A list of Moroccan Darija Datasets grouped by name, data source, region and size.
# Moroccan Darija Datasets :morocco:
This repository aims to group Moroccan Darija Datasets to help make them available at once without the need to spend endless amounts of time looking for one. The datasets are grouped by name, data source, region, and size to provide as much information as possible to select the best dataset for the task at hand.
## Datasets `(13)`
Dataset
Data source
Region
Size
Link
Reference
1
Moroccan Arabic Sentiment Analysis Corpus
Twitter
Maghrebi (Moroccan)
2000 entries
source
2018 [1]
2
IADD: An integrated Arabic dialect identification dataset
Varied
Twitter
Facebook comments
Manual transcription from recordings of conversations movies or shows
Readers’ comments in websites of Arabic news papers
-->
Maghrebi, Levantine, Egyptian and Gulf
135,804 texts
source
2022 [2]
3
Dialectal Arabic Datasets
Twitter
Maghrebi, Levantine, Egyptian and Gulf
350 tweets per region
source
2018 [3]
4
MSDA Open Datasets
Social media posts
Arabic
-
source
2020 [4]
5
Moroccan Dialect Darija Open Dataset
Open source contributions
Maghrebi (Moroccan)
More than 13K
source
2021 [5]
6
Goud.ma: a News Dataset for Summarization in Moroccan Darija
goud.ma
Maghrebi (Moroccan)
158k news articles
source
2022 [6]
7
MNAD : Moroccan News Articles Dataset
Moroccan news websites
Maghrebi (Moroccan)
418 563 documents
source
2021 [7]
8
QADI: QCRI Arabic Dialect Identification
Twitter
Maghrebi, Levantine, Egyptian and Gulf
540k tweets
source
2020 [8]
9
Dvoice : An open source dataset for Automatic Speech Recognition on Moroccan dialectal Arabic
Voice recordings + text transcriptions
Maghrebi (Moroccan)
2392 training and 600 testing files
source
2021 [9]
10
ASAYAR: A Dataset for Arabic-Latin Scene Text Localization in Highway Traffic Panels
Images collected on different Moroccan highways, annotated manually.
Maghrebi (Moroccan)
1763 images
source
2020 [10]
11
OMCD: Offensive Moroccan Comments Dataset
A collection of comments from YouTube that have be …