Datasets:

arbml

/

masader

like 10

Follow

Arabic Machine Learning 261

AQMAR

https://huggingface.co/datasets/arbml/AQMAR

https://www.cs.cmu.edu/~ark/ArabicNER/

CC BY-SA 3.0

ar

ar-MSA: (Arabic (Modern Standard Arabic))

wikipedia

text

crawling and annotation(other)

This is a 74,000-token corpus of 28 Arabic Wikipedia articles hand-annotated for named entities.

74,000

tokens

Low

CMU

nan

Recall-Oriented Learning of Named Entities in Arabic Wikipedia

https://aclanthology.org/E12-1017.pdf

Arab

No

other

Free

nan

No

named entity recognition

EACL

64.0

conference

European Chapter of the Association for Computational Linguistics

B. Mohit,Nathan Schneider,Rishav Bhowmick,Kemal Oflazer,Noah A. Smith

,,,,

We consider the problem of NER in Arabic Wikipedia, a semisupervised domain adaptation setting for which we have no labeled training data in the target domain. To facilitate evaluation, we obtain annotations for articles in four topical groups, allowing annotators to identify domain-specific entity types in addition to standard categories. Standard supervised learning on newswire text leads to poor target-domain recall. We train a sequence model and show that a simple modification to the online learner---a loss function encouraging it to "arrogantly" favor recall over precision---substantially improves recall and F1. We then adapt our model with self-training on unlabeled target-domain data; enforcing the same recall-oriented bias in the self-training stage yields marginal gains.

Zaid Alyafeai

MASC

https://huggingface.co/datasets/abdusahmbzuai/masc_dev

https://github.com/almoslmi/masc

custom

ar

mixed

reviews

text

crawling and annotation(other)

Multi-domain Arabic Sentiment Corpus (MASC) with a size of 8860 positive and negative reviews from different domains

8,860

sentences

Low

Multiple institutions

nan

Arabic senti-lexicon: Constructing publicly available language resources for Arabic sentiment analysis

https://journals.sagepub.com/doi/full/10.1177/0165551516683908

Arab

No

GitHub

Free

nan

No

sentiment analysis, dialect identification, topic classification

JIS

54.0

journal

Journal of Information Science

Tareq Al-Moslmi,M. Albared,Adel Al-Shabi,N. Omar,S. Abdullah

,,,,

Sentiment analysis is held to be one of the highly dynamic recent research fields in Natural Language Processing, facilitated by the quickly growing volume of Web opinion data. Most of the approaches in this field are focused on English due to the lack of sentiment resources in other languages such as the Arabic language and its large variety of dialects. In most sentiment analysis applications, good sentiment resources play a critical role. Based on that, in this article, several publicly available sentiment analysis resources for Arabic are introduced. This article introduces the Arabic senti-lexicon, a list of 3880 positive and negative synsets annotated with their part of speech, polarity scores, dialects synsets and inflected forms. This article also presents a Multi-domain Arabic Sentiment Corpus (MASC) with a size of 8860 positive and negative reviews from different domains. In this article, an in-depth study has been conducted on five types of feature sets for exploiting effective features and investigating their effect on performance of Arabic sentiment analysis. The aim is to assess the quality of the developed language resources and to integrate different feature sets and classification algorithms to synthesise a more accurate sentiment analysis method. The Arabic senti-lexicon is used for generating feature vectors. Five well-known machine learning algorithms: naïve Bayes, k-nearest neighbours, support vector machines (SVMs), logistic linear regression and neural network are employed as base-classifiers for each of the feature sets. A wide range of comparative experiments on standard Arabic data sets were conducted, discussion is presented and conclusions are drawn. The experimental results show that the Arabic senti-lexicon is a very useful resource for Arabic sentiment analysis. Moreover, results show that classifiers which are trained on feature vectors derived from the corpus using the Arabic sentiment lexicon are more accurate than classifiers trained using the raw corpus.

Zaid Alyafeai

Arabic senti-lexicon

https://huggingface.co/datasets/arbml/Senti_Lexicon

https://github.com/almoslmi/masc

custom

ar

ar-MSA: (Arabic (Modern Standard Arabic))

other

text

manual curation

a list of 3880 positive and negative synsets annotated with their part of speech, polarity scores, dialects synsets and inflected forms

3,880

tokens

Low

Multiple institutions

nan

Arabic senti-lexicon: Constructing publicly available language resources for Arabic sentiment analysis

https://journals.sagepub.com/doi/full/10.1177/0165551516683909

Arab

No

GitHub

Free

nan

No

part of speech tagging, sentiment analysis

JIS

54.0

journal

Journal of Information Science

Tareq Al-Moslmi,M. Albared,Adel Al-Shabi,N. Omar,S. Abdullah

,,,,

Sentiment analysis is held to be one of the highly dynamic recent research fields in Natural Language Processing, facilitated by the quickly growing volume of Web opinion data. Most of the approaches in this field are focused on English due to the lack of sentiment resources in other languages such as the Arabic language and its large variety of dialects. In most sentiment analysis applications, good sentiment resources play a critical role. Based on that, in this article, several publicly available sentiment analysis resources for Arabic are introduced. This article introduces the Arabic senti-lexicon, a list of 3880 positive and negative synsets annotated with their part of speech, polarity scores, dialects synsets and inflected forms. This article also presents a Multi-domain Arabic Sentiment Corpus (MASC) with a size of 8860 positive and negative reviews from different domains. In this article, an in-depth study has been conducted on five types of feature sets for exploiting effective features and investigating their effect on performance of Arabic sentiment analysis. The aim is to assess the quality of the developed language resources and to integrate different feature sets and classification algorithms to synthesise a more accurate sentiment analysis method. The Arabic senti-lexicon is used for generating feature vectors. Five well-known machine learning algorithms: naïve Bayes, k-nearest neighbours, support vector machines (SVMs), logistic linear regression and neural network are employed as base-classifiers for each of the feature sets. A wide range of comparative experiments on standard Arabic data sets were conducted, discussion is presented and conclusions are drawn. The experimental results show that the Arabic senti-lexicon is a very useful resource for Arabic sentiment analysis. Moreover, results show that classifiers which are trained on feature vectors derived from the corpus using the Arabic sentiment lexicon are more accurate than classifiers trained using the raw corpus.

Zaid Alyafeai

CANERCorpus

https://huggingface.co/datasets/caner

https://github.com/RamziSalah/Classical-Arabic-Named-Entity-Recognition-Corpus

unknown

ar

ar-MSA: (Arabic (Modern Standard Arabic))

other

text

manual curation

It is freely available and manual annotation by human experts, containing more than 7,000 Hadiths

72,108

tokens

Low

Universiti Kebangsaan

nan

Building the Classical Arabic Named Entity Recognition Corpus (CANERCorpus)

https://ieeexplore.ieee.org/document/8464820/authors#authors

Arab

No

GitHub

Free

nan

No

named entity recognition

CAMP

0.0

conference

International Conference on Information Retrieval and Knowledge Management

Ramzi Salah,Lailatul Qadri Binti Zakaria

,

The past decade has witnessed construction of the background information resources to overcome several challenges in text mining tasks. For non-English languages with poor knowledge sources such as Arabic, these challenges have become more salient especially for handling the natural language processing applications that require human annotation. In the Named Entity Recognition (NER) task, several researches have been introduced to address the complexity of Arabic in terms of morphological and syntactical variations. However, there are a small number of studies dealing with Classical Arabic (CA) that is the official language of Quran and Hadith. CA was also used for archiving the Islamic topics that contain a lot of useful information which could of great value if extracted. Therefore, in this paper, we introduce Classical Arabic Named Entity Recognition corpus as a new corpus of tagged data that can be useful for handling the issues in recognition of Arabic named entities. It is freely available and manual annotation by human experts, containing more than 7,000 Hadiths. Based on Islamic topics, we classify named entities into 20 types which include the specific-domain entities that have not been handled before such as Allah, Prophet, Paradise, Hell, and Religion. The differences between the standard and classical Arabic are described in details during this work. Moreover, the comprehensive statistical analysis is introduced to measure the factors that play important role in manual human annotation.

Zaid Alyafeai

SAD

https://huggingface.co/datasets/arbml/SAD

http://www.cs.stir.ac.uk/~lss/arabic/

unknown

ar

ar-MSA: (Arabic (Modern Standard Arabic))

transcribed audio

spoken

manual curation

The Arabic speech corpus for isolated words contains 9992 utterances of 20 words spoken by 50 native male Arabic speakers.

6

hours

Low

University of Stirling

nan

On Improving the Classification Capability of Reservoir Computing for Arabic Speech Recognition

https://link.springer.com/content/pdf/10.1007%2F978-3-319-11179-7_29.pdf

Arab

No

other

Free

nan

Yes

speech recognition

ICANN

13.0

conference

International Conference on Artificial Neural Networks

A. Alalshekmubarak,Leslie S. Smith

,

Designing noise-resilient systems is a major challenge in the field of automated speech recognition (ASR). These systems are crucial for real-world applications where high levels of noise tend to be present. We introduce a noise robust system based on Echo State Networks and Extreme Kernel machines which we call ESNEKM. To evaluate the performance of the proposed system, we used our recently released public Arabic speech dataset and the well-known spoken Arabic digits (SAD) dataset. Different feature extraction methods considered in this study include mel-frequency cepstral coefficients (MFCCs), perceptual linear prediction (PLP) and RASTA- perceptual linear prediction. These extracted features were fed to the ESNEKM and the result compared with a baseline hidden Markov model (HMM), so that nine models were compared in total. ESNEKM models outperformed HMM models under all the feature extraction methods, noise levels, and noise types. The best performance was obtained by the model that combined RASTA-PLP with ESNEKM.

Zaid Alyafeai

ANTCORPUS

https://huggingface.co/datasets/arbml/antcorpus

https://github.com/antcorpus/antcorpus.data

custom

ar

ar-MSA: (Arabic (Modern Standard Arabic))

news articles

text

crawling

ANT Corpus, which is collected from RSS Feeds.

6,005

documents

Low

Multiple institutions

nan

ANT Corpus: An Arabic News Text Collection for Textual Classification

https://ieeexplore.ieee.org/abstract/document/8308275/authors#authors

Arab

No

GitHub

Free

nan

No

topic classification

AICCSA

17.0

conference

International Conference on Computer Systems and Applications

Amina Chouigui,Oussama Ben Khiroun,Bilel Elayeb

,,

We propose in this paper a new online Arabic corpus of news articles, named ANT Corpus, which is collected from RSS Feeds. Each document represents an article structured in the standard XML TREC format. We use the ANT Corpus for Text Classification (TC) by applying the SVM and Naive Bayes (NB) classifiers to assign to each article its accurate predefined category. We study also in this work the contribution of terms weighting, stop-words removal and light stemming on Arabic TC. The experimental results prove that the text length affects considerably the TC accuracy and that titles words are not sufficiently significant to perform good classification rates. As a conclusion, the SVM method gives the best results of classification of both titles and texts parts.

Zaid Alyafeai

Religious Hate Speech

https://huggingface.co/datasets/arbml/Religious_Hate_Speech

https://github.com/nuhaalbadi/Arabic_hatespeech

unknown

ar

mixed

social media

text

crawling and annotation(other)

training dataset contains 5,569 examples, while the testing dataset contains 567 examples collected from twittter

6,136

sentences

High

Multiple institutions

nan

Are they Our Brothers? Analysis and Detection of Religious Hate Speech in the Arabic Twittersphere

https://ieeexplore.ieee.org/document/8508247/authors#authors

Arab

No

GitHub

Free

nan

Yes

hate speech detection

ASONAM

72.0

conference

IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining

Nuha Albadi,Maram Kurdi,Shivakant Mishra

,,

Religious hate speech in the Arabic Twittersphere is a notable problem that requires developing automated tools to detect messages that use inflammatory sectarian language to promote hatred and violence against people on the basis of religious affiliation. Distinguishing hate speech from other profane and vulgar language is quite a challenging task that requires deep linguistic analysis. The richness of the Arabic morphology and the limited available resources for the Arabic language make this task even more challenging. To the best of our knowledge, this paper is the first to address the problem of identifying speech promoting religious hatred in the Arabic Twitter. In this work, we describe how we created the first publicly available Arabic dataset annotated for the task of religious hate speech detection and the first Arabic lexicon consisting of terms commonly found in religious discussions along with scores representing their polarity and strength. We then developed various classification models using lexicon-based, n-gram-based, and deep-learning-based approaches. A detailed comparison of the performance of different models on a completely new unseen dataset is then presented. We find that a simple Recurrent Neural Network (RNN) architecture with Gated Recurrent Units (GRU) and pre-trained word embeddings can adequately detect religious hate speech with 0.84 Area Under the Receiver Operating Characteristic curve (AUROC).

Zaid Alyafeai

omcca

nan

https://github.com/AhmedObaidi/omcca

unknown

ar

mixed

reviews

text

crawling and annotation(other)

Opinion Mining Corpus for Colloquial Variety of Arabic language

28,576

sentences

Low

Multiple institutions

nan

Opinion Mining: Analysis of Comments Written in Arabic Colloquial

http://www.iaeng.org/publication/WCECS2016/WCECS2016_pp470-475.pdf

Arab

No

GitHub

Free

nan

No

dialect identification, sentiment analysis

WCECS

8.0

conference

World Congress on Engineering and Computer Science

Ahmed Y. Al-Obaidi,V. Samawi

,

In Arab nations, people used to express their opinions using colloquial dialects depending on the country to which they belong to. Analyzing reviews written in various Arabic dialects is a challenging problem. This is because some words could have different meanings in various dialects. Furthermore, dialects could contain words that do not belong to classical Arabic language.‎This research tackles the problem of sentiment analysis of reviews and comments written in colloquial dialects of Arabic language, at which the ability of different machine learning algorithms and features are examined in polarity determination. In this work, people's reviews (written in different dialects) are classified into positive or negative opinions. Each dialect comes with its own stopwords list. Consequently, a list of stop-words that suits different dialects in addition to modern standard Arabic (MSA) is suggested. In this paper, a light stemmer that suits dialects is developed. Two feature sets are utilized (bag of words (BoW), and N-gram of words) to investigate their effectiveness in sentiment analysis. Finally, Naïve-Bayes, Support vector machine (SVM), and Maximum Entropy machine learning algorithms are applied to study their performance in opinion mining. F1-measure is used to evaluate the performance of these machine learning algorithms. To train and test the suggested system performance, we built a corpus1 of reviews by collecting reviews written in two dialects (Saudi dialect and Jordanian dialect). The testing results show that Maximum Entropy outperforms the other two machine learning algorithms. Using N-gram (with N=3) as features set improves the performance of the three machine learning algorithms.

Zaid Alyafeai

Corpus of Offensive Language in Arabic

nan

https://onedrive.live.com/?authkey=!ACDXj_ZNcZPqzy0&id=6EF6951FBF8217F9!105&cid=6EF6951FBF8217F9

unknown

ar

mixed

social media

text

crawling and annotation(other)

a dataset of YouTube comments in Arabic, specifically designed to be used for thedetection of offensive language in a machine learning scenario

15,050

sentences

High

University of Limerick

nan

Dataset Construction for the Detection of Anti-Social Behaviour in Online Communication in Arabic

https://core.ac.uk/download/pdf/211161742.pdf

Arab

No

OneDrive

Free

nan

No

offensive language detection, hate speech detection

ACLING

33.0

conference

nternational Conference on AI in Computational Linguistics

Azalden Alakrot,Liam Murray,Nikola S. Nikolov

,,

Abstract Warning: this paper contains a range of words which may cause offence. In recent years, many studies target anti-social behaviour such as offensive language and cyberbullying in online communication. Typically, these studies collect data from various reachable sources, the majority of the datasets being in English. However, to the best of our knowledge, there is no dataset collected from the YouTube platform targeting Arabic text and overall there are only a few datasets of Arabic text, collected from other social platforms for the purpose of offensive language detection. Therefore, in this paper we contribute to this field by presenting a dataset of YouTube comments in Arabic, specifically designed to be used for the detection of offensive language in a machine learning scenario. Our dataset contains a range of offensive language and flaming in the form of YouTube comments. We document the labelling process we have conducted, taking into account the difference in the Arab dialects and the diversity of perception of offensive language throughout the Arab world. Furthermore, statistical analysis of the dataset is presented, in order to make it ready for use as a training dataset for predictive modelling.

Zaid Alyafeai

Qatari heritage corpus

nan

https://data.world/saraalmulla/qatari-heritage-expressions

CDLA-Permissive-1.0

ar

ar-QA: (Arabic (Qatar))

other

text

crawling and annotation(other)

qatari heritage expressions dataset with translations

1,000

sentences

Low

Hamad Bin Khalifa University

nan

Building a Corpus of Qatari Arabic Expressions

https://aclanthology.org/2020.osact-1.4.pdf

Arab

No

data.world

Upon-Request

nan

No

machine translation

OSACT

1.0

workshop

Workshop on Open-Source Arabic Corpora and Processing Tools

Sara Al-Mulla,W. Zaghouani

,

The current Arabic natural language processing resources are mainly build to address the Modern Standard Arabic (MSA), while we witnessed some scattered efforts to build resources for various Arabic dialects such as the Levantine and the Egyptian dialects. We observed a lack of resources for Gulf Arabic and especially the Qatari variety. In this paper, we present the first Qatari idioms and expression corpus of 1000 entries. The corpus was created from on-line and printed sources in addition to transcribed recorded interviews. The corpus covers various Qatari traditional expressions and idioms. To this end, audio recordings were collected from interviews and an online survey questionnaire was conducted to validate our data. This corpus aims to help advance the dialectal Arabic Speech and Natural Language Processing tools and applications for the Qatari dialect.

Zaid Alyafeai

KSU Rich Arabic Speech Database

nan

https://catalog.ldc.upenn.edu/LDC2014S02

custom

ar

ar-MSA: (Arabic (Modern Standard Arabic))

transcribed audio

spoken

crawling and annotation(other)

it has 752 speakers; the speakers are from different ethnic groups: Saudis, Arabs, and non-Arabs;

590

hours

Low

King Saud University

nan

KSU Rich Arabic Speech Database

https://catalog.ldc.upenn.edu/docs/LDC2014S02/KSU-Rich-Arabic-Speech-Database.pdf

Arab

No

LDC

With-Fee

2000 $

No

speech recognition

other

20.0

preprint

nan

M. Alsulaiman,G. Muhammad,M. Bencherif,A. Mahmood,Z. Ali

,,,,

Arabic is one of the major languages in the world. Unfortunately not so much research in Arabic speaker recognition has been done. One main reason for this lack of research is the unavailability of rich Arabic speech databases. In this paper, we present a rich and comprehensive Arabic speech database that we developed for the Arabic speaker / speech recognition research and/or applications. The database is rich in different aspects: (a) it has 752 speakers; (b) the speakers are from different ethnic groups: Saudis, Arabs, and non-Arabs; (c) utterances are both read text and spontaneous; (d) scripts are of different dimensions, such as, isolated words, digits, phonetically rich words, sentences, phonetically balanced sentences, paragraphs, etc.; (e) different sets of microphones with medium and high quality; (f) telephony and non-telephony speech; (g) three different recording environments: office, sound proof room, and cafeteria; (h) three different sessions, where the recording sessions are scheduled at least with 2 weeks interval. Because of the richness of this database, it can be used in many Arabic, and non-Arabic, speech processing researches, such as speaker / speech recognition, speech analysis, accent identification, ethnic groups / nationality recognition, etc. The richness of the database makes it a valuable resource for research in Arabic speech processing in particular and for research in speech processing in general. The database was carefully manually verified. The manual verification was complemented with automatic verification. Validation was performed on a subset of the database where the recognition rate reached 100% for Saudi speakers and 96% for non-Saudi speakers by using a system with 12 Mel frequency Cepstral coefficients, and 32 Gaussian mixtures.

Zaid Alyafeai

XTREME

https://huggingface.co/datasets/xtreme

https://github.com/google-research/xtreme

Apache-2.0

multilingual

ar-MSA: (Arabic (Modern Standard Arabic))

other

text

other

contains many datasets from different benchmarks like XNLI, TYDiQA, etc.

nan

sentences

Low

Google

Tydiqa, xnli, etc.

XTREME: A Massively Multilingual Multi-task Benchmark for Evaluating Cross-lingual Generalization

https://arxiv.org/pdf/2003.11080.pdf

Arab

No

GitHub

Free

nan

Yes

natural language inference,part of speech tagging,named entity recognition,question answering,machine translation

ICML

209.0

conference

International Conference on Machine Learning

Junjie Hu,Sebastian Ruder,Aditya Siddhant,Graham Neubig,Orhan Firat,M. Johnson

,DeepMind,,,,

Much recent progress in applications of machine learning models to NLP has been driven by benchmarks that evaluate models across a wide variety of tasks. However, these broad-coverage benchmarks have been mostly limited to English, and despite an increasing interest in multilingual models, a benchmark that enables the comprehensive evaluation of such methods on a diverse range of languages and tasks is still missing. To this end, we introduce the Cross-lingual TRansfer Evaluation of Multilingual Encoders XTREME benchmark, a multi-task benchmark for evaluating the cross-lingual generalization capabilities of multilingual representations across 40 languages and 9 tasks. We demonstrate that while models tested on English reach human performance on many tasks, there is still a sizable gap in the performance of cross-lingually transferred models, particularly on syntactic and sentence retrieval tasks. There is also a wide spread of results across languages. We release the benchmark to encourage research on cross-lingual learning methods that transfer linguistic knowledge across a diverse and representative set of languages and tasks.

Zaid Alyafeai

Arabic document classification dataset

https://huggingface.co/datasets/arbml/Document_Classification

https://diab.edublogs.org/dataset-for-arabic-document-classification/

unknown

ar

ar-MSA: (Arabic (Modern Standard Arabic))

other

text

manual curation

The dataset contains nine major disciplines: Art, Literature, Religion, Politics, Law, Economy, Sport, Health, and Technology.

2,700

documents

Low

Multiple institutions

nan

On the Impact of Dataset Characteristics on Arabic Document Classification

https://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.800.5666&rep=rep1&type=pdf

Arab

Yes

other

Free

nan

No

topic classification

IJCA

11.0

journal

International Journal of Computer Applications

Diab Abuaiadah,Jihad El Sana,Walid Abusalah

,,

paper describes the impact of dataset characteristics on the results of Arabic document classification algorithms using TF-IDF representations. The experiments compared different stemmers, different categories and different training set sizes, and found that different dataset characteristics produced widely differing results, in one case attaining a remarkable 99% recall (accuracy). The use of a standard dataset would eliminate this variability and enable researchers to gain comparable knowledge from the published results.

Mustafa Ghaleb

KSUCCA Corpus

https://huggingface.co/datasets/arbml/KSUCCA

https://sourceforge.net/projects/ksucca-corpus/

unknown

ar

ar-CLS: (Arabic (Classic))

books

text

crawling

KSUCCA is a pioneering 50+ million word corpus that captures the culture of a nation.

410

documents

Low

King Saud University

nan

The Design and Construction of the 50 Million Words KSUCCA

https://eprints.whiterose.ac.uk/81860/1/TheDesignConstruction50MillionWordKSUCCA.pdf

Arab

No

sourceforge

Free

nan

No

semantic similarity

other

30.0

preprint

nan

Maha AlRabiah,A. Al-Salman,E. Atwell

,,

In this paper, we report the design and construction of King Saud University Corpus of Classical Arabic (KSUCCA), which is part of ongoing research that attempts to study the meanings of words used in the holy Quran, through analysis of their distributional semantics in contemporaneous texts. The holy Quranic text was revealed in pure Classical Arabic, which forms the basis of Arabic linguistic theory and which is well understood by the educated Arabic reader. Therefore, it is necessary to investigate the distributional lexical semantics of the Quran's words in the light of similar texts (corpus) that are written in pure Classical Arabic. To the best of our knowledge, there exist only two corpora of Classical Arabic; one is part of the King Abdulaziz City for Science and Technology Arabic Corpus (KACST Arabic Corpus) and the other is the Classical Arabic Corpus (CAC) (Elewa, 2009). However, neither of the two corpora is adequate for our research; the former does not cover many genres such as: Linguistics, Literature, Science, Sociology and Biography; and it only contains 17+ million words, so it is not very large. While the latter is even smaller with only 5 million words. Therefore, we made an effort to carefully design and compose our own corpus bearing in mind that it should be large enough, balanced, and representative so that any result obtained from it can be generalized for Classical Arabic. In addition, we tried to make the design general enough in order to make the corpus also appropriate for other research.

Mustafa Ghaleb

XNLI

https://huggingface.co/datasets/xnli

https://github.com/facebookresearch/XNLI

CC BY-NC 4.0

multilingual

ar-MSA: (Arabic (Modern Standard Arabic))

other

text

human translation

evaluation set for XLU by extending the development and test sets of the Multi-Genre Natural Language Inference Corpus (MultiNLI) to 15 languages,

7,500

sentences

Low

Facebook, NYU

nan

XNLI: Evaluating Cross-lingual Sentence Representations

https://arxiv.org/pdf/1809.05053.pdf

Arab

No

GitHub

Free

nan

Yes

natural language inference

EMNLP

354.0

conference

Conference on Empirical Methods in Natural Language Processing

Alexis Conneau,Guillaume Lample,Ruty Rinott,Adina Williams,Samuel R. Bowman,Holger Schwenk,Veselin Stoyanov

,,,Facebook AI Research,New York University,,

State-of-the-art natural language processing systems rely on supervision in the form of annotated data to learn competent models. These models are generally trained on data in a single language (usually English), and cannot be directly used beyond that language. Since collecting data in every language is not realistic, there has been a growing interest in cross-lingual language understanding (XLU) and low-resource cross-language transfer. In this work, we construct an evaluation set for XLU by extending the development and test sets of the Multi-Genre Natural Language Inference Corpus (MultiNLI) to 14 languages, including low-resource languages such as Swahili and Urdu. We hope that our dataset, dubbed XNLI, will catalyze research in cross-lingual sentence understanding by providing an informative standard evaluation task. In addition, we provide several baselines for multilingual sentence understanding, including two based on machine translation systems, and two that use parallel data to train aligned multilingual bag-of-words and LSTM encoders. We find that XNLI represents a practical and challenging evaluation suite, and that directly translating the test data yields the best performance among available baselines.

Zaid Alyafeai

OSAC

https://huggingface.co/datasets/arbml/OSAC_CNN

https://sourceforge.net/projects/ar-text-mining/files/Arabic-Corpora/

unknown

ar

ar-MSA: (Arabic (Modern Standard Arabic))

news articles

text

crawling

collecting the largest free accessible Arabic corpus, OSAC, which contains about 18M words and about 0.5M district keywords.

22,429

documents

Low

Islamic University of Gaza

nan

OSAC: Open Source Arabic Corpora

http://site.iugaza.edu.ps/wp-content/uploads/mksaad-OSAC-OpenSourceArabicCorpora-EECS10-rev9(1).pdf

Arab

No

sourceforge

Free

nan

No

topic classification

other

101.0

preprint

nan

Motaz Saad,W. Ashour

The Islamic University of Gaza,

Arabic Linguistics is promising research field. The acute lack of free public accessible Arabic corpora is one of the major difficulties that Arabic linguistics researches face. The effort of this paper is a step towards supporting Arabic linguistics research field. This paper presents the complex nature of Arabic language, pose the problems of: (1) lacking free public Arabic corpora, (2) the lack of high-quality, wellstructured Arabic digital contents. The paper finally presents OSAC, the largest free accessible that we collected.

Zaid Alyafeai

wikiann

https://huggingface.co/datasets/wikiann

https://drive.google.com/drive/folders/1Q-xdT99SeaCghihGa7nRkcXGwRGUIsKN

unknown

multilingual

ar-MSA: (Arabic (Modern Standard Arabic))

wikipedia

text

crawling and annotation(other)

Both name tagging and linking results for 282 languages are promising on Wikipedia data and on-Wikipedia data

185,000

tokens

Low

Multiple institutions

nan

Cross-lingual Name Tagging and Linking for 282 Languages

https://aclanthology.org/P17-1178.pdf

Arab

No

Gdrive

Free

nan

Yes

named entity recognition

ACL

168.0

conference

Assofications of computation linguisitcs

Xiaoman Pan,Boliang Zhang,Jonathan May,J. Nothman,Kevin Knight,Heng Ji

,,,,,

The ambitious goal of this work is to develop a cross-lingual name tagging and linking framework for 282 languages that exist in Wikipedia. Given a document in any of these languages, our framework is able to identify name mentions, assign a coarse-grained or fine-grained type to each mention, and link it to an English Knowledge Base (KB) if it is linkable. We achieve this goal by performing a series of new KB mining methods: generating “silver-standard” annotations by transferring annotations from English to other languages through cross-lingual links and KB properties, refining annotations through self-training and topic selection, deriving language-specific morphology features from anchor links, and mining word translation pairs from cross-lingual links. Both name tagging and linking results for 282 languages are promising on Wikipedia data and on-Wikipedia data.

Zaid Alyafeai

Tanzil

https://huggingface.co/datasets/tanzil

https://opus.nlpl.eu/Tanzil.php

custom

multilingual

ar-CLS: (Arabic (Classic))

other

text

human translation

This is a collection of Quran translations compiled by the Tanzil project

12,472

sentences

Low

OPUS

nan

Parallel Data, Tools and Interfaces in OPUS

http://www.lrec-conf.org/proceedings/lrec2012/pdf/463_Paper.pdf

Arab

No

other

Free

nan

No

machine translation

LREC

1006.0

conference

International Conference on Language Resources and Evaluation

J. Tiedemann

nan

This paper presents the current status of OPUS, a growing language resource of parallel corpora and related tools. The focus in OPUS is to provide freely available data sets in various formats together with basic annotation to be useful for applications in computational linguistics, translation studies and cross-linguistic corpus studies. In this paper, we report about new data sets and their features, additional annotation tools and models provided from the website and essential interfaces and on-line services included in the project.

Zaid Alyafeai

Tatoeba

https://huggingface.co/datasets/tatoeba

https://tatoeba.org/en/sentences/show_all_in/ara/none/none/indifferent

CC BY 2.0

multilingual

ar-MSA: (Arabic (Modern Standard Arabic))

other

text

human translation

Tatoeba covers 338 languages and contains a total of 7.8 million sentences.

36,792

sentences

Low

OPUS

nan

Parallel Data, Tools and Interfaces in OPUS

http://www.lrec-conf.org/proceedings/lrec2012/pdf/463_Paper.pdf

Arab

No

other

Free

nan

No

machine translation

LREC

1006.0

conference

International Conference on Language Resources and Evaluation

J. Tiedemann

nan

This paper presents the current status of OPUS, a growing language resource of parallel corpora and related tools. The focus in OPUS is to provide freely available data sets in various formats together with basic annotation to be useful for applications in computational linguistics, translation studies and cross-linguistic corpus studies. In this paper, we report about new data sets and their features, additional annotation tools and models provided from the website and essential interfaces and on-line services included in the project.

Zaid Alyafeai

TaPaCo

https://huggingface.co/datasets/tapaco

https://zenodo.org/record/3707949#.YRKu0ogzaUk

CC BY 2.0

multilingual

ar-MSA: (Arabic (Modern Standard Arabic))

other

text

other

The corpus contains a total of 1.9 million sentences, with 200 – 250 000 sentences per language

6,446

sentences

Low

University of Helsinki

Tatoeba

TaPaCo: A Corpus of Sentential Paraphrases for 73 Languages

https://aclanthology.org/2020.lrec-1.848.pdf

Arab

No

zenodo

Free

nan

No

paraphrasing

LREC

4.0

conference

International Conference on Language Resources and Evaluation

Yves Scherrer

nan

This paper presents TaPaCo, a freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 - 250 000 sentences per language. It covers a range of languages for which, to our knowledge, no other paraphrase dataset exists. The dataset is available at https://doi.org/10.5281/zenodo.3707949.

Zaid Alyafeai

LASER

nan

https://github.com/facebookresearch/LASER

BSD

multilingual

ar-MSA: (Arabic (Modern Standard Arabic))

other

text

human translation

aligned sentences in 112 languages

1,000

sentences

Low

Multiple institutions

Tatoeba

Massively Multilingual Sentence Embeddings for Zero-Shot Cross-Lingual Transfer and Beyond

https://arxiv.org/pdf/1812.10464.pdf

Arab

No

GitHub

Free

nan

No

machine translation

TACL

374.0

journal

Transactions of the Association for Computational Linguistics

Mikel Artetxe,Holger Schwenk

,

Abstract We introduce an architecture to learn joint multilingual sentence representations for 93 languages, belonging to more than 30 different families and written in 28 different scripts. Our system uses a single BiLSTM encoder with a shared byte-pair encoding vocabulary for all languages, which is coupled with an auxiliary decoder and trained on publicly available parallel corpora. This enables us to learn a classifier on top of the resulting embeddings using English annotated data only, and transfer it to any of the 93 languages without any modification. Our experiments in cross-lingual natural language inference (XNLI data set), cross-lingual document classification (MLDoc data set), and parallel corpus mining (BUCC data set) show the effectiveness of our approach. We also introduce a new test set of aligned sentences in 112 languages, and show that our sentence embeddings obtain strong results in multilingual similarity search even for low- resource languages. Our implementation, the pre-trained encoder, and the multilingual test set are available at https://github.com/facebookresearch/LASER.

Zaid Alyafeai

NArabizi treebank

nan

https://parsiti.github.io/NArabizi/

unknown

ar

ar-DZ: (Arabic (Algeria))

other

text

crawling and annotation(other)

fully annotated in morpho-syntax and Universal Dependency syntax, with full translation at both the word and the sentence levels

1,500

sentences

Low

Inria,Sorbonne Université

nan

Building a User-Generated Content North-African Arabizi Treebank: Tackling Hell

https://aclanthology.org/2020.acl-main.107.pdf

Arab-Latn

Yes

GitHub

Free

nan

Yes

part of speech tagging,dependency parsing,machine translation

ACL

14.0

conference

Assofications of computation linguisitcs

Djamé Seddah,Farah Essaidi,Amal Fethi,Matthieu Futeral,Benjamin Müller,Pedro Javier Ortiz Suárez,B. Sagot,Abhishek Srivastava

Inria;Sorbonne Université,,,,,Inria;Sorbonne Université,,

We introduce the first treebank for a romanized user-generated content variety of Algerian, a North-African Arabic dialect known for its frequent usage of code-switching. Made of 1500 sentences, fully annotated in morpho-syntax and Universal Dependency syntax, with full translation at both the word and the sentence levels, this treebank is made freely available. It is supplemented with 50k unlabeled sentences collected from Common Crawl and web-crawled data using intensive data-mining techniques. Preliminary experiments demonstrate its usefulness for POS tagging and dependency parsing. We believe that what we present in this paper is useful beyond the low-resource language community. This is the first time that enough unlabeled and annotated data is provided for an emerging user-generated content dialectal language with rich morphology and code switching, making it an challenging test-bed for most recent NLP approaches.

Zaid Alyafeai

NArabizi corpus

https://huggingface.co/datasets/arbml/NArabizi

https://github.com/SamiaTouileb/Narabizi

unknown

ar

ar-DZ: (Arabic (Algeria))

other

text

crawling and annotation(other)

extension of NArabizi treebank by adding to annotations.

1,500

sentences

Low

University of Oslo

NArabizi treebank

The interplay between language similarity and script on a novel multi-layer Algerian dialect corpus

https://arxiv.org/pdf/2105.07400.pdf

Arab-Latn

Yes

GitHub

Free

nan

Yes

part of speech tagging,dependency parsing,machine translation, sentiment analysis,transliteration, topic classification

FINDINGS

1.0

conference

Findings of the Association for Computational Linguistics

Samia Touileb,Jeremy Barnes

,University of the Basque Country UPV/EHU

Recent years have seen a rise in interest for cross-lingual transfer between languages with similar typology, and between languages of various scripts. However, the interplay between language similarity and difference in script on cross-lingual transfer is a less studied problem. We explore this interplay on cross-lingual transfer for two supervised tasks, namely part-of-speech tagging and sentiment analysis. We introduce a newly annotated corpus of Algerian user-generated comments comprising parallel annotations of Algerian written in Latin, Arabic, and code-switched scripts, as well as annotations for sentiment and topic categories. We perform baseline experiments by fine-tuning multi-lingual language models. We further explore the effect of script vs. language similarity in cross-lingual transfer by fine-tuning multi-lingual models on languages which are a) typologically distinct, but use the same script, b) typologically similar, but use a distinct script, or c) are typologically similar and use the same script. We find there is a delicate relationship between script and typology for part-of-speech, while sentiment analysis is less sensitive.

Zaid Alyafeai

DefArabicQA

nan

https://sites.google.com/site/anlprg/outils-et-corpus-realises

unknown

ar

ar-MSA: (Arabic (Modern Standard Arabic))

wikipedia

text

crawling and annotation(other)

2000 snippets returned by Google search engine and Wikipedia Arabic version and a set of 50 organization definition questions

2,000

sentences

Low

Multiple institutions

nan

DefArabicQA: Arabic Definition Question Answering System

http://personales.upv.es/prosso/resources/TriguiEtAl_LREC10.pdf

Arab

No

other

Free

nan

No

question answering

LREC

51.0

conference

International Conference on Language Resources and Evaluation

Omar Trigui,L. Belguith,P. Rosso

,,

Today the Web is the largest resource of knowledge and, therefore, sometimes this makes it difficult to find precise information. Current search engines can only return ranked snippets containing the effective answers to a query user. But, they can not return the exact answers. Question Answering systems present the solution to obtain effective and exact answers to a user question asked in natural language question instead of keywords query. Unfortunately, Question Answering task for the Arabic language has not been investigated enough in the last decade, compared to other languages. In this paper, we tackle the definition Question Answering task for the Arabic language. We propose an Arabic definitional Question Answering system based on a pattern approach to identify exact and accurate definitions about organization using Web resources. We experimented this system using 2000 snippets returned by Google search engine and Wikipedia Arabic version and a set of 50 organization definition questions. The obtained results are very encouraging: (90%) of the questions used have complete (vital) definitions in the top-five answers and (64%) of them have complete definitions in the top-one answer. MRR was (0.81).

Zaid Alyafeai

TuDiCoI

https://huggingface.co/datasets/arbml/TuDiCoI

https://sites.google.com/site/anlprg/outils-et-corpus-realisess

unknown

ar

ar-TN: (Arabic (Tunisia))

transcribed audio

spoken

manual curation

The corpus consists of 434 1465 staff utterances and 1615 client utterances

127

sentences

Low

University of Sfax

nan

LEXICAL STUDY OF A SPOKEN DIALOGUE CORPUS IN TUNISIAN DIALECT

https://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.452.7847&rep=rep1&type=pdf

Arab

No

other

Free

nan

No

speech recognition

other

15.0

preprint

nan

M. Graja,M. Jaoua,L. Belguith

,,

The aim of this paper is to present a lexical study of a spoken dialogue corpus in Tunisian dialect since such resources does not currently existing. The lexical analysis permits to take into account the specificity of Tunisian dialect by identifying lexical varieties and significant elements used in the spoken dialogue. This can lead us to provide a useful characterization for dialogue systems and help us to develop models and methods specifically designed for Tunisian dialect.

Zaid Alyafeai

AFEWC

https://huggingface.co/datasets/arbml/AFEWC

https://sourceforge.net/projects/crlcl/

CC BY-NC 2.0

multilingual

ar-MSA: (Arabic (Modern Standard Arabic))

wikipedia

text

crawling

a multilingual comparable text articles in Arabic, French, and English languages

40,290

documents

Low

INRIA

nan

Extracting Comparable Articles from Wikipedia and Measuring their Comparabilities

https://www.sciencedirect.com/science/article/pii/S1877042813041402

Arab

No

sourceforge

Free

nan

No

machine translation

PSBS

21.0

journal

Procedia Social and Behavioral Sciences

Motaz K. Saad,David Langlois,Kamel Sma

,,

Parallel corpora are not available for all domains and languages, but statistical methods in multilingual research domains require huge parallel/comparable corpora. Comparable corpora can be used when the parallel is not sufficient or not available for specific domains and languages. In this paper, we propose a method to extract all comparable articles from Wikipedia for multiple languages based on interlanguge links. We also extract comparable articles from Euro News website. We also present two comparability measures (CM) to compute the degree of comparability of multilingual articles. We extracted about 40K and 34K comparable articles from Wikipedia and Euro News respectively in three languages including Arabic, French, and English. Experimental results of comparability measures show that our measure can capture the comparability of multilingual corpora and allow to retrieve articles from different language concerning the same topic.

Zaid Alyafeai

Arabic Named Entities

https://huggingface.co/datasets/arbml/Arabic_Named_Entities

https://sourceforge.net/projects/arabicnes/

unknown

ar

ar-MSA: (Arabic (Modern Standard Arabic))

other

text

crawling and annotation(other)

we have extracted approximately 45,000 Arabic NE

45,000

tokens

Low

Multiple institutions

nan

An automatically built Named Entity lexicon for Arabic

http://doras.dcu.ie/15979/1/An_automatically_built_Named_Entity_lexicon_for_Arabic.pdf

Arab

No

sourceforge

Free

nan

No

named entity recognition

LREC

41.0

conference

International Conference on Language Resources and Evaluation

Mohammed Attia,Antonio Toral,L. Tounsi,M. Monachini,Josef van Genabith

,University of Groningen,,,

We have successfully adapted and extended the automatic Multilingual, Interoperable Named Entity Lexicon approach to Arabic, using Arabic WordNet (AWN) and Arabic Wikipedia (AWK). First, we extract AWN’s instantiable nouns and identify the corresponding categories and hyponym subcategories in AWK. Then, we exploit Wikipedia inter-lingual links to locate correspondences between articles in ten different languages in order to identify Named Entities (NEs). We apply keyword search on AWK abstracts to provide for Arabic articles that do not have a correspondence in any of the other languages. In addition, we perform a post-processing step to fetch further NEs from AWK not reachable through AWN. Finally, we investigate diacritization using matching with geonames databases, MADA-TOKAN tools and different heuristics for restoring vowel marks of Arabic NEs. Using this methodology, we have extracted approximately 45,000 Arabic NEs and built, to the best of our knowledge, the largest, most mature and well-structured Arabic NE lexical resource to date. We have stored and organised this lexicon following the Lexical Markup Framework (LMF) ISO standard. We conduct a quantitative and qualitative evaluation of the lexicon against a manually annotated gold standard and achieve precision scores from 95.83% (with 66.13% recall) to 99.31% (with 61.45% recall) according to different values of a threshold.

Zaid Alyafeai

AraFacts

https://huggingface.co/datasets/arbml/AraFacts

https://gitlab.com/bigirqu/AraFacts/

CC BY-NC 4.0

ar

mixed

other

text

crawling and annotation(other)

consists of 6,222 claims along with their factual labels and additional metadata, such as fact-checking article content, topical category, and links to posts or Web pages spreading the claim

6,222

sentences

Low

Qatar University

nan

AraFacts: The First Large Arabic Dataset of Naturally Occurring Claims

https://aclanthology.org/2021.wanlp-1.26.pdf

Arab

No

GitLab

Free

nan

No

topic classification, claim verification

WANLP

0.0

workshop

Arabic Natural Language Processing Workshop

Zien Sheikh Ali,Watheq Mansour,Tamer Elsayed,A. Al-Ali

,,,

We introduce AraFacts, the first large Arabic dataset of naturally occurring claims collected from 5 Arabic fact-checking websites, e.g., Fatabyyano and Misbar, and covering claims since 2016. Our dataset consists of 6,121 claims along with their factual labels and additional metadata, such as fact-checking article content, topical category, and links to posts or Web pages spreading the claim. Since the data is obtained from various fact-checking websites, we standardize the original claim labels to provide a unified label rating for all claims. Moreover, we provide revealing dataset statistics and motivate its use by suggesting possible research applications. The dataset is made publicly available for the research community.

Zaid Alyafeai

ASAD

nan

https://wti.kaust.edu.sa/solve/Arabic-Sentiment-Analysis-Challenge

unknown

ar

mixed

social media

text

crawling and annotation(other)

ASAD is a large, high-quality annotated dataset (including 95K tweets), with three-class sentiment labels (positive, negative and neutral)

100,000

sentences

Medium

Multiple institutions

nan

ASAD: A TWITTER-BASED BENCHMARK ARABIC SENTIMENT ANALYSIS DATASET

https://arxiv.org/pdf/2011.00578.pdf

Arab

No

other

Upon-Request

nan

Yes

sentiment analysis

ArXiv

2.0

preprint

ArXiv

Basma Alharbi,Hind Alamro,Manal Alshehri,Zuhair Khayyat,Manal Kalkatawi,I. Jaber,X. Zhang

,,,,,,

This paper provides a detailed description of a new Twitter-based benchmark dataset for Arabic Sentiment Analysis (ASAD), which is launched in a competition3, sponsored by KAUST for awarding 10000 USD, 5000 USD and 2000 USD to the first, second and third place winners, respectively. Compared to other publicly released Arabic datasets, ASAD is a large, high-quality annotated dataset(including 95K tweets), with three-class sentiment labels (positive, negative and neutral). We presents the details of the data collection process and annotation process. In addition, we implement several baseline models for the competition task and report the results as a reference for the participants to the competition.

Zaid Alyafeai

ASTAD

https://huggingface.co/datasets/arbml/Sentiment_Analysis_Tweets

https://github.com/motazsaad/arabic-sentiment-analysis

Apache-2.0

ar

mixed

social media

text

crawling and annotation(other)

It contains 58K Arabic tweets (47K training, 11K test) tweets annotated in positive and negative labels

58,000

sentences

Medium

Multiple institutions

nan

An Arabic Tweets Sentiment Analysis Dataset (ATSAD) using Distant Supervision and Self Training

https://aclanthology.org/2020.osact-1.1.pdf

Arab

No

GitHub

Free

nan

Yes

sentiment analysis

OSACT

4.0

workshop

Workshop on Open-Source Arabic Corpora and Processing Tools

Kathrein Abu Kwaik,S. Chatzikyriakidis,Simon Dobnik,Motaz Saad,Richard Johansson

,,,The Islamic University of Gaza,

As the number of social media users increases, they express their thoughts, needs, socialise and publish their opinions reviews. For good social media sentiment analysis, good quality resources are needed, and the lack of these resources is particularly evident for languages other than English, in particular Arabic. The available Arabic resources lack of from either the size of the corpus or the quality of the annotation. In this paper, we present an Arabic Sentiment Analysis Corpus collected from Twitter, which contains 36K tweets labelled into positive and negative. We employed distant supervision and self-training approaches into the corpus to annotate it. Besides, we release an 8K tweets manually annotated as a gold standard. We evaluated the corpus intrinsically by comparing it to human classification and pre-trained sentiment analysis models, Moreover, we apply extrinsic evaluation methods exploiting sentiment analysis task and achieve an accuracy of 86%.

Zaid Alyafeai

PAAD

https://huggingface.co/datasets/arbml/PAAD

https://data.mendeley.com/datasets/spvbf5bgjs/2

CC BY 4.0

ar

ar-MSA: (Arabic (Modern Standard Arabic))

news articles

text

crawling and annotation(other)

he dataset is 206 articles distributed into three categories as (Reform, Conservative and Revolutionary) that we offer to the research community on Arabiccomputational linguistics.

206

documents

Low

Multiple institutions

nan

PAAD:POLITICAL ARABIC ARTICLES DATASET FOR AUTOMATIC TEXT CATEGORIZATION

https://ijci.uoitc.edu.iq/index.php/ijci/article/view/246/174

Arab

No

Mendeley Data

Free

nan

No

topic classification

IJCI

3.0

journal

Iraqi journal of computer and informatics

Dhafar Hamed Abd,Ahmed T. Sadiq,Ayad R. Abbas

,,

Now day’s text Classification and Sentiment analysis is considered as one of the popular Natural Language Processing (NLP) tasks. This kind of technique plays significant role in human activities and has impact on the daily behaviours. Each article in different fields such as politics and business represent different opinions according to the writer tendency. A huge amount of data will be acquired through that differentiation. The capability to manage the political orientation of an online article automatically. Therefore, there is no corpus for political categorization was directed towards this task in Arabic, due to the lack of rich representative resources for training an Arabic text classifier. However, we introduce political Arabic articles dataset (PAAD) of textual data collected from newspapers, social network, general forum and ideology website. The dataset is 206 articles distributed into three categories as (Reform, Conservative and Revolutionary) that we offer to the research community on Arabic computational linguistics. We anticipate that this dataset would make a great aid for a variety of NLP tasks on Modern Standard Arabic, political text classification purposes. We present the data in raw form and excel file. Excel file will be in four types such as V1 raw data, V2 preprocessing, V3 root stemming and V4 light stemming.

Zaid Alyafeai

NLP dataset for Arabic dialects

nan

https://msda.um6p.ma/msda_datasets

unknown

ar

mixed

social media

text

crawling and annotation(other)

This data was collected from the Twitter social network and consists on +50K twits in five (5) national dialects

52,210

sentences

Medium

Multiple institutions

nan

An open access NLP dataset for Arabic dialects : data collection, labeling, and model construction

https://arxiv.org/pdf/2102.11000.pdf

Arab

No

other

Free

nan

No

dialect identification, topic classification, sentiment analysis

ArXiv

0.0

preprint

ArXiv

ElMehdi Boujou,Hamza Chataoui,Abdellah El Mekki,Saad Benjelloun,I. Chairi,Ismail Berrada

,,Mohammed VI Polytechnic University,,,

Natural Language Processing (NLP) is today a very active field of research and innovation. Many applications need however big sets of data for supervised learning, suitably labelled for the training purpose. This includes applications for the Arabic language and its national dialects. However, such open access labeled data sets in Arabic and its dialects are lacking in the Data Science ecosystem and this lack can be a burden to innovation and research in this field. In this work, we present an open data set of social data content in several Arabic dialects. This data was collected from the Twitter social network and consists on +50K twits in five (5) national dialects. Furthermore, this data was labeled for several applications, namely dialect detection, topic detection and sentiment analysis. We publish this data as an open access data to encourage innovation and encourage other works in the field of NLP for Arabic dialects and social media. A selection of models were built using this data set and are presented in this paper along with their performances.

nan

LK-Hadith-Corpus

https://huggingface.co/datasets/arbml/LK_Hadith

https://github.com/ShathaTm/LK-Hadith-Corpus

unknown

multilingual

ar-CLS: (Arabic (Classic))

books

text

other

a bilingual parallel corpus of Islamic Hadith

39,038

sentences

Low

King Saud University

nan

The Arabic–English Parallel Corpus of Authentic Hadith

https://www.researchgate.net/publication/341359917_The_Arabic-English_Parallel_Corpus_of_Authentic_Hadith

Arab-Latn

Yes

GitHub

Free

nan

No

machine translation

IJASAT

nan

journal

International Journal on Islamic Applications in Computer Science And Technologies

nan

nan

We present a bilingual parallel corpus of Islamic Hadith, which is the set of narratives reporting different aspects of the prophet Muhammad's life. The Hadith collection is extracted from the six canonical Hadith books which possess unique linguistic features and patterns that are automatically extracted and annotated using a domain-specific tool for Hadith segmentation. In this article, we present the methodology of creating the corpus of 39,038 annotated Hadiths which will be freely available for the research community.

Mustafa Ghaleb

BBN Blog Posts Sentiment Corpus

https://huggingface.co/datasets/arbml/BBN_Blog_Posts

https://saifmohammad.com/WebPages/ArabicSA.html

unknown

ar

ar-LEV: (Arabic(Levant))

social media

text

crawling and annotation(other)

A random subset of 1200 Levantine dialectal sentences chosen from the BBN Arabic-Dialect–English Parallel Text

1,200

sentences

Medium

Multiple institutions

nan

Sentiment after Translation: A Case-Study on Arabic Social Media Posts

https://aclanthology.org/N15-1078.pdf

Arab

No

other

Free

nan

No

sentiment analysis,machine translation

NAACL

125.0

conference

North American Chapter of the Association for Computational Linguistics

Mohammad Salameh,Saif M. Mohammad,Svetlana Kiritchenko

,National Research Council Canada,

When text is translated from one language into another, sentiment is preserved to varying degrees. In this paper, we use Arabic social media posts as stand-in for source language text, and determine loss in sentiment predictability when they are translated into English, manually and automatically. As benchmarks, we use manually and automatically determined sentiment labels of the Arabic texts. We show that sentiment analysis of English translations of Arabic texts produces competitive results, w.r.t. Arabic sentiment analysis. We discover that even though translation significantly reduces the human ability to recover sentiment, automatic sentiment systems are still able to capture sentiment information from the translations.

Zaid Alyafeai

Syria Tweets Sentiment Corpus

https://huggingface.co/datasets/arbml/Syria_Tweet_Sentiment

https://saifmohammad.com/WebPages/ArabicSA.html

unknown

ar

ar-SY: (Arabic (Syria))

social media

text

crawling and annotation(other)

A dataset of 2000 tweets originating from Syria

2,000

sentences

Medium

Multiple institutions

nan

Sentiment after Translation: A Case-Study on Arabic Social Media Posts

https://aclanthology.org/N15-1078.pdf

Arab

No

other

Free

nan

No

sentiment analysis,machine translation

NAACL

125.0

conference

North American Chapter of the Association for Computational Linguistics

Mohammad Salameh,Saif M. Mohammad,Svetlana Kiritchenko

,National Research Council Canada,

When text is translated from one language into another, sentiment is preserved to varying degrees. In this paper, we use Arabic social media posts as stand-in for source language text, and determine loss in sentiment predictability when they are translated into English, manually and automatically. As benchmarks, we use manually and automatically determined sentiment labels of the Arabic texts. We show that sentiment analysis of English translations of Arabic texts produces competitive results, w.r.t. Arabic sentiment analysis. We discover that even though translation significantly reduces the human ability to recover sentiment, automatic sentiment systems are still able to capture sentiment information from the translations.

Zaid Alyafeai

Coronavirus

nan

https://github.com/aseelad/Coronavirus-Public-Arabic-Twitter-Data-Set/

CC BY-NC-SA 4.0

ar

mixed

social media

text

other

contains data collected from December 1st 2019 until April 11th 2020

707,829

sentences

Low

Imam Mohammad Bin Saud University

nan

Coronavirus: Public Arabic Twitter Dataset

https://www.preprints.org/manuscript/202004.0263/v1

Arab

No

GitHub

Free

nan

No

text generation, language modeling

other

3.0

preprint

nan

Aseel Addawood

nan

The COVID-19 pandemic spread of the coronavirus across the globe has affected our lives on many different levels. The world we knew before the spread of the virus has become another one. Every country has taken preventive measures, including social distancing, travel restrictions, and curfew, to control the spread of the disease. With these measures implemented, people have shifted to social media platforms in the online sphere, such as Twitter, to maintain connections. In this paper, we describe a coronavirus data set of Arabic tweets collected from January 1, 2020, primarily from hashtags populated from Saudi Arabia. This data set is available to the research community to glean a better understanding of the societal, economical, and political effects of the outbreak and to help policy makers make better decisions for fighting this epidemic.

Zaid Alyafeai

Botta

nan

https://camel.abudhabi.nyu.edu/botta/

custom

ar

ar-EG: (Arabic (Egypt))

other

text

other

files available for chatbot systems

nan

documents

Low

NYU Abu Dhabi

nan

Botta: An Arabic Dialect Chatbot

https://aclanthology.org/C16-2044.pdf

Arab

No

CAMeL Resources

Upon-Request

nan

No

dialogue generation

COLING

14.0

conference

International Conference on Computational Linguistics

Dana Abu Ali,Nizar Habash

,

This paper presents BOTTA, the first Arabic dialect chatbot. We explore the challenges of creating a conversational agent that aims to simulate friendly conversations using the Egyptian Arabic dialect. We present a number of solutions and describe the different components of the BOTTA chatbot. The BOTTA database files are publicly available for researchers working on Arabic chatbot technologies. The BOTTA chatbot is also publicly available for any users who want to chat with it online.

Zaid Alyafeai

AR-ASAG

nan

https://data.mendeley.com/datasets/dj95jh332j/1

CC BY 4.0

ar

ar-MSA: (Arabic (Modern Standard Arabic))

other

text

other

reported evaluations relate to answers submitted for three different exams submitted to three classes of students.

2,133

sentences

Low

Bouira University

nan

AR-ASAG An ARabic Dataset for Automatic Short Answer Grading Evaluation

https://aclanthology.org/2020.lrec-1.321.pdf

Arab

No

Mendeley Data

Free

nan

No

answer grading evaluation

LREC

0.0

conference

International Conference on Language Resources and Evaluation

Leila Ouahrani,Djamel Bennouar

,

Automatic short answer grading is a significant problem in E-assessment. Several models have been proposed to deal with it. Evaluation and comparison of such solutions need the availability of Datasets with manual examples. In this paper, we introduce AR-ASAG, an Arabic Dataset for automatic short answer grading. The Dataset contains 2133 pairs of (Model Answer, Student Answer) in several versions (txt, xml, Moodle xml and .db). We explore then an unsupervised corpus based approach for automatic grading adapted to the Arabic Language. We use COALS (Correlated Occurrence Analogue to Lexical Semantic) algorithm to create semantic space for word distribution. The summation vector model is combined to term weighting and common words to achieve similarity between a teacher model answer and a student answer. The approach is particularly suitable for languages with scarce resources such as Arabic language where robust specific resources are not yet available. A set of experiments were conducted to analyze the effect of domain specificity, semantic space dimension and stemming techniques on the effectiveness of the grading model. The proposed approach gives promising results for Arabic language. The reported results may serve as baseline for future research work evaluation

Zaid Alyafeai

DAICT

nan

https://www.hbku.edu.qa/en/DAICT

custom

ar

mixed

social media

text

crawling and annotation(other)

The dataset includes 5,588 tweets -- written in both MSA and dialectual Arabic -- manually annotated by two professional linguistics from HBKU

5,588

sentences

Medium

Hamad Bin Khalifa University

nan

DAICT: A Dialectal Arabic Irony Corpus Extracted from Twitter

http://www.lrec-conf.org/proceedings/lrec2020/pdf/2020.lrec-1.768.pdf

Arab

No

other

Free

nan

No

irony detection

LREC

15.0

conference

International Conference on Language Resources and Evaluation

Ines Abbes,W. Zaghouani,Omaima El-Hardlo,Faten Ashour

,,,

Identifying irony in user-generated social media content has a wide range of applications; however to date Arabic content has received limited attention. To bridge this gap, this study builds a new open domain Arabic corpus annotated for irony detection. We query Twitter using irony-related hashtags to collect ironic messages, which are then manually annotated by two linguists according to our working definition of irony. Challenges which we have encountered during the annotation process reflect the inherent limitations of Twitter messages interpretation, as well as the complexity of Arabic and its dialects. Once published, our corpus will be a valuable free resource for developing open domain systems for automatic irony recognition in Arabic language and its dialects in social media text.

Zaid Alyafeai

WAW

nan

https://alt.qcri.org/resources/wawcorpus/

custom

ar

ar-MSA: (Arabic (Modern Standard Arabic))

transcribed audio

spoken

human translation

an interpreting corpus for English/Arabic

31

hours

Low

QCRI,Hamad Bin Khalifa University

nan

The WAW Corpus: The First Corpus of Interpreted Speeches and their Translations for English and Arabic

https://alt.qcri.org/resources/wawcorpus/

Arab-Latn

No

QCRI Resources

Free

nan

No

speech recognition, machine translation

LREC

1.0

conference

International Conference on Language Resources and Evaluation

Ahmed Abdelali,Irina Temnikova,S. Hedaya,S. Vogel

,University of Wolverhampton, Bulgarian Academy of Sciences, Qatar Computing Research Institute, Mitra Translations,,

This article presents the WAW Corpus, an interpreting corpus for English/Arabic, which can be used for teaching interpreters, studying the characteristics of interpreters’ work, as well as to train machine translation systems. The corpus contains recordings of lectures and speeches from international conferences, their interpretations, the transcripts of the original speeches and of their interpretations, as well as human translations of both kinds of transcripts into the opposite language of the language pair. The article presents the corpus curation, statistics, assessment, as well as a case study of the corpus use.

Zaid Alyafeai

SenZi

https://huggingface.co/datasets/arbml/SenZi

https://tahatobaili.github.io/project-rbz/

custom

ar

ar-LB: (Arabic (Lebanon))

social media

text

crawling and annotation(other)

by translating, annotating, and transliterating other resources to have an initial set of 2K sentiment words. We expanded it to 24.6K sentiment words by importing inflectional and orthographic forms using word embeddings

24,600

tokens

Medium

Multiple institutions

nan

SenZi: A Sentiment Analysis Lexicon for the Latinised Arabic (Arabizi)

https://aclanthology.org/R19-1138.pdf

Arab-Latn

No

other

Free

nan

No

sentiment analysis, transliteration

RANLP

5.0

conference

Recent Advances in Natural Language Processing

Taha Tobaili,Miriam Fernández,Harith Alani,S. Sharafeddine,Hazem M. Hajj,Goran Glavas

,,,,,

Arabizi is an informal written form of dialectal Arabic transcribed in Latin alphanumeric characters. It has a proven popularity on chat platforms and social media, yet it suffers from a severe lack of natural language processing (NLP) resources. As such, texts written in Arabizi are often disregarded in sentiment analysis tasks for Arabic. In this paper we describe the creation of a sentiment lexicon for Arabizi that was enriched with word embeddings. The result is a new Arabizi lexicon consisting of 11.3K positive and 13.3K negative words. We evaluated this lexicon by classifying the sentiment of Arabizi tweets achieving an F1-score of 0.72. We provide a detailed error analysis to present the challenges that impact the sentiment analysis of Arabizi.

Zaid Alyafeai

ArSAS

https://huggingface.co/datasets/arbml/ArSAS

https://homepages.inf.ed.ac.uk/wmagdy/resources.htm

unknown

ar

mixed

social media

text

crawling and annotation(other)

A set of 21K Arabic tweets labeled for 4 classes of sentiment and 6 classes of speech-act

21,000

sentences

Medium

Multiple institutions

nan

ArSAS: An Arabic Speech-Act and Sentiment Corpus of Tweets

http://lrec-conf.org/workshops/lrec2018/W30/pdf/22_W30.pdf

Arab

No

other

Free

nan

No

sentiment analysis, speech act classification

LREC

25.0

conference

International Conference on Language Resources and Evaluation

AbdelRahim Elmadany,Hamdy Mubarak,Walid Magdy

University of British Columbia,,The University of Edinburgh

Speech acts are the type of communicative acts within a conversation. Speech act recognition (aka classification) has been an active research in recent years. However, much less attention was directed towards this task in Arabic due to the lack of resources for training an Arabic speech-act classifier. In this paper we present ArSAS , an Arabic corpus of tweets annotated for the tasks of speech-act recognition and sentiment analysis. A large set of 21k Arabic tweets covering multiple topics were collected, prepared and annotated for six different classes of speech-act labels, such as expression, assertion, and question. In addition, the same set of tweets were also annotated with four classes of sentiment. We aim to have this corpus promoting the research in both speech-act recognition and sentiment analysis tasks for Arabic language.

Zaid Alyafeai

ANS CORPUS: claim verification

https://huggingface.co/datasets/arbml/ANS_stance

https://github.com/latynt/ans

Apache-2.0

ar

ar-MSA: (Arabic (Modern Standard Arabic))

news articles

text

crawling and annotation(other)

corpus comes in two perspectives: a version consisting of 4,547 true and false claims and a version consisting of 3,786 pairs (claim, evidence).

4,547

sentences

Low

Latynt

nan

Stance Prediction and Claim Verification: An Arabic Perspective

https://arxiv.org/pdf/2005.10410.pdf

Arab

No

GitHub

Free

nan

Yes

stance detection, claim verification

FEVER

8.0

workshop

Workshop on Fact Extraction and Verification

Jude Khouja

nan

This work explores the application of textual entailment in news claim verification and stance prediction using a new corpus in Arabic. The publicly available corpus comes in two perspectives: a version consisting of 4,547 true and false claims and a version consisting of 3,786 pairs (claim, evidence). We describe the methodology for creating the corpus and the annotation process. Using the introduced corpus, we also develop two machine learning baselines for two proposed tasks: claim verification and stance prediction. Our best model utilizes pretraining (BERT) and achieves 76.7 F1 on the stance prediction task and 64.3 F1 on the claim verification task. Our preliminary experiments shed some light on the limits of automatic claim verification that relies on claims text only. Results hint that while the linguistic features and world knowledge learned during pretraining are useful for stance prediction, such learned representations from pretraining are insufficient for verifying claims without access to context or evidence.

Mustafa Ghaleb

OntoNotes 5.0

https://huggingface.co/datasets/conll2012_ontonotesv5

https://catalog.ldc.upenn.edu/LDC2013T19

LDC User Agreement

multilingual

ar-MSA: (Arabic (Modern Standard Arabic))

news articles

text

crawling and annotation(other)

The Arabic portion of OntoNotes 5.0 includes 300K words of newswire data.

300,000

tokens

Low

Multiple institutions

nan

OntoNotes Release 5.0 with OntoNotes DB Tool v0.999 beta

https://catalog.ldc.upenn.edu/docs/LDC2013T19/OntoNotes-Release-5.0.pdf

Arab

Yes

LDC

Upon-Request

nan

No

machine translation

other

nan

preprint

nan

nan

nan

nan

Mustafa Ghaleb

Arabic-Dialect/English Parallel Text

nan

https://catalog.ldc.upenn.edu/LDC2012T09

LDC User Agreement

multilingual

ar-LEV: (Arabic(Levant))

other

text

crawling and annotation(translation)

it uses crowdsourcing to cheaply and quickly build LevantineEnglish and Egyptian-English parallel corpora, consisting of 1.1M words and 380k words, respectively.

1,500,000

tokens

Low

Multiple institutions

nan

Machine Translation of Arabic Dialects

https://aclanthology.org/N12-1006.pdf

Arab-Latn

Yes

LDC

With-Fee

2250 $

No

machine translation

NAACL

158.0

conference

North American Chapter of the Association for Computational Linguistics

Rabih Zbib,Erika Malchiodi,J. Devlin,D. Stallard,S. Matsoukas,R. Schwartz,J. Makhoul,Omar Zaidan,Chris Callison-Burch

,,,,,,,,

Arabic Dialects present many challenges for machine translation, not least of which is the lack of data resources. We use crowdsourcing to cheaply and quickly build Levantine-English and Egyptian-English parallel corpora, consisting of 1.1M words and 380k words, respectively. The dialectal sentences are selected from a large corpus of Arabic web text, and translated using Amazon's Mechanical Turk. We use this data to build Dialectal Arabic MT systems, and find that small amounts of dialectal data have a dramatic impact on translation quality. When translating Egyptian and Levantine test sets, our Dialectal Arabic MT system performs 6.3 and 7.0 BLEU points higher than a Modern Standard Arabic MT system trained on a 150M-word Arabic-English parallel corpus.

Mustafa Ghaleb

FLORES-101

https://huggingface.co/datasets/gsarti/flores_101

https://github.com/facebookresearch/flores/tree/master/floresv1/data

CC BY-SA 4.0

multilingual

ar-MSA: (Arabic (Modern Standard Arabic))

other

text

crawling and annotation(other)

Low Resource MT Benchmark

3,100,000

tokens

Low

Facebook

OPTUS

The FLORES-101 Evaluation Benchmark for Low-Resource and Multilingual Machine Translation

https://arxiv.org/pdf/2106.03193.pdf

Arab

Yes

GitHub

Free

nan

Yes

machine translation

ArXiv

1.0

preprint

ArXiv

Naman Goyal,Cynthia Gao,Vishrav Chaudhary,Guillaume Wenzek,Da Ju,Sanjan Krishnan,Marc'Aurelio Ranzato,Francisco Guzmán,Angela Fan

,,,,Facebook AI Research,,,,

One of the biggest challenges hindering progress in low-resource and multilingual machine translation is the lack of good evaluation benchmarks. Current evaluation benchmarks either lack good coverage of low-resource languages, consider only restricted domains, or are low quality because they are constructed using semi-automatic procedures. In this work, we introduce the FLORES-101 evaluation benchmark, consisting of 3001 sentences extracted from English Wikipedia and covering a variety of different topics and domains. These sentences have been translated in 101 languages by professional translators through a carefully controlled process. The resulting dataset enables better assessment of model quality on the long tail of low-resource languages, including the evaluation of many-to-many multilingual translation systems, as all translations are multilingually aligned. By publicly releasing such a highquality and high-coverage dataset, we hope to foster progress in the machine translation community and beyond.

Maraim Masoud

GEM

https://huggingface.co/datasets/gem/xlsumm

https://gem-benchmark.com/

Apache-2.0

multilingual

ar-MSA: (Arabic (Modern Standard Arabic))

wikipedia

text

crawling and annotation(other)

benchmark environment for Natural Language Generation with a focus on its Evaluation, both through human annotations and automated Metric

29,229

documents

Low

Multiple institutions

MLSUM, XSUM, Wikilingua,WebNLG, CommonGen, E2E,DART, Czech Restaurant, ToTTo, wiki-Auto,TirkCorpus, ASSET, Schema-Guided Dialog

The GEM Benchmark: Natural Language Generation, its Evaluation and Metrics

https://aclanthology.org/2021.gem-1.10.pdf

Arab

No

other

Free

nan

Yes

summarization

GEM

27.0

workshop

Generation Evaluation and Metrics

Sebastian Gehrmann,Tosin P. Adewumi,Karmanya Aggarwal,Pawan Sasanka Ammanamanchi,Aremu Anuoluwapo,Antoine Bosselut,Khyathi Raghavi Chandu,Miruna Adriana Clinciu,Dipanjan Das,Kaustubh D. Dhole,Wanyu Du,Esin Durmus,Ondrej Dusek,Chris C. Emezue,Varun Gangal,Cristina Garbacea,Tatsunori B. Hashimoto,Yufang Hou,Yacine Jernite,Harsh Jhamtani,Yangfeng Ji,Shailza Jolly,Mihir Kale,Dhruv Kumar,Faisal Ladhak,Aman Madaan,Mounica Maddela,Khyati Mahajan,Saad Mahamood,Bodhisattwa Prasad Majumder,Pedro Henrique Martins,Angelina McMillan-Major,Simon Mille,Emiel van Miltenburg,Moin Nadeem,Shashi Narayan,Vitaly Nikolaev,Rubungo Andre Niyongabo,Salomey Osei,Ankur P. Parikh,Laura Perez-Beltrachini,Niranjan Rao,Vikas Raunak,Juan Diego Rodríguez,Sashank Santhanam,João Sedoc,Thibault Sellam,Samira Shaikh,Anastasia Shimorina,Marco Antonio Sobrevilla Cabezudo,Hendrik Strobelt,Nishant Subramani,W. Xu,Diyi Yang,Akhila Yerukola,Jiawei Zhou

,,,,,EPFL,,Edinburgh Centre for Robotics,,,,Stanford University,Charles University,,Carnegie Mellon University,,,,FAIR,Carnegie Mellon University,University of Virginia,,,,,,,,,University of California San Diego,,,,Tilburg University,Massachusetts Institute of Technology;MIT,,,,,,,,,The University of Texas at Austin,,,,,,Institute of Mathematics and Computer Sciences;University of São Paulo;Pontifical Catholic University of Peru,,Allen Institute for AI;Masakhane,,,,

We introduce GEM, a living benchmark for natural language Generation (NLG), its Evaluation, and Metrics. Measuring progress in NLG relies on a constantly evolving ecosystem of automated metrics, datasets, and human evaluation standards. Due to this moving target, new models often still evaluate on divergent anglo-centric corpora with well-established, but flawed, metrics. This disconnect makes it challenging to identify the limitations of current models and opportunities for progress. Addressing this limitation, GEM provides an environment in which models can easily be applied to a wide set of tasks and in which evaluation strategies can be tested. Regular updates to the benchmark will help NLG research become more multilingual and evolve the challenge alongside models. This paper serves as the description of the data for the 2021 shared task at the associated GEM Workshop.

Maraim Masoud

GEM - XLSum

https://huggingface.co/datasets/GEM/xlsum

https://github.com/csebuetnlp/xl-sum

CC BY-NC-SA 4.0

multilingual

ar-MSA: (Arabic (Modern Standard Arabic))

news articles

text

crawling and annotation(other)

Large-Scale Multilingual Abstractive Summarization for 44 Languages"

46,897

sentences

Low

Multiple institutions

nan

XL-Sum: Large-Scale Multilingual Abstractive Summarization for 44 Languages

https://aclanthology.org/2021.findings-acl.413.pdf

Arab

No

other

Free

nan

Yes

summarization

FINDINGS

0.0

conference

Findings of the Association for Computational Linguistics

Tahmid Hasan,Abhik Bhattacharjee,Md. Saiful Islam,Kazi Samin,Yuan-Fang Li,Yong-Bin Kang,M. Rahman,Rifat Shahriyar

,,,,,,,

Contemporary works on abstractive text summarization have focused primarily on highresource languages like English, mostly due to the limited availability of datasets for low/midresource ones. In this work, we present XLSum, a comprehensive and diverse dataset comprising 1 million professionally annotated article-summary pairs from BBC, extracted using a set of carefully designed heuristics. The dataset covers 44 languages ranging from low to high-resource, for many of which no public dataset is currently available. XL-Sum is highly abstractive, concise, and of high quality, as indicated by human and intrinsic evaluation. We fine-tune mT5, a state-of-theart pretrained multilingual model, with XLSum and experiment on multilingual and lowresource summarization tasks. XL-Sum induces competitive results compared to the ones obtained using similar monolingual datasets: we show higher than 11 ROUGE-2 scores on 10 languages we benchmark on, with some of them exceeding 15, as obtained by multilingual training. Additionally, training on low-resource languages individually also provides competitive performance. To the best of our knowledge, XL-Sum is the largest abstractive summarization dataset in terms of the number of samples collected from a single source and the number of languages covered. We are releasing our dataset and models to encourage future research on multilingual abstractive summarization. The resources can be found at https://github. com/csebuetnlp/xl-sum.

Maraim Masoud

GEM - WikiLingua

https://huggingface.co/datasets/wiki_lingua

https://github.com/esdurmus/Wikilingua

CC0

multilingual

ar-MSA: (Arabic (Modern Standard Arabic))

wikipedia

text

crawling and annotation(other)

New Benchmark Dataset for Multilingual Abstractive Summarization

29,229

sentences

Low

Multiple institutions

nan

WikiLingua: A New Benchmark Dataset for Cross-Lingual Abstractive Summarization

WikiLingua: A New Benchmark Dataset for Cross-Lingual Abstractive Summarization

Arab

No

GitHub

Free

nan

Yes

summarization

EMNLP

9.0

conference

Conference on Empirical Methods in Natural Language Processing

Faisal Ladhak,Esin Durmus,Claire Cardie,K. McKeown

,Stanford University,,

We introduce WikiLingua, a large-scale, multilingual dataset for the evaluation of crosslingual abstractive summarization systems. We extract article and summary pairs in 18 languages from WikiHow, a high quality, collaborative resource of how-to guides on a diverse set of topics written by human authors. We create gold-standard article-summary alignments across languages by aligning the images that are used to describe each how-to step in an article. As a set of baselines for further studies, we evaluate the performance of existing cross-lingual abstractive summarization methods on our dataset. We further propose a method for direct crosslingual summarization (i.e., without requiring translation at inference time) by leveraging synthetic data and Neural Machine Translation as a pre-training step. Our method significantly outperforms the baseline approaches, while being more cost efficient during inference.

Maraim Masoud

UDP (UDP-NYUAD)

nan

https://github.com/UniversalDependencies/UD_Arabic-NYUAD

CC BY-SA 4.0

ar

ar-MSA: (Arabic (Modern Standard Arabic))

other

text

manual curation

The treebank consists of 19,738 sentences (738889 tokens), and its domain is mainly newswire.

738,889

tokens

Low

Multiple institutions, NYUAD

The NYUAD Arabic UD treebank is based on the Penn Arabic Treebank (PATB), parts 1, 2, and 3, through conversion to CATiB dependency trees.

Universal Dependencies for Arabic

https://aclanthology.org/W17-1320.pdf

Arab

Yes

GitHub

Free

nan

Yes

morphological attribute tagging, dependency parsing, part of speech tagging

WANLP

12.0

workshop

Arabic Natural Language Processing Workshop

Dima Taji,Nizar Habash,Daniel Zeman

,,

We describe the process of creating NUDAR, a Universal Dependency treebank for Arabic. We present the conversion from the Penn Arabic Treebank to the Universal Dependency syntactic representation through an intermediate dependency representation. We discuss the challenges faced in the conversion of the trees, the decisions we made to solve them, and the validation of our conversion. We also present initial parsing results on NUDAR.

Maraim Masoud

WikiANN NER(MMNER)

nan

https://github.com/afshinrahimi/mmner

unknown

multilingual

ar-MSA: (Arabic (Modern Standard Arabic))

wikipedia

text

crawling and annotation(other)

Cross-lingual name tagging and linking for 282 languages

30,000

tokens

Low

The university of Melbourne

nan

Massively Multilingual Transfer for NER

https://aclanthology.org/P19-1015.pdf

Arab

Yes

GitHub

Free

nan

Yes

named entity recognition

ACL

54.0

conference

Assofications of computation linguisitcs

Afshin Rahimi,Yuan Li,Trevor Cohn

University of Melbourne,,University of Melbourne

In cross-lingual transfer, NLP models over one or more source languages are applied to a low-resource target language. While most prior work has used a single source model or a few carefully selected models, here we consider a “massive” setting with many such models. This setting raises the problem of poor transfer, particularly from distant languages. We propose two techniques for modulating the transfer, suitable for zero-shot or few-shot learning, respectively. Evaluating on named entity recognition, we show that our techniques are much more effective than strong baselines, including standard ensembling, and our unsupervised method rivals oracle selection of the single best individual model.

Maraim Masoud

WinoMT (MT_Gender)

https://huggingface.co/datasets/arbml/mt_gender_ar

https://github.com/gabrielStanovsky/mt_gender

MIT License

multilingual

ar-MSA: (Arabic (Modern Standard Arabic))

other

text

machine translation

Evaluating Gender Bias in Machine Translation

3,888

sentences

Low

Multiple Institutions

nan

Evaluating Gender Bias in Machine Translation

https://arxiv.org/pdf/1906.00591.pdf

Arab

No

GitHub

Free

nan

Yes

machine translation

ACL

99.0

conference

Assofications of computation linguisitcs

Gabriel Stanovsky,Noah A. Smith,Luke Zettlemoyer

Allen Institute for Artificial Intelligence;University of Washington,,University of Washington;Facebook

We present the first challenge set and evaluation protocol for the analysis of gender bias in machine translation (MT). Our approach uses two recent coreference resolution datasets composed of English sentences which cast participants into non-stereotypical gender roles (e.g., “The doctor asked the nurse to help her in the operation”). We devise an automatic gender bias evaluation method for eight target languages with grammatical gender, based on morphological analysis (e.g., the use of female inflection for the word “doctor”). Our analyses show that four popular industrial MT systems and two recent state-of-the-art academic MT models are significantly prone to gender-biased translation errors for all tested target languages. Our data and code are publicly available at https://github.com/gabrielStanovsky/mt_gender.

Maraim Masoud

LinCE - MSA-EGY (NER - Code Switching)

nan

https://ritual.uh.edu/lince/datasets

unknown

ar

mixed

social media

text

crawling and annotation(other)

Modern Standard ArabicDialectal Arabic (MSA-DA)

11,224

sentences

Medium

Multiple institutions

nan

Named Entity Recognition on Code-Switched Data: Overview of the CALCS 2018 Shared Task

https://aclanthology.org/W18-3219.pdf

Arab-Latn

No

other

Upon-Request

nan

Yes

named entity recognition

CALCS

39.0

workshop

Workshop on Computational Approaches to Code Switching

Gustavo Aguilar,F. Alghamdi,Víctor Soto,Mona T. Diab,Julia Hirschberg,T. Solorio

University of Houston,,,,,

In the third shared task of the Computational Approaches to Linguistic Code-Switching (CALCS) workshop, we focus on Named Entity Recognition (NER) on code-switched social-media data. We divide the shared task into two competitions based on the English-Spanish (ENG-SPA) and Modern Standard Arabic-Egyptian (MSA-EGY) language pairs. We use Twitter data and 9 entity types to establish a new dataset for code-switched NER benchmarks. In addition to the CS phenomenon, the diversity of the entities and the social media challenges make the task considerably hard to process. As a result, the best scores of the competitions are 63.76% and 71.61% for ENG-SPA and MSA-EGY, respectively. We present the scores of 9 participants and discuss the most common challenges among submissions.

Maraim Masoud

LinCE - MSA-DA (LID - Code Switching )

nan

https://ritual.uh.edu/lince/datasets

unknown

ar

mixed

social media

text

crawling and annotation(other)

Twitter data and 9 entity types to establish a new dataset for code-switched NER benchmarks.

11,241

sentences

Medium

Multiple Institutions

nan

Overview for the Second Shared Task on Language Identification in Code-Switched Data

https://aclanthology.org/W16-5805.pdf

Arab-Latn

No

other

Upon-Request

nan

Yes

code switch detection

CALCS

98.0

workshop

Workshop on Computational Approaches to Code Switching

Giovanni Molina,F. Alghamdi,Mahmoud A. Ghoneim,A. Hawwari,Nicolas Rey-Villamizar,Mona T. Diab,T. Solorio

,,,,,,

We present an overview of the first shared task on language identification on codeswitched data. The shared task included code-switched data from four language pairs: Modern Standard ArabicDialectal Arabic (MSA-DA), MandarinEnglish (MAN-EN), Nepali-English (NEPEN), and Spanish-English (SPA-EN). A total of seven teams participated in the task and submitted 42 system runs. The evaluation showed that language identification at the token level is more difficult when the languages present are closely related, as in the case of MSA-DA, where the prediction performance was the lowest among all language pairs. In contrast, the language pairs with the higest F-measure where SPA-EN and NEP-EN. The task made evident that language identification in code-switched data is still far from solved and warrants further research.

Maraim Masoud

AutoTweet

nan

https://sites.google.com/view/bigir/resources/datasets#h.p_sSqR81gM9CWFF

unknown

ar

mixed

social media

text

crawling and annotation(other)

classification of Arabic tweets into automated or manual.

3,503

sentences

Medium

Qatar University

nan

Detecting Automatically-Generated Arabic Tweets

https://link.springer.com/chapter/10.1007/978-3-319-28940-3_10

Arab

No

Dropbox

Free

nan

No

generated text detection

AIRS

10.0

conference

Conference on Alliance of Information and Referral Systems

H. Almerekhi,T. Elsayed

,

Recently, Twitter, one of the most widely-known social media platforms, got infiltrated by several automation programs, commonly known as “bots”. Bots can be easily abused to spread spam and hinder information extraction applications by posting lots of automatically-generated tweets that occupy a good portion of the continuous stream of tweets. This problem heavily affects users in the Arab region due to the recent developing political events as automated tweets can disturb communication and waste time needed in filtering such tweets.

Zaid Alyafeai

Arabic SMS/Chat

nan

https://catalog.ldc.upenn.edu/LDC2017T07

LDC User Agreement

ar

mixed

other

text

other

5,691 conversations totaling 1,029,248 words across 262,026 messages. Messages were natively written in either Arabic orthography or romanized Arabizi

262,026

sentences

Low

LDC

nan

Transliteration of Arabizi into Arabic Orthography: Developing a Parallel Annotated Arabizi-Arabic Script SMS/Chat Corpus

https://www.ldc.upenn.edu/sites/www.ldc.upenn.edu/files/emnlp2014-transliteration-aribizi-into-arabic-orthography.pdf

Arab-Latn

No

LDC

With-Fee

1750 $

No

transliteration, machine translation

WANLP

39.0

workshop

Arabic Natural Language Processing workshop

Ann Bies,Zhiyi Song,M. Maamouri,Stephen Grimes,Haejoong Lee,Jonathan Wright,S. Strassel,Nizar Habash,R. Eskander,Owen Rambow

,,,,,,,,,

This paper describes the process of creating a novel resource, a parallel Arabizi-Arabic script corpus of SMS/Chat data. The language used in social media expresses many differences from other written genres: its vocabulary is informal with intentional deviations from standard orthography such as repeated letters for emphasis; typos and nonstandard abbreviations are common; and nonlinguistic content is written out, such as laughter, sound representations, and emoticons. This situation is exacerbated in the case of Arabic social media for two reasons. First, Arabic dialects, commonly used in social media, are quite different from Modern Standard Arabic phonologically, morphologically and lexically, and most importantly, they lack standard orthographies. Second, Arabic speakers in social media as well as discussion forums, SMS messaging and online chat often use a non-standard romanization called Arabizi. In the context of natural language processing of social media Arabic, transliterating from Arabizi of various dialects to Arabic script is a necessary step, since many of the existing state-of-the-art resources for Arabic dialect processing expect Arabic script input. The corpus described in this paper is expected to support Arabic NLP by providing this resource.

Zaid Alyafeai

AraNews

nan

https://github.com/UBC-NLP/wanlp2020_arabic_fake_news_detection

custom

ar

ar-MSA: (Arabic (Modern Standard Arabic))

news articles

text

crawling and annotation(other)

a large-scale, multi-topic, and multi-country Arabic news dataset

1,000,000

documents

Low

UBC

nan

Machine Generation and Detection of Arabic Manipulated and Fake News

https://aclanthology.org/2020.wanlp-1.7.pdf

Arab

No

GitHub

Upon-Request

nan

Yes

fake news detection

WANLP

6.0

workshop

Arabic Natural Language Processing Workshop

E. Nagoudi,AbdelRahim Elmadany,Muhammad Abdul-Mageed,Tariq Alhindi,H. Cavusoglu

,University of British Columbia,,Columbia University;King Abdulaziz City for Science and Technology,

Fake news and deceptive machine-generated text are serious problems threatening modern societies, including in the Arab world. This motivates work on detecting false and manipulated stories online. However, a bottleneck for this research is lack of sufficient data to train detection models. We present a novel method for automatically generating Arabic manipulated (and potentially fake) news stories. Our method is simple and only depends on availability of true stories, which are abundant online, and a part of speech tagger (POS). To facilitate future work, we dispense with both of these requirements altogether by providing AraNews, a novel and large POS-tagged news dataset that can be used off-the-shelf. Using stories generated based on AraNews, we carry out a human annotation study that casts light on the effects of machine manipulation on text veracity. The study also measures human ability to detect Arabic machine manipulated text generated by our method. Finally, we develop the first models for detecting manipulated Arabic news and achieve state-of-the-art results on Arabic fake news detection (macro F1=70.06). Our models and data are publicly available.

Zaid Alyafeai

Khawas

nan

https://sourceforge.net/projects/kacst-acptool/

unknown

ar

ar-MSA: (Arabic (Modern Standard Arabic))

news articles

text

crawling

a corpus containing more than two million words and a corpora processing tool that is specifically designed for Arabic

2,910

documents

Low

Multiple Institutions

nan

New Language Resources for Arabic: Corpus Containing More Than Two Million Words and a Corpus Processing Tool

https://ieeexplore.ieee.org/abstract/document/6646005

Arab

No

sourceforge

Free

nan

No

topic classification

IALP

19.0

conference

International Conference on Asian Language Processing

A. Al-Thubaity,Marwa Khan,Manal Al-Mazrua,Maram Al-Mousa

,,,

Arabic is a resource-poor language relative to other languages with a similar number of speakers. This situation negatively affects corpus-based linguistic studies in Arabic and, to a lesser extent, Arabic language processing. This paper presents a brief overview of recent freely available Arabic corpora and corpora processing tools, and it examines some of the issues that may be preventing Arabic linguists from using the same. These issues reveal the need for new language resources to enrich and foster Arabic corpus-based studies. Accordingly, this paper introduces the design of a new Arabic corpus that includes modern standard Arabic varieties based on newspapers from all Arab countries and that comprises more than two million words, it also describes the main features of a corpus processing tool specifically designed for Arabic, called "Khawas ÛæÇÕ" ("diver" in English). Khawas provides more features than any other freely available corpus processing tool for Arabic, including n-gram frequency and concordance, collocations, and statistical comparison of two corpora. Finally, we outline modifications and improvements that could be made in future works.

Zaid Alyafeai

MultiUN v2

https://huggingface.co/datasets/un_multi

http://www.euromatrixplus.net/multi-un/

unknown

multilingual

ar-MSA: (Arabic (Modern Standard Arabic))

other

text

human translation

6 official languages of the UN, consisting of around 300 million words per language

65,156

documents

Low

DFKI

nan

MultiUN: A Multilingual Corpus from United Nation Documents

https://www.dfki.de/fileadmin/user_upload/import/4790_686_Paper.pdf

Arab

No

other

Free

nan

Yes

machine translation

LREC

223.0

conference

International Conference on Language Resources and Evaluation

A. Eisele

nan

This paper describes the acquisition, preparation and properties of a corpus extracted from the official documents of the United Nations (UN). This corpus is available in all 6 official languages of the UN, consisting of around 300 million words per language. We describe the methods we used for crawling, document formatting, and sentence alignment. This corpus also includes a common test set for machine translation. We present the results of a French-Chinese machine translation experiment performed on this corpus.

Zaid Alyafeai

WDC

https://huggingface.co/datasets/arbml/WDC

https://github.com/Maha-J-Althobaiti/Arabic_NER_Wiki-Corpus

CC BY 3.0

ar

ar-MSA: (Arabic (Modern Standard Arabic))

wikipedia

text

crawling

contains around 6 million tokens representing different genres, as Wikipedia is considered an open domain

6,000,000

tokens

Low

University of Essex

nan

Automatic Creation of Arabic Named Entity Annotated Corpus Using Wikipedia

https://aclanthology.org/E14-3012.pdf

Arab

No

GitHub

Free

nan

No

named entity recognition

EACL

14.0

conference

European Chapter of the Association for Computational Linguistics

M. Althobaiti,Udo Kruschwitz,Massimo Poesio

,University of Regensburg,

In this paper we propose a new methodology to exploit Wikipedia features and structure to automatically develop an Arabic NE annotated corpus. Each Wikipedia link is transformed into an NE type of the target article in order to produce the NE annotation. Other Wikipedia features - namely redirects, anchor texts, and inter-language links - are used to tag additional NEs, which appear without links in Wikipedia texts. Furthermore, we have developed a filtering algorithm to eliminate ambiguity when tagging candidate NEs. Herein we also introduce a mechanism based on the high coverage of Wikipedia in order to address two challenges particular to tagging NEs in Arabic text: rich morphology and the absence of capitalisation. The corpus created with our new method (WDC) has been used to train an NE tagger which has been tested on different domains. Judging by the results, an NE tagger trained on WDC can compete with those trained on manually annotated corpora.

Zaid Alyafeai

CALYOU

https://huggingface.co/datasets/arbml/CAYLOU

https://github.com/abidikarima/CALYOU

unknown

ar

ar-DZ: (Arabic (Algeria))

social media

text

crawling

A Comparable Spoken Algerian Corpus Harvested from YouTube

5,190

sentences

Medium

Multiple Institutions

nan

CALYOU: A Comparable Spoken Algerian Corpus Harvested from YouTube

https://hal.archives-ouvertes.fr/hal-01531591/document

Arab-Latn

No

GitHub

Free

nan

No

machine translation

INTERSPEECH

16.0

conference

Conference of the International Speech Communication Association

K. Abidi,M. Menacer,Kamel Smaïli

,,

This paper addresses the issue of comparability of comments extracted from Youtube. The comments concern spoken Algerian which could be either local Arabic, Modern Standard Arabic or French. This diversity of expression arises a huge number of problems concerning the data processing. In this article, several methods of alignment will be proposed and tested. The method which permits to best align is Word2Vec-based approach that will be used iteratively. This recurrent call of Word2Vec allows to improve significantly the results of comparability. In fact, a dictionary-based approach leads to a Recall of 4, while our approach allows to get a Recall of 33 at rank 1. Thanks to this approach, we built from Youtube CALYOU, a Comparable Corpus of the spoken Algerian.

Zaid Alyafeai

ADPBC

nan

https://github.com/salsama/Arabic-Information-Extraction-Corpus

unknown

ar

ar-MSA: (Arabic (Modern Standard Arabic))

other

text

crawling and annotation(other)

This corpus contains the words and their dependency relation produced by performing some steps

16

documents

Low

Multiple Institutions

nan

ADPBC: Arabic Dependency Parsing Based Corpora for Information Extraction

http://www.mecs-press.org/ijitcs/ijitcs-v13-n1/IJITCS-V13-N1-4.pdf

Arab

No

GitHub

Free

nan

No

part of speech tagging, dependency parsing, topic classification

IJITCS

0.0

journal

International Journal of Information Technology and Computer Science

Sally Mohamed,M. Hussien.,Hamdy M. Mousa

,,

There is a massive amount of different information and data in the World Wide Web, and the number of Arabic users and contents is widely increasing. Information extraction is an essential issue to access and sort the data on the web. In this regard, information extraction becomes a challenge, especially for languages, which have a complex morphology like Arabic. Consequently, the trend today is to build a new corpus that makes the information extraction easier and more precise. This paper presents Arabic linguistically analyzed corpus, including dependency relation. The collected data includes five fields; they are a sport, religious, weather, news and biomedical. The output is CoNLL universal lattice file format (CoNLL-UL). The corpus contains an index for the sentences and their linguistic meta-data to enable quick mining and search across the corpus. This corpus has seventeenth morphological annotations and eight features based on the identification of the textual structures help to recognize and understand the grammatical characteristics of the text and perform the dependency relation. The parsing and dependency process conducted by the universal dependency model and corrected manually. The results illustrated the enhancement in the dependency relation corpus. The designed Arabic corpus helps to quickly get linguistic annotations for a text and make the information Extraction techniques easy and clear to learn. The gotten results illustrated the average enhancement in the dependency relation corpus.

Zaid Alyafeai

AraSenCorpus

https://huggingface.co/datasets/arbml/AraSenCorpus

https://github.com/yemen2016/AraSenCorpus

MIT License

ar

mixed

social media

text

crawling and annotation(other)

contains 4.5 million tweets and covers both modern standard Arabic and some of the Arabic dialects

4,500,000

sentences

Medium

Multiple Institutions

nan

AraSenCorpus: A Semi-Supervised Approach for Sentiment Annotation of a Large Arabic Text Corpus

https://github.com/yemen2016/AraSenCorpus

Arab

No

GitHub

Free