    year = {2021},
}