arbml/masader · Datasets at Hugging Face

Multiple-Translation Arabic (MTA) Part 2

[]

nan

https://catalog.ldc.upenn.edu/LDC2005T05

LDC User Agreement for Non-Members

2,005

multilingual

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

news articles

text

other

All source data was drawn from January and February 2003. Here's a breakdown of the data amounts by source contained in this corpus:

100

documents

Low

LDC

nan

No

LDC

With-Fee

1,000.00 $

No

cross-lingual information retrieval,language teaching,machine translation

nan

Zaid Alyafeai

Arabic English Parallel News Part 1

[]

nan

https://catalog.ldc.upenn.edu/LDC2004T18

LDC User Agreement for Non-Members

2,004

multilingual

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

news articles

text

other

LDC collected the data in this corpus via Ummah Press Service from January 2001 to September 2004. It totals 8,439 story pairs, 68,685 sentence pairs. The corpus is aligned at sentence level. All data files are SGML documents.

2,000,000

tokens

Low

LDC

nan

No

LDC

With-Fee

3,000.00 $

No

machine translation

nan

Zaid Alyafeai

Arabic News Translation Text Part 1

[]

nan

https://catalog.ldc.upenn.edu/LDC2004T17

LDC User Agreement for Non-Members

2,004

multilingual

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

news articles

text

other

Here is a breakdown of the Arabic material by source:

1,526

documents

Low

LDC

nan

No

LDC

With-Fee

3,000.00 $

No

cross-lingual information retrieval,language teaching,machine translation

nan

Zaid Alyafeai

Arabic Treebank: Part 2 v 2.0

[]

nan

https://catalog.ldc.upenn.edu/LDC2004T02

LDC User Agreement for Non-Members

2,004

ar

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

news articles

text

other

The following table gives a breakdown of the data contained in the entire Arabic Treebank project, with discrepancies between versions for Parts 1, 2, and 3. The fields include source, number of stories, total number of tokens, number of tokens after clitic separation, and number of Arabic word tokens after punctuation, numbers, and latin strings have been taken out. The totals given at the bottom are calculated from the latest versions where discrepencies exist, and do not include tokens after clitic separation since that number is missing from Part 4.

688,549

tokens

Low

LDC

nan

No

LDC

With-Fee

4,000.00 $

No

information retrieval,cross-lingual information retrieval,information detection

nan

Zaid Alyafeai

Arabic Treebank: Part 3 v 1.0

[]

nan

https://catalog.ldc.upenn.edu/LDC2004T11

LDC User Agreement for Non-Members

2,004

ar

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

news articles

text

other

The following table gives a breakdown of the data contained in the entire Arabic Treebank project, with discrepancies between versions for Parts 1 and 3. The fields include source, number of stories, total number of tokens, number of tokens after clitic separation, and number of Arabic word tokens after punctuation, numbers, and latin strings have been taken out. The totals given at the bottom are calculated from the latest versions where discrepencies exist, and do not include tokens after clitic separation since that number is missing from Part 4.

300,000

tokens

Low

LDC

nan

No

LDC

With-Fee

3,500.00 $

No

information retrieval,cross-lingual information retrieval,information detection

nan

Zaid Alyafeai

Buckwalter Arabic Morphological Analyzer Version 2.0

[]

nan

https://catalog.ldc.upenn.edu/LDC2004L02

LDC User Agreement for Non-Members

2,004

multilingual

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

nan

spoken

other

The data consists primarily of three Arabic-English lexicon files: prefixes (299 entries), suffixes (618 entries), and stems (82158 entries representing 38600 lemmas). The lexicons are supplemented by three morphological compatibility tables used for controlling prefix-stem combinations (1648 entries), stem-suffix combinations (1285 entries), and prefix-suffix combinations (598 entries). The actual code for morphology analysis and POS tagging is contained in a Perl script. The documentation consists of a readme file with a description of the lexicon files, the morphological compatibility tables, the morphology analysis algorithm, a summary of stem morphological categories, and a table with the authors Arabic transliteration system.

299

tokens

Low

LDC

nan

No

LDC

With-Fee

N/A $

No

machine translation,information retrieval

nan

Zaid Alyafeai

Prague Arabic Dependency Treebank 1.0

[]

nan

https://catalog.ldc.upenn.edu/LDC2004T23

LDC User Agreement for Non-Members

2,004

ar

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

news articles

text

other

The corpus of PADT 1.0 consists of morphologically and analytically annotated newswire texts of Modern Standard Arabic, which originate from the Arabic Gigaword (LDC2003T12) and the plain data of Arabic Treebank: Part 1 v 2.0 (LDC2003T06) and Arabic Treebank: Part 2 v 2.0 (LDC2004T02).

212,500

tokens

Low

LDC

nan

No

LDC

With-Fee

100.00 $

No

cross-lingual information retrieval,information extraction,information retrieval,language modeling,language teaching,machine translation,parsing

nan

Zaid Alyafeai

Arabic Gigaword

[]

nan

https://catalog.ldc.upenn.edu/LDC2003T12

LDC User Agreement for Non-Members

2,003

ar

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

news articles

text

other

There are 319 files, totalling approximately 1.1GB in compressed form (4348 MB uncompressed, and 391619 Kwords).

1,256,719

documents

Low

LDC

nan

No

LDC

With-Fee

3,000.00 $

No

information retrieval,language modeling

nan

Zaid Alyafeai

Arabic Treebank: Part 1 - 10K-word English Translation

[]

nan

https://catalog.ldc.upenn.edu/LDC2003T07

LDC User Agreement for Non-Members

2,003

ar

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

news articles

text

other

The project targets the translation of a written Modern Standard Arabic corpus from the Agence France Presse (AFP) newswire archives for July 2000 (the files are dated 07/15/2000). The corpus consists of 49 source stories, which is a subset of the 734 stories published in Arabic Treebank: Part 1 v 2.0 (LDC2003T06). These 49 source files comprise 418 paragraphs and 9,981 words.

9,981

tokens

Low

LDC

nan

No

LDC

With-Fee

1,500.00 $

No

information retrieval,cross-lingual information retrieval,information detection

nan

Zaid Alyafeai

Arabic Treebank: Part 1 v 2.0

[]

nan

https://catalog.ldc.upenn.edu/LDC2003T06

LDC User Agreement for Non-Members

2,003

ar

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

news articles

text

other

The following table gives a breakdown of the data contained in the entire Arabic Treebank project, with discrepancies between versions for Parts 1, 2, and 3. The fields include source, number of stories, total number of tokens, number of tokens after clitic separation, and number of Arabic word tokens after punctuation, numbers, and Latin strings have been taken out. The totals given at the bottom are calculated from the latest versions where discrepancies exist, and do not include tokens after clitic separation since that number is missing from Part 4.

688,549

tokens

Low

LDC

nan

No

LDC

With-Fee

3,000.00 $

No

information retrieval,cross-lingual information retrieval,information detection

nan

Zaid Alyafeai

Multiple-Translation Arabic (MTA) Part 1

[]

nan

https://catalog.ldc.upenn.edu/LDC2003T18

LDC User Agreement for Non-Members

2,003

multilingual

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

news articles

text

other

Here's a breakdown of the data amounts by source contained in this corpus:

141

documents

Low

LDC

nan

No

LDC

With-Fee

1,000.00 $

No

cross-lingual information retrieval,language teaching,machine translation

nan

Zaid Alyafeai

1997 HUB5 Arabic Evaluation

[]

nan

https://catalog.ldc.upenn.edu/LDC2002S22

LDC User Agreement for Non-Members

2,002

ar

ar-EG: (Arabic (Egypt))

transcribed audio

spoken

other

This publication contains 20 sphere files encoded in two channel interleaved mulaw with a sampling rate of 8 KHz, for a total of 424,160,000 bytes (405 Mbytes) of sphere data. The sphere headers have been modified from the original Evaluation data by the addition of sample checksums to the CALLHOME data files.

20

documents

Low

LDC

nan

No

LDC

With-Fee

1,500.00 $

No

speech recognition

nan

Zaid Alyafeai

1997 HUB5 Arabic Transcripts

[]

nan

https://catalog.ldc.upenn.edu/LDC2002T39

LDC User Agreement for Non-Members

2,002

ar

ar-EG: (Arabic (Egypt))

transcribed audio

text

other

There are 40 data files. Each of the 20 calls has transcripts in two formats: .txt and .scr.

40

documents

Low

LDC

nan

No

LDC

With-Fee

500.00 $

No

speech recognition

nan

Zaid Alyafeai

Buckwalter Arabic Morphological Analyzer Version 1.0

[]

nan

https://catalog.ldc.upenn.edu/LDC2002L49

LDC User Agreement for Non-Members

2,002

multilingual

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

other

text

other

The data consists primarily of three Arabic-English lexicon files: prefixes (299 entries), suffixes (618 entries), and stems (82,158 entries representing 38,600 lemmas). The lexicons are supplemented by three morphological compatibility tables used for controlling prefix-stem combinations (1,648 entries), stem-suffix combinations (1,285 entries), and prefix-suffix combinations (598 entries). The actual code for morphology analysis and POS tagging is contained in a Perl script. The documentation consists of a readme file with a description of the lexicon files, the morphological compatibility tables, the morphology analysis algorithm, a summary of stem morphological categories, and a table with the author's Arabic transliteration system.

299

tokens

Low

LDC

nan

No

LDC

Upon-Request

nan

No

information retrieval,machine translation

nan

Zaid Alyafeai

CALLHOME Egyptian Arabic Speech Supplement

[]

nan

https://catalog.ldc.upenn.edu/LDC2002S37

LDC User Agreement for Non-Members

2,002

ar

ar-EG: (Arabic (Egypt))

transcribed audio

spoken

other

There are 20 data files in sphere format. The files are 8 KHz shorten-compressed two-channel mulaw. 12 of the files were recorded from domestic phone calls (both parties living in the continental U.S.), while the other eight are overseas calls (a participant in the U.S. called a friend or relative in Egypt or some other overseas country).

20

documents

Low

LDC

nan

No

LDC

With-Fee

1,000.00 $

No

speech recognition

nan

Zaid Alyafeai

CALLHOME Egyptian Arabic Transcripts Supplement

[]

nan

https://catalog.ldc.upenn.edu/LDC2002T38

LDC User Agreement for Non-Members

2,002

ar

ar-EG: (Arabic (Egypt))

transcribed audio

text

other

There are 40 data files. Each of the 20 calls has transcripts in two formats: .txt and .scr.

40

documents

Low

LDC

nan

No

LDC

With-Fee

750.00 $

No

speech recognition

nan

Zaid Alyafeai

Arabic Newswire Part 1

[]

nan

https://catalog.ldc.upenn.edu/LDC2001T55

LDC User Agreement for Non-Members

2,002

ar

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

news articles

text

other

The data is in 2,337 compressed (zipped) Arabic text data files. There are 209 Mb of compressed data (869 Mb uncompressed) with approximately 383,872 documents containing 76 million tokens over approximately 666,094 unique words.

383,872

documents

Low

LDC

nan

No

LDC

With-Fee

1,200.00 $

No

information retrieval,language modeling

nan

Zaid Alyafeai

Egyptian Colloquial Arabic Lexicon

[]

nan

https://catalog.ldc.upenn.edu/LDC99L22

LDC User Agreement for Non-Members

2,002

ar

ar-EG: (Arabic (Egypt))

transcribed audio

text

other

The lexicon contains 51,202 entries, drawn from 140 CALLHOME telephone conversations among native speakers of Colloquial Egyptian Arabic, collected and published by the LDC as follows: CALLHOME Egyptian Arabic Speech LDC97S45, CALLHOME Egyptian Arabic Transcripts LDC97T19, CALLHOME Egyptian Arabic Speech Supplement LDC200237 and CALLHOME Egyptian Arabic Transcripts Supplement LDC2002T38. The lexicon also contains entries derived manually from the Badawi & Hines dictionary of Colloquial Egyptian Arabic.

51,202

tokens

Low

LDC

nan

No

LDC

With-Fee

2,000.00 $

No

speech recognition

nan

Zaid Alyafeai

TDT4 Multilingual Text and Annotations

[]

nan

https://catalog.ldc.upenn.edu/LDC2005T16

LDC User Agreement for Non-Members

2,005

multilingual

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

news articles

text

other

The TDT4 corpus contains news data collected daily from 20 news sources in three languages over a period of four months (October 2000 through January 2001).

100,000

documents

Low

LDC

nan

Arab

No

LDC

With-Fee

2,000.00 $

No

topic detection and tracking

nan

Zaid Alyafeai

CALLHOME Egyptian Arabic Speech

[]

nan

https://catalog.ldc.upenn.edu/LDC97S45

LDC User Agreement for Non-Members

2,002

ar

ar-EG: (Arabic (Egypt))

transcribed audio

spoken

other

All calls, which lasted up to 30 minutes, originated in North America and were placed to locations overseas (typically Egypt). Most participants called family members or close friends.

120

sentences

Low

LDC

nan

No

LDC

With-Fee

1,500.00 $

No

speech recognition

nan

Zaid Alyafeai

CALLHOME Egyptian Arabic Transcripts

[]

nan

https://catalog.ldc.upenn.edu/LDC97T19

LDC User Agreement for Non-Members

2,002

ar

ar-EG: (Arabic (Egypt))

transcribed audio

text

other

The transcripts are timestamped by speaker turn for alignment with the speech signal and are provided in standard orthography.

120

sentences

Low

LDC

nan

No

LDC

With-Fee

1,000.00 $

No

speech recognition

nan

Zaid Alyafeai

TIDES Extraction (ACE) 2003 Multilingual Training Data

[]

nan

https://catalog.ldc.upenn.edu/LDC2004T09

LDC User Agreement for Non-Members

2,004

multilingual

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

transcribed audio

text

other

Annotations for this corpus were produced by Linguistic Data Consortium to support the following tasks broken down by language:

42,197

tokens

Low

LDC

nan

Arab

No

LDC

With-Fee

3,000.00 $

No

information retrieval,information detection

nan

Zaid Alyafeai

CALLFRIEND Egyptian Arabic

[]

nan

https://catalog.ldc.upenn.edu/LDC96S49

LDC User Agreement for Non-Members

2,002

ar

ar-EG: (Arabic (Egypt))

transcribed audio

spoken

other

The corpus consists of 60 unscripted telephone conversations, lasting between 5-30 minutes. The corpus also includes documentation describing speaker information (sex, age, education, callee telephone number) and call information (channel quality, number of speakers).

25

hours

Low

LDC

nan

No

LDC

With-Fee

1,000.00 $

No

language identification

nan

Zaid Alyafeai

Sudanese Dialect tweets about ridesharing companies

[]

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

https://docs.google.com/spreadsheets/d/1bNwimEQFMWtjlsKtL8PH_RNFNjg-b6p3/edit?usp=sharing&ouid=101796411348671465142&rtpof=true&sd=true

unknown

2,020

ar

ar-SD: (Arabic (Sudan))

social media

text

crawling and annotation(other)

Sentiment Analysis dataset collected from Twitter. It contains people's opinions on Sudanese Ridesharing companies.

2,116

sentences

High

University of Khartoum

nan

Sentiment Analysis for Sudanese Arabic Dialect Using comparative Supervised Learning approach

https://ieeexplore.ieee.org/document/9429560

Arab

No

Gdrive

Free

nan

No

sentiment analysis

ICCCEEE

0.0

conference

2020 International Conference on Computer, Control, Electrical, and Electronics Engineering

Shahad Abuuznien, Zena Abdelmohsin, Ehsan Abdu, Izzeldein Amin

University of Khartoum

Sentiment analysis is several methods, techniques, and tools that are used to determine the polarity of the text (positive, negative, or neutral). The most popular approaches to address this problem, is the machine learning approach, lexicon-based approach, and hybrid approach. This project focuses on extracting and analyzing Sudanese social media feeds about ridesharing services. This project aims to tackle the issue of Sudanese Arabic dialect analysis by conducting a comparative analysis to measure the performance of the machine learning algorithms using Sudanese dialect corpus comparing different preprocessing approaches. For this study, a stop word list that combines a modern standard Arabic list and a Sudanese stop word list was built to be conducted through the analysis as one of the preprocessing steps. with four classifiers applied on a dataset consist of 2116 tweets. In particular, Naïve Bayes (NB), Support vector machine (SVM), Logistic Regression, and K-Nearest Neighbor (KNN) had been trained and measured the performance. The results of the selected classifiers against the dataset which had been applied to various preprocessing steps revealed that SVM with stemming only gives the highest F1-score (0.71), and the best accuracy (0.95).

Khalid N. Elmadani

mozilla foundation common voice dataset

[]

https://huggingface.co/mozilla-foundation

https://github.com/common-voice/cv-dataset/tree/main/datasets

MPL-2.0

2,020

multilingual

mixed

transcribed audio

spoken

other

The Common Voice dataset consists of a unique MP3 and corresponding text file. Many of the 20217 recorded hours in the dataset also include demographic metadata like age, sex, and accent that can help improve the accuracy of speech recognition engines.

87

hours

Low

Mozilla, Indiana University, Artie, Inc.

nan

Common Voice: A Massively-Multilingual Speech Corpus

https://arxiv.org/pdf/1912.06670.pdf

Arab

No

HuggingFace

Free

nan

Yes

speech recognition, language identification

LREC

369.0

conference

International Conference on Language Resources and Evaluation

Rosana Ardila, Megan Branson, Kelly Davis, Michael Henretty, Michael Kohler, Josh Meyer, Reuben Morais, Lindsay Saunders, Francis M. Tyers, Gregor Weber

Mozilla, Indiana University, Artie, Inc.

The Common Voice corpus is a massively-multilingual collection of transcribed speech intended for speech technology research and development. Common Voice is designed for Automatic Speech Recognition purposes but can be useful in other domains (e.g. language identification). To achieve scale and sustainability, the Common Voice project employs crowdsourcing for both data collection and data validation. The most recent release includes 29 languages, and as of November 2019 there are a total of 38 languages collecting data. Over 50,000 individuals have participated so far, resulting in 2,500 hours of collected audio. To our knowledge this is the largest audio corpus in the public domain for speech recognition, both in terms of number of hours and number of languages. As an example use case for Common Voice, we present speech recognition experiments using Mozilla’s DeepSpeech Speech-to-Text toolkit. By applying transfer learning from a source English model, we find an average Character Error Rate improvement of 5.99 ± 5.48 for twelve target languages (German, French, Italian, Turkish, Catalan, Slovenian, Welsh, Irish, Breton, Tatar, Chuvash, and Kabyle). For most of these languages, these are the first ever published results on end-to-end Automatic Speech Recognition.

Wafaa Mohammed

Ultimate Arabic News Dataset

[]

https://huggingface.co/datasets/khalidalt/ultimate_arabic_news

https://data.mendeley.com/datasets/jz56k5wxz7/1

CC BY 4.0

2,022

ar

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

news articles

text

crawling

The Ultimate Arabic News Dataset is a collection of single-label modern Arabic texts that are used in news websites and press articles.

381,000

sentences

Low

Yalova Universitesi

nan

Arab

No

Mendeley Data

Free

nan

No

topic classification

nan

Al-Dulaimi, Ahmed Hashim

Yalova Universitesi

nan

Khalid Almubarak

Goud-sum

[]

https://huggingface.co/datasets/Goud/Goud-sum

https://github.com/issam9/goud-summarization-dataset

unknown

2,022

ar

ar-MA: (Arabic (Morocco))

news articles

text

crawling

Goud-sum contains 158k articles and their headlines extracted from Goud.ma news website. The articles are written in the Arabic script. All headlines are in Moroccan Darija, while articles may be in Moroccan Darija, in Modern Standard Arabic, or a mix of both (code-switched Moroccan Darija).

158,000

documents

Low

nan

GOUD.MA: A NEWS ARTICLE DATASET FOR SUMMARIZATION IN MOROCCAN DARIJA

https://openreview.net/pdf?id=BMVq5MELb9

Arab

No

HuggingFace

Free

nan

Yes

summarization

ICLR

0.0

workshop

International Conference on Learning Representations

Abderrahmane Issam, Khalil Mrini

Archipel Cognitive, University of California San Diego

Moroccan Darija is a vernacular spoken by over 30 million people primarily in Morocco. Despite a high number of speakers, it remains a low-resource language. In this paper, we introduce GOUD.MA: a dataset of over 158k news articles for automatic summarization in code-switched Moroccan Darija. We analyze the dataset and find that it requires a high level of abstractive reasoning. We fine-tune the Arabic-language BERT (AraBERT), and the language models for the Moroccan (DarijaBERT), and Algerian (DziriBERT) national vernaculars for summarization on GOUD.MA. The results show that GOUD.MA is a challenging summarization benchmark dataset. We release our dataset publicly in an effort to encourage the diversity of evaluation tasks to improve language modeling in Moroccan Darija.

Abderrahmane Issam

ElecMorocco2016

[]

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

https://github.com/sentiprojects/ElecMorocco2016

unknown

2,016

ar

ar-MA: (Arabic (Morocco))

social media

text

crawling

A sentiment analysis dataset containing 10254 Arabic facebook comments about the Moroccan elections of 2016. The comments are written in standard arabic and morrocan dialect.

10,000

sentences

Low

nan

Collecting and Processing Arabic Facebook Comments for Sentiment Analysis

https://link.springer.com/chapter/10.1007/978-3-319-66854-3_20

Arab

No

GitHub

Free

nan

No

sentiment analysis

MEDI

5.0

conference

International Conference on Model and Data Engineering

Abdeljalil Elouardighi, Mohcine Maghfour, Hafdalla Hammia

nan

Social networks platforms such as Facebook are becoming one of the most powerful sources for information. The produced and shared data are important in volume, in velocity and in variety. Processing these data in the raw state to extract useful information can be a very difficult task and a big challenge. Furthermore, the Arabic language under its modern standard or dialectal shape is one of the languages producing an important quantity of data in social networks and the least analyzed. The characteristics and the specificity of the Arabic language present a big challenge for sentiment analysis, especially if this analysis is performed on Arabic Facebook comments. In this paper, we present a methodology that we have elaborated, for collecting and preprocessing Facebook comments written in Modern Standard Arabic (MSA) or in Moroccan Dialectal Arabic (MDA) for Sentiment Analysis (SA) using supervised classification methods. In this methodology, we have detailed the processing applied to the comments’ text as well as various schemes of features’ construction (words or groups of words) useful for supervised sentiments’ classification. This methodology was tested on comments written in MSA or in MDA collected from Facebook for the sentiment analysis on a political phenomenon. The experiments’ results obtained are promising and this encourages us to continue working on this topic.

Abderrahmane Issam

MASC: Massive Arabic Speech Corpus

[]

https://huggingface.co/datasets/pain/MASC

https://ieee-dataport.org/open-access/masc-massive-arabic-speech-corpus

CC BY 4.0

2,022

ar

mixed

transcribed audio

spoken

crawling and annotation(other)

This corpus is a dataset that contains 1,000 hours of speech sampled at 16~kHz and crawled from over 700 YouTube channels.

1,000

hours

nan

Arab

No

other

Free

nan

Yes

speech recognition

nan

Mohammad Al-Fetyani,Muhammad Al-Barham,Gheith Abandah,Adham Alsharkawi, Maha Dawas

nan

This paper releases and describes the creation of the Massive Arabic Speech Corpus (MASC). This corpus is a dataset that contains 1,000 hours of speech sampled at 16~kHz and crawled from over 700 YouTube channels. MASC is multi-regional, multi-genre, and multi-dialect dataset that is intended to advance the research and development of Arabic speech technology with the special emphasis on Arabic speech recognition. In addition to MASC, a pre-trained 3-gram language model and a pre-trained automatic speech recognition model are also developed and made available for interested researches. For a better language model, a new and unified Arabic speech corpus is required, and thus, a dataset of 12~M unique Arabic words is created and released. To make practical and convenient use of MASC, the whole dataset is stratified based on dialect into clean and noisy portions. Each of the two portions is then stratified and divided into three subsets: development, test, and training sets. The best word error rate achieved by the speech recognition model is 19.8% for the clean development set and 21.8% for the clean test set.

Mohammad Amjad Al-Fetyani

AttImam

[]

nan

https://catalog.ldc.upenn.edu/LDC2022T02

LDC User Agreement

2,022

ar

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

news articles

text

crawling and annotation(other)

AttImam was developed by Al-Imam Mohammad Ibn Saud Islamic University and consists of approximately 2,000 attribution relations applied to Arabic newswire text from Arabic Treebank: Part 1 v 4.1 (LDC2010T13). Attribution refers to the process of reporting or assigning an utterance to the correct speaker.

2,000

sentences

Low

Al-Imam Mohammad Ibn Saud Islamic University

nan

Arab-Latn

No

LDC

With-Fee

250.00 $

No

discourse analysis, entity extraction, language identification

nan

Amal Alsaif, Tasniem Alyahya, Madawi Alotibi, Huda Almuzaini, Abeer Alqahtani

nan

Ahmed Ruby

RATS Speaker Identification

[]

nan

https://catalog.ldc.upenn.edu/LDC2021S08

LDC User Agreement for Non-Members

2,021

multilingual

ar-LEV: (Arabic(Levant))

transcribed audio

spoken

other

The source audio consists of conversational telephone speech recordings collected by LDC specifically for the RATS program from Levantine Arabic, Pashto, Urdu, Farsi and Dari native speakers. Annotations on the audio files include start time, end time, speech activity detection (SAD) label, SAD provenance, speaker ID, speaker ID provenance, language ID, and language ID provenance.

1,900

hours

Low

LDC

nan

Arab

No

LDC

With-Fee

7,500.00 $

No

speaker identification

nan

Zaid Alyafeai

2018 NIST Speaker Recognition Evaluation Test Set

[]

nan

https://catalog.ldc.upenn.edu/LDC2020S04

LDC User Agreement for Non-Members

2,020

multilingual

ar-TN: (Arabic (Tunisia))

transcribed audio

spoken

other

The telephone speech data was drawn from the Call My Net 2 (CMN2) collection conducted by LDC in Tunisia in which Tunisian Arabic speakers called friends or relatives who agreed to record their telephone conversations lasting between 8-10 minutes. The speech segments include PSTN (public switched telephone network) and VOIP (voice over IP) data.

396

hours

Low

LDC

nan

Arab

No

LDC

With-Fee

750.00 $

No

speaker identification

nan

Zaid Alyafeai

BOLT Egyptian-English Word Alignment -- Discussion Forum Training

[]

nan

https://catalog.ldc.upenn.edu/LDC2019T06

LDC User Agreement for Non-Members

2,019

multilingual

ar-EG: (Arabic (Egypt))

commentary

text

other

This release consists of Egyptian source discussion forum threads harvested from the Internet by LDC using a combination of manual and automatic processes. The source data is released as BOLT Arabic Discussion Forums (LDC2018T10).

400,448

tokens

Low

LDC

nan

Arab

No

LDC

With-Fee

1,750.00 $

No

information retrieval,machine translation

nan

Zaid Alyafeai

2011 NIST Language Recognition Evaluation Test Set

[]

nan

https://catalog.ldc.upenn.edu/LDC2018S06

LDC User Agreement for Non-Members

2,018

multilingual

mixed

transcribed audio

spoken

other

This release includes training data for nine language varieties that had not been represented in prior LRE cycles -- Arabic (Iraqi), Arabic (Levantine), Arabic (Maghrebi), Arabic (Standard), Czech, Lao, Punjabi, Polish, and Slovak -- contained in 893 audited segments of roughly 30 seconds duration and in 400 full-length CTS recordings. The evaluation test set comprises a total of 29,511 audio files, all manually audited at LDC for language and divided equally into three different test conditions according to the nominal amount of speech content per segment.

204

hours

Low

LDC

nan

Arab

No

LDC

With-Fee

500.00 $

No

language identification

nan

Zaid Alyafeai

BOLT Information Retrieval Comprehensive Training and Evaluation

[]

nan

https://catalog.ldc.upenn.edu/LDC2018T18

LDC User Agreement for Non-Members

2,018

multilingual

ar-EG: (Arabic (Egypt))

commentary

text

other

BOLT Information Retrieval Comprehensive Training and Evaluation contains the pilot, dry run, and evaluation data developed for each phase of the BOLT IR task, including: (1) natural-language IR queries, system responses to queries, and manually-generated assessment judgments for system responses; (2) discussion forum source documents in Arabic, Chinese and English; (3) scoring software for each evaluation phase; and (4) experimental data developed in Phase 2.

nan

Low

LDC

nan

Arab

No

LDC

With-Fee

2,500.00 $

No

information retrieval,machine translation

nan

Zaid Alyafeai

RATS Language Identification

[]

nan

https://catalog.ldc.upenn.edu/LDC2018S10

LDC User Agreement for Non-Members

2,018

multilingual

ar-LEV: (Arabic(Levant))

transcribed audio

spoken

other

The source audio consists of conversational telephone speech recordings from: (1) conversational telephone speech (CTS) recordings, taken either from previous LDC CTS corpora, or from CTS data collected specifically for the RATS program from Levantine Arabic, Pashto, Urdu, Farsi and Dari native speakers; and (2) portions of VOA broadcast news recordings, taken from data used in the 2009 NIST Language Recognition Evaluation. The 2009 LRE Test Set is available from LDC as LDC2014S06.

600

hours

Low

LDC

nan

Arab

No

LDC

With-Fee

7,500.00 $

No

language identification

nan

Zaid Alyafeai

KSUEmotions

[]

nan

https://catalog.ldc.upenn.edu/LDC2017S12

LDC User Agreement for Non-Members

2,017

ar

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

transcribed audio

spoken

other

Audio was recorded in each participant's home. Audio is presented as 16-bit 16 kHz flac compressed wav. In addition to speech files and metadata about the speakers, timeless label files and automatic time segmentation alignment files are included. Text is presented as UTF-8 plain text.

5

hours

Low

LDC

nan

Arab

No

LDC

With-Fee

1,000.00 $

No

prosody,speech recognition

nan

Zaid Alyafeai

RATS Keyword Spotting

[]

nan

https://catalog.ldc.upenn.edu/LDC2017S20

LDC User Agreement for Non-Members

2,017

multilingual

ar-LEV: (Arabic(Levant))

transcribed audio

spoken

other

The source audio consists of conversational telephone speech recordings collected by LDC: (1) data collected for the RATS program from Levantine Arabic and Farsi speakers; and (2) material from Levantine Arabic QT Training Data Set 5, Speech (LDC2006S29) and CALLFRIEND Farsi Second Edition Speech (LDC2014S01).

400

hours

Low

LDC

nan

Arab

No

LDC

With-Fee

7,500.00 $

No

keyword spotting

nan

Zaid Alyafeai

JANA: A Human-Human Dialogues Corpus for Egyptian Dialect

[]

nan

https://catalog.ldc.upenn.edu/LDC2016T24

LDC User Agreement for Non-Members

2,016

ar

ar-EG: (Arabic (Egypt))

transcribed audio

text

other

The transcribed dialogues consist of 52 telephone calls and 30 instant messaging conversations, amounting to approximately 20,311 words. The data contains roughly 3,001 conversation turns, with an average of 6.7 words per turn, and 4,725 utterances, with an average of 4.3 words per utterance. The data was transcribed using Transcriber.

20,311

tokens

Low

LDC

nan

Arab

No

LDC

With-Fee

1,650.00 $

No

dialogue generation

nan

Zaid Alyafeai

RATS Speech Activity Detection

[]

nan

https://catalog.ldc.upenn.edu/LDC2015S02

LDC User Agreement for Non-Members

2,015

multilingual

ar-LEV: (Arabic(Levant))

transcribed audio

spoken

other

The source audio consists of conversational telephone speech recordings collected by LDC: (1) data collected for the RATS program from Levantine Arabic, Farsi, Pashto and Urdu speakers; and (2) material from the Fisher English (LDC2004S13, LDC2005S13), and Fisher Levantine Arabic telephone studies (LDC2007S02), as well as from CALLFRIEND Farsi (LDC2014S01).

350

hours

Low

LDC

nan

Arab

No

LDC

With-Fee

7,500.00 $

No

speech activity detection

nan

Zaid Alyafeai

ACE 2007 Multilingual Training Corpus

[]

nan

https://catalog.ldc.upenn.edu/LDC2014T18

LDC User Agreement for Non-Members

2,014

multilingual

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

news articles

text

other

The Arabic data is composed of newswire (60%) published in October 2000-December 2000 and weblogs (40%) published during the period November 2004-February 2005. The Spanish data set consists entirely of newswire material from multiple sources published in January 2005-April 2005.

98,353

tokens

Low

LDC

nan

Arab

No

LDC

With-Fee

1,000.00 $

No

information retrieval

nan

Zaid Alyafeai

HyTER Networks of Selected OpenMT08/09 Sentences

[]

nan

https://catalog.ldc.upenn.edu/LDC2014T09

LDC User Agreement for Non-Members

2,014

multilingual

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

news articles

text

other

The source material is comprised of Arabic and Chinese newswire and web data collected by LDC in 2007. Annotators created meaning-equivalent annotations under three annotation protocols. In the first protocol, foreign language native speakers built English networks starting from foreign language sentences. In the second, English native speakers built English networks from the best translation of a foreign language sentence as identified by NIST (National Institute of Standards and Technology). In the third protocol, English native speakers built English networks starting from the best translation, but those annotators also had access to three additional, independently produced human translations. Networks created by different annotators for each sentence were combined and evaluated.

102

sentences

Low

LDC

nan

Arab

No

LDC

With-Fee

150.00 $

No

machine translation

nan

Zaid Alyafeai

NIST 2012 Open Machine Translation (OpenMT) Progress Test Five Language Source

[]

nan

https://catalog.ldc.upenn.edu/LDC2014T02

LDC User Agreement for Non-Members

2,014

multilingual

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

web pages

text

other

This release consists of 20 files, four for each of the five languages, presented in XML with an included DTD. The four files are source and reference data from the same source data in the following two styles:

4

documents

Low

LDC

nan

Arab

No

LDC

With-Fee

150.00 $

No

machine translation

nan

Zaid Alyafeai

United Nations Proceedings Speech

[]

nan

https://catalog.ldc.upenn.edu/LDC2014S08

LDC User Agreement for Non-Members

2,014

multilingual

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

transcribed audio

spoken

other

Data is presented either as mp3 or flac compressed wav and are 16-bit single channel files in either 22,050 or 8,000 Hz organized by committee and session number, then language. The folder labeled "Floor" indicates the microphone used by the particular speaker. Those files may include other languages, for instance, if the speaker's language was not among the six official UN languages.

8,500

hours

Low

LDC

nan

Arab

No

LDC

With-Fee

5,000.00 $

No

speech recognition,language identification

nan

Zaid Alyafeai

1993-2007 United Nations Parallel Text

[]

nan

https://catalog.ldc.upenn.edu/LDC2013T06

LDC User Agreement for Non-Members

2,013

multilingual

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

other

text

other

The data is presented as raw text and word-aligned text. The raw text is very close to what was extracted from the original word processing documents in UN ODS (e.g., Word, WordPerfect, PDF), converted to UTF-8 encoding.

520,283

documents

Low

LDC

nan

Arab

No

LDC

With-Fee

175.00 $

No

machine translation

nan

Zaid Alyafeai

MADCAT Phase 2 Training Set

[]

nan

https://catalog.ldc.upenn.edu/LDC2013T09

LDC User Agreement for Non-Members

2,013

ar

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

news articles

text

other

This release includes 27,814 annotation files in both GEDI XML and MADCAT XML formats (gedi.xml and madcat.xml) along with their corresponding scanned image files in TIFF format. The annotation results in GEDI XML output files include ground truth annotations and source transcripts.

27,814

documents

Low

LDC

nan

Arab

No

LDC

With-Fee

2,500.00 $

No

handwriting recognition,machine translation

nan

Zaid Alyafeai

MADCAT Phase 3 Training Set

[]

nan

https://catalog.ldc.upenn.edu/LDC2013T15

LDC User Agreement for Non-Members

2,013

ar

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

news articles

text

other

This release includes 4,540 annotation files in both GEDI XML and MADCAT XML formats (gedi.xml and madcat.xml) along with their corresponding scanned image files in TIFF format. The annotation results in GEDI XML files include ground truth annotations and source transcripts.

4,540

documents

Low

LDC

nan

Arab

No

LDC

With-Fee

1,000.00 $

No

handwriting recognition,machine translation

nan

Zaid Alyafeai

NIST 2008-2012 Open Machine Translation (OpenMT) Progress Test Sets

[]

nan

https://catalog.ldc.upenn.edu/LDC2013T07

LDC User Agreement for Non-Members

2,013

multilingual

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

web pages

text

other

This release contains 2,748 documents with corresponding source and reference files, the latter of which contains four independent human reference translations of the source data. The source data is comprised of Arabic and Chinese newswire and web data collected by LDC in 2007. The table below displays statistics by source, genre, documents, segments and source tokens.

2,748

documents

Low

LDC

nan

Arab

No

LDC

With-Fee

150.00 $

No

machine translation

nan

Zaid Alyafeai

2006 NIST Speaker Recognition Evaluation Test Set Part 2

[]

nan

https://catalog.ldc.upenn.edu/LDC2012S01

LDC User Agreement for Non-Members

2,012

multilingual

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

transcribed audio

spoken

other

The speech data in this release was collected by LDC as part of the Mixer project, in particular Mixer Phases 1, 2, and 3. The Mixer project supports the development of robust speaker recognition technology by providing carefully collected and audited speech from a large pool of speakers recorded simultaneously across numerous microphones and in different communicative situations and/or in multiple languages. The data is mostly English speech, but includes some speech in Arabic, Bengali, Chinese, Farsi, Hindi, Korean, Russian, Spanish, Thai, and Urdu.

568

hours

Low

LDC

nan

Arab

No

LDC

With-Fee

350.00 $

No

speaker identification

nan

Zaid Alyafeai

MADCAT Phase 1 Training Set

[]

nan

https://catalog.ldc.upenn.edu/LDC2012T15

LDC User Agreement for Non-Members

2,012

ar

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

news articles

text

other

This release includes 9,693 annotation files in MADCAT XML format (.madcat.xml) along with their corresponding scanned image files in TIFF format.

9,693

documents

Low

LDC

nan

Arab

No

LDC

With-Fee

2,000.00 $

No

machine translation,handwriting recognition

nan

Zaid Alyafeai

2005 NIST Speaker Recognition Evaluation Test Data

[]

nan

https://catalog.ldc.upenn.edu/LDC2011S04

LDC User Agreement for Non-Members

2,011

multilingual

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

transcribed audio

spoken

other

The speech data consists of conversational telephone speech with multi-channel data collected by LDC simultaneously from a number of auxiliary microphones. The files are organized into two segments: 10 second two-channel excerpts (continuous segments from single conversations that are estimated to contain approximately 10 seconds of actual speech in the channel of interest) and five minute two-channel conversations.

525

hours

Low

LDC

nan

Arab

No

LDC

With-Fee

400.00 $

No

speaker identification

nan

Zaid Alyafeai

2005 NIST Speaker Recognition Evaluation Training Data

[]

nan

https://catalog.ldc.upenn.edu/LDC2011S01

LDC User Agreement for Non-Members

2,011

multilingual

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

transcribed audio

spoken

other

The speech data consists of conversational telephone speech with multi-channel data collected simultaneously from a number of auxiliary microphones. The files are organized into two segments: 10 second two-channel excerpts (continuous segments from single conversations that are estimated to contain approximately 10 seconds of actual speech in the channel of interest) and five minute two-channel conversations.

392

hours

Low

LDC

nan

Arab

No

LDC

With-Fee

350.00 $

No

speaker identification

nan

Zaid Alyafeai

2006 NIST Speaker Recognition Evaluation Test Set Part 1

[]

nan

https://catalog.ldc.upenn.edu/LDC2011S10

LDC User Agreement for Non-Members

2,011

multilingual

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

transcribed audio

spoken

other

The speech data in this release was collected by LDC as part of the Mixer project, in particular Mixer Phases 1, 2, and 3. The Mixer project supports the development of robust speaker recognition technology by providing carefully collected and audited speech from a large pool of speakers recorded simultaneously across numerous microphones and in different communicative situations and/or in multiple languages. The data is mostly English speech, but includes some speech in Arabic, Bengali, Chinese, Farsi, Hindi, Korean, Russian, Spanish, Thai, and Urdu.

437

hours

Low

LDC

nan

Arab

No

LDC

With-Fee

300.00 $

No

speaker identification

nan

Zaid Alyafeai

2006 NIST Speaker Recognition Evaluation Training Set

[]

nan

https://catalog.ldc.upenn.edu/LDC2011S09

LDC User Agreement for Non-Members

2,011

multilingual

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

transcribed audio

spoken

other

The speech data in this release was collected by LDC as part of the Mixer project, in particular Mixer Phases 1, 2, and 3. The Mixer project supports the development of robust speaker recognition technology by providing carefully collected and audited speech from a large pool of speakers recorded simultaneously across numerous microphones and in different communicative situations and/or in multiple languages. The data is mostly English speech, but includes some speech in Arabic, Bengali, Chinese, Hindi, Korean, Russian, Thai, and Urdu.

595

hours

Low

LDC

nan

Arab

No

LDC

With-Fee

350.00 $

No

speaker identification

nan

Zaid Alyafeai

2006 NIST Spoken Term Detection Development Set

[]

nan

https://catalog.ldc.upenn.edu/LDC2011S02

LDC User Agreement for Non-Members

2,011

multilingual

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

transcribed audio

spoken

other

The development corpus consists of three data genres: broadcast news (BNews), conversational telephone speech (CTS) and conference room meetings (CONFMTG). The broadcast news material was collected in 2001 by LDCs broadcast collection system from the following sources: ABC (English), China Broadcasting System (Chinese), China Central TV (Chinese), China National Radio (Chinese), China Television System (Chinese), CNN (English), MSNBC/NBC (English), Nile TV (Arabic), Public Radio International (English) and Voice of America (Arabic, Chinese, English). The CTS data was taken from the Switchboard data sets (e.g., Switchboard-2 Phase 1 LDC98S75, Switchboard-2 Phase 2 LDC99S79) and the Fisher corpora (e.g., Fisher English Training Speech Part 1 LDC2004S13), also collected by LDC. The conference room meeting material consists of goal-oriented, small group roundtable meetings and was collected in 2001, 2004 and 2005 by NIST, the International Computer Science Institute (Berkely, California), Carnegie Mellon University (Pittsburgh, PA) and Virginia Polytechnic Institute and State University (Blacksburg, VA) as part of the AMI corpus project.

18

hours

Low

LDC

nan

Arab

No

LDC

With-Fee

800.00 $

No

spoken term detection

nan

Zaid Alyafeai

2006 NIST Spoken Term Detection Evaluation Set

[]

nan

https://catalog.ldc.upenn.edu/LDC2011S03

LDC User Agreement for Non-Members

2,011

multilingual

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

transcribed audio

spoken

other

The evaluation corpus consists of three data genres: broadcast news (BNews), conversational telephone speech (CTS) and conference room meetings (CONFMTG). The broadcast news material was collected in 2003 and 2004 by LDCs broadcast collection system from the following sources: ABC (English), Aljazeera (Arabic), China Central TV (Chinese), CNN (English), CNBC (English), Dubai TV (Arabic), New Tang Dynasty TV (Chinese), Public Radio International (English) and Radio Free Asia (Chinese). The CTS data was taken from the Switchboard data sets (e.g., Switchboard-2 Phase 1 LDC98S75, Switchboard-2 Phase 2 LDC99S79) and the Fisher corpora (e.g., Fisher English Training Speech Part 1 LDC2004S13), also collected by LDC. The conference room meeting material consists of goal-oriented, small group roundtable meetings and was collected in 2004 and 2005 by NIST, the International Computer Science Institute (Berkeley, California), Carnegie Mellon University (Pittsburgh, PA), TNO (The Netherlands) and Virginia Polytechnic Institute and State University (Blacksburg, VA) as part of the AMI corpus project.

18

hours

Low

LDC

nan

Arab

No

LDC

With-Fee

800.00 $

No

spoken term detection

nan

Zaid Alyafeai

2008 NIST Speaker Recognition Evaluation Test Set

[]

nan

https://catalog.ldc.upenn.edu/LDC2011S08

LDC User Agreement for Non-Members

2,011

multilingual

mixed

transcribed audio

spoken

other

The speech data in this release was collected in 2007 by LDC at its Human Subjects Collection facility in Philadelphia and by the International Computer Science Institute (ICSI) at the University of California, Berkeley. This collection was part of the Mixer 5 project, which was designed to support the development of robust speaker recognition technology by providing carefully collected and audited speech from a large pool of speakers recorded simultaneously across numerous microphones and in different communicative situations and/or in multiple languages. Mixer participants were native English and bilingual English speakers. The telephone speech in this corpus is predominantly English, but also includes the above languages. All interview segments are in English. Telephone speech represents approximately 368 hours of the data, whereas microphone speech represents the other 574 hours.

942

hours

Low

LDC

nan

Arab

No

LDC

With-Fee

600.00 $

No

speaker identification

nan

Zaid Alyafeai

2008 NIST Speaker Recognition Evaluation Training Set Part 1

[]

nan

https://catalog.ldc.upenn.edu/LDC2011S05

LDC User Agreement for Non-Members

2,011

multilingual

mixed

transcribed audio

spoken

other

The speech data in this release was collected in 2007 by LDC at its Human Subjects Collection facility in Philadelphia and by the International Computer Science Institute (ICSI) at the University of California, Berkley. This collection was part of the Mixer 5 project, which was designed to support the development of robust speaker recognition technology by providing carefully collected and audited speech from a large pool of speakers recorded simultaneously across numerous microphones and in different communicative situations and/or in multiple languages.

640

hours

Low

LDC

nan

Arab

No

LDC

With-Fee

400.00 $

No

speaker identification

nan

Zaid Alyafeai

2008 NIST Speaker Recognition Evaluation Training Set Part 2

[]

nan

https://catalog.ldc.upenn.edu/LDC2011S07

LDC User Agreement for Non-Members

2,011

multilingual

mixed

transcribed audio

spoken

other

The speech data in this release was collected in 2007 by LDC at its Human Subjects Data Collection Laboratories in Philadelphia and by the International Computer Science Institute (ICSI) at the University of California, Berkeley. This collection was part of the Mixer 5 project, which was designed to support the development of robust speaker recognition technology by providing carefully collected and audited speech from a large pool of speakers recorded simultaneously across numerous microphones and in different communicative situations and/or in multiple languages. Mixer participants were native English speakers and bilingual English speakers. The telephone speech in this corpus is predominately English, but also includes the above languages. All interview segments are in English. Telephone speech represents approximately 523 hours of the data, and microphone speech represents the other 427 hours.

950

hours

Low

LDC

nan

Arab

No

LDC

With-Fee

600.00 $

No

speaker identification

nan

Zaid Alyafeai

2008/2010 NIST Metrics for Machine Translation (MetricsMaTr) GALE Evaluation Set

[]

nan

https://catalog.ldc.upenn.edu/LDC2011T05

LDC User Agreement for Non-Members

2,011

multilingual

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

web pages

text

other

This release contains 149 documents with corresponding reference translations (Arabic-to-English and Chinese-to-English), system translations and human assessments. The human assessments include the following: Adequacy7 (a 7-point scale for judging the meaning of a system translation with respect to the reference translation) Adequacy Yes/No (whether the given system segment meant essentially the same as the reference translation) Preference (the judges preference between two candidate translations when compared to a human reference translation) and HTER (Human Targeted Error Rate, human edits to a system translation to have the same meaning as a reference translation).

149

documents

Low

LDC

nan

Arab

No

LDC

With-Fee

250.00 $

No

machine translation

nan

Zaid Alyafeai

OntoNotes Release 4.0

[]

nan

https://catalog.ldc.upenn.edu/LDC2011T03

LDC User Agreement for Non-Members

2,011

multilingual

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

news articles

text

other

Documents describing the annotation guidelines and the routines for deriving various views of the data from the database are included in the documentation directory of this release. The annotation is provided both in separate text files for each annotation layer (Treebank, PropBank, word sense, etc.) and in the form of an integrated relational database (ontonotes-v4.0.sql.gz) with a Python API to provide convenient cross-layer access.

300,000

tokens

Low

LDC

nan

Arab

No

LDC

Upon-Request

nan

No

information retrieval,information extraction

nan

Zaid Alyafeai

NIST 2002 Open Machine Translation (OpenMT) Evaluation

[]

nan

https://catalog.ldc.upenn.edu/LDC2010T10

LDC User Agreement for Non-Members

2,010

multilingual

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

news articles

text

other

This evaluation kit includes a single perl script (mteval-v09.pl) that may be used to produce a translation quality score for one (or more) MT systems. The script works by comparing the system output translation with a set of (expert) reference translations of the same source text. Comparison is based on finding sequences of words in the reference translations that match word sequences in the system output translation. More information on the evaluation algorithm may be obtained from the paper detailing the algorithm: BLEU: a Method for Automatic Evaluation of Machine Translation (Papineni et al, 2002).

nan

Low

LDC

nan

Arab

No

LDC

With-Fee

150.00 $

No

machine translation

nan

Zaid Alyafeai

NIST 2003 Open Machine Translation (OpenMT) Evaluation

[]

nan

https://catalog.ldc.upenn.edu/LDC2010T11

LDC User Agreement for Non-Members

2,010

multilingual

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

news articles

text

other

This evaluation kit includes a single perl script (mteval-v09c.pl) that may be used to produce a translation quality score for one (or more) MT systems. The script works by comparing the system output translation with a set of (expert) reference translations of the same source text. Comparison is based on finding sequences of words in the reference translations that match word sequences in the system output translation. More information on the evaluation algorithm may be obtained from the paper detailing the algorithm: BLEU: a Method for Automatic Evaluation of Machine Translation (Papineni et al, 2002).

nan

Low

LDC

nan

Arab

No

LDC

With-Fee

150.00 $

No

machine translation

nan

Zaid Alyafeai

NIST 2004 Open Machine Translation (OpenMT) Evaluation

[]

nan

https://catalog.ldc.upenn.edu/LDC2010T12

LDC User Agreement for Non-Members

2,010

multilingual

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

news articles

text

other

This evaluation kit includes a single Perl script (mteval-v11a.pl) that may be used to produce a translation quality score for one (or more) MT systems. The script works by comparing the system output translation with a set of (expert) reference translations of the same source text. Comparison is based on finding sequences of words in the reference translations that match word sequences in the system output translation. More information on the evaluation algorithm may be obtained from the paper detailing the algorithm: BLEU: a Method for Automatic Evaluation of Machine Translation (Papineni et al, 2002).

150

documents

Low

LDC

nan

Arab

No

LDC

With-Fee

150.00 $

No

machine translation

nan

Zaid Alyafeai

NIST 2005 Open Machine Translation (OpenMT) Evaluation

[]

nan

https://catalog.ldc.upenn.edu/LDC2010T14

LDC User Agreement for Non-Members

2,010

multilingual

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

news articles

text

other

This evaluation kit includes a single Perl script (mteval-v11b.pl) that may be used to produce a translation quality score for one (or more) MT systems. The script works by comparing the system output translation with a set of (expert) reference translations of the same source text. Comparison is based on finding sequences of words in the reference translations that match word sequences in the system output translation. More information on the evaluation algorithm may be obtained from the paper detailing the algorithm: BLEU: a Method for Automatic Evaluation of Machine Translation (Papineni et al, 2002).

100

documents

Low

LDC

nan

Arab

No

LDC

With-Fee

150.00 $

No

machine translation

nan

Zaid Alyafeai

NIST 2006 Open Machine Translation (OpenMT) Evaluation

[]

nan

https://catalog.ldc.upenn.edu/LDC2010T17

LDC User Agreement for Non-Members

2,010

multilingual

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

news articles

text

other

This evaluation kit includes a single Perl script (mteval-v11b.pl) that may be used to produce a translation quality score for one (or more) MT systems. The script works by comparing the system output translation with a set of (expert) reference translations of the same source text. Comparison is based on finding sequences of words in the reference translations that match word sequences in the system output translation. More information on the evaluation algorithm may be obtained from the paper detailing the algorithm: BLEU: a Method for Automatic Evaluation of Machine Translation (Papineni et al, 2002).

357

documents

Low

LDC

nan

Arab

No

LDC

With-Fee

150.00 $

No

machine translation

nan

Zaid Alyafeai

NIST 2008 Open Machine Translation (OpenMT) Evaluation

[]

nan

https://catalog.ldc.upenn.edu/LDC2010T21

LDC User Agreement for Non-Members

2,010

multilingual

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

news articles

text

other

This evaluation kit includes a single Perl script (mteval-v11b.pl) that may be used to produce a translation quality score for one (or more) MT systems. The script works by comparing the system output translation with a set of (expert) reference translations of the same source text. Comparison is based on finding sequences of words in the reference translations that match word sequences in the system output translation. More information on the evaluation algorithm may be obtained from the paper detailing the algorithm: BLEU: a Method for Automatic Evaluation of Machine Translation (Papineni et al, 2002).

373

documents

Low

LDC

nan

Arab

No

LDC

With-Fee

150.00 $

No

machine translation

nan

Zaid Alyafeai

NIST 2009 Open Machine Translation (OpenMT) Evaluation

[]

nan

https://catalog.ldc.upenn.edu/LDC2010T23

LDC User Agreement for Non-Members

2,010

multilingual

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

news articles

text

other

This evaluation kit includes a single Perl script (mteval-v11b.pl) that may be used to produce a translation quality score for one (or more) MT systems. The script works by comparing the system output translation with a set of (expert) reference translations of the same source text. Comparison is based on finding sequences of words in the reference translations that match word sequences in the system output translation. More information on the evaluation algorithm may be obtained from the paper detailing the algorithm: BLEU: a Method for Automatic Evaluation of Machine Translation (Papineni et al, 2002).

373

documents

Low

LDC

nan

Arab

No

LDC

With-Fee

150.00 $

No

machine translation

nan

Zaid Alyafeai

NIST Open MT 2008 Evaluation (MT08) Selected References and System Translations

[]

nan

https://catalog.ldc.upenn.edu/LDC2010T01

LDC User Agreement for Non-Members

2,010

multilingual

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

web pages

text

other

120 documents with 1312 segments, output from 17 machine translation systems.

1,312

sentences

Low

LDC

nan

Arab

No

LDC

With-Fee

200.00 $

No

machine translation

nan

Zaid Alyafeai

TRECVID 2006 Keyframes

[]

nan

https://catalog.ldc.upenn.edu/LDC2010V02

LDC User Agreement for Non-Members

2,010

multilingual

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

other

The video stills that compose this corpus are drawn from approximately 158.6 hours of English, Arabic, and Chinese language broadcast programming data collected by LDC from NBC ("NBC Nightly News"), CNN ("Live From..", "Anderson Cooper 360"), MSNBC ("MSNBC News live"), New Tang Dynsaty TV ("Economic Frontier", "Focus Interactive"), Phoenix TV ("Good Morning China"), Lebanese Broadcasting Corp. ("Naharkum Saiid", "News on LBC"), Alhurra TV ("Alhurra News") and China Central TV ("CCTV_News").

158.6

hours

Low

LDC

nan

Arab

No

LDC

With-Fee

500.00 $

No

information extraction,event detection,information retrieval

nan

Zaid Alyafeai

2007 NIST Language Recognition Evaluation Supplemental Training Set

[]

nan

https://catalog.ldc.upenn.edu/LDC2009S05

LDC User Agreement for Non-Members

2,009

multilingual

ar-EG: (Arabic (Egypt))

transcribed audio

spoken

other

The supplemental training material in this release consists of the following:

118

hours

Low

LDC

nan

Arab

No

LDC

With-Fee

500.00 $

No

language identification

nan

Zaid Alyafeai

2007 NIST Language Recognition Evaluation Test Set

[]

nan

https://catalog.ldc.upenn.edu/LDC2009S04

LDC User Agreement for Non-Members

2,009

multilingual

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

transcribed audio

spoken

other

Each speech file in the test data is one side of a 4-wire telephone conversation represented as 8-bit 8-kHz mu-law format. There are 7530 speech files in SPHERE (.sph) format for a total of 66 hours of speech. The speech data was compiled from LDCs CALLFRIEND, Fisher Spanish, and Mixer 3 corpora and from data collected by Oregon Health and Science University (OHSU), Beaverton, Oregon.

66

hours

Low

LDC

nan

Arab

No

LDC

With-Fee

500.00 $

No

language identification

nan

Zaid Alyafeai

2008 NIST Metrics for Machine Translation (MetricsMATR08) Development Data

[]

nan

https://catalog.ldc.upenn.edu/LDC2009T05

LDC User Agreement for Non-Members

2,009

multilingual

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

news articles

text

other

The MetricsMATR08 development data set released here is reflective of the test data set only to a degree; the evaluation data set contains more varied data -- from more genres, more source languages, more systems and different evaluations -- than this development data set. There are also more types of human assessments for the test data. The MetricsMATR08 test data remains unseen to allow for repeated use as test data.

249

sentences

Low

LDC

nan

Arab

No

LDC

With-Fee

150.00 $

No

machine translation

nan

Zaid Alyafeai

Language Understanding Annotation Corpus

[]

nan

https://catalog.ldc.upenn.edu/LDC2009T10

LDC User Agreement for Non-Members

2,009

multilingual

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

transcribed audio

text

other

The resulting corpus contains over 9000 words of English text (6949 words) and Arabic text (2183 words) annotated for committed belief, event and entity coreference, dialog acts and temporal relations.

2,183

tokens

Low

LDC

nan

Arab

No

LDC

Upon-Request

nan

No

pragmatics

nan

Zaid Alyafeai

OntoNotes Release 3.0

[]

nan

https://catalog.ldc.upenn.edu/LDC2009T24

LDC User Agreement for Non-Members

2,009

multilingual

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

news articles

text

other

Each data directory has been stored as a Gnu Zipped Tar File (.tgz) due to the complexity and depth of each directory and the limitations of the ISO CD9660 file system for CD and DVD media. These directories may be easily unpacked using the Unix command line or using utilities such as StuffIt or WinZip under Windows.

200,000

tokens

Low

LDC

nan

Arab

No

LDC

Upon-Request

nan

No

information extraction,information retrieval

nan

Zaid Alyafeai

REFLEX Entity Translation Training/DevTest

[]

nan

https://catalog.ldc.upenn.edu/LDC2009T11

LDC User Agreement for Non-Members

2,009

multilingual

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

news articles

text

other

Please use this link for a sample.

22,500

tokens

Low

LDC

nan

Arab

No

LDC

Upon-Request

nan

No

named entity recognition,information retrieval

nan

Zaid Alyafeai

Unified Linguistic Annotation Text Collection

[]

nan

https://catalog.ldc.upenn.edu/LDC2009T07

LDC User Agreement for Non-Members

2,009

multilingual

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

nan

text

other

Please view this LDC2009T10 sample and LDC2009T11 sample.

22,500

tokens

Low

LDC

nan

Arab

No

LDC

Upon-Request

nan

No

summarization,sociolinguistics,question-answering,psycholinguistics,pragmatics,information retrieval

nan

Zaid Alyafeai

2003 NIST Rich Transcription Evaluation Data

[]

nan

https://catalog.ldc.upenn.edu/LDC2007S10

LDC User Agreement for Non-Members

2,007

multilingual

mixed

transcribed audio

spoken

other

The BN datasets were selected from TDT-4 sources collected in February 2001. The evaluation excerpts were transcribed to the nearest story boundary. The English BN dataset is approximately three hours long and is composed of 30-minute excerpts from six different broadcasts. The Mandarin Chinese BN dataset is approximately one hour long, consisting of 12-minute excerpts from five different broadcasts. The Arabic BN dataset is also approximately one hour long and contains 30-minute excerpts from two different broadcasts.

1

hours

Low

LDC

nan

Arab

No

LDC

With-Fee

2,000.00 $

No

speech recognition

nan

Zaid Alyafeai

GALE Phase 1 Distillation Training

[]

nan

https://catalog.ldc.upenn.edu/LDC2007T20

LDC User Agreement for Non-Members

2,007

multilingual

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

other

text

other

The annotation task involves responding to a series of user queries. For each query, annotators first find relevant documents and identify snippets (strings of contiguous text that answer the query) in the Arabic, Chinese or English source document. Annotators then create a nugget for each fact expressed in the snippet. Semantically equivalent nuggets are grouped into cross-language, cross-document "supernugs". Judges at BAE Systems finally provide relevance weights for each supernug.

81

sentences

Low

LDC

nan

Arab

No

LDC

With-Fee

2,000.00 $

No

topic detection and tracking,metadata extraction,message understanding,information retrieval,information extraction,distillation,information retrieval

nan

Zaid Alyafeai

TRECVID 2005 Keyframes & Transcripts

[]

nan

https://catalog.ldc.upenn.edu/LDC2007V01

LDC User Agreement for Non-Members

2,007

multilingual

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

other

The source data is Arabic, Chinese and English language broadcast programming collected in November 2004 from the following sources: Lebanese Broadcasting Corp. (Arabic); China Central TV and New Tang Dynasty TV (Chinese); and CNN and MSNBC/NBC (English).

nan

Low

LDC

nan

Arab

No

LDC

With-Fee

500.00 $

No

information retrieval,information extraction,event detection

nan

Zaid Alyafeai

2003 NIST Language Recognition Evaluation

[]

nan

https://catalog.ldc.upenn.edu/LDC2006S31

LDC User Agreement for Non-Members

2,006

multilingual

ar-EG: (Arabic (Egypt))

transcribed audio

spoken

other

Each speech file is one side of a "four wire" telephone conversation represented as 8-bit, 8-kHz mulaw data. There are 11,830 speech files in SPHERE (.sph) format. The speech data was compiled from LDC's CALLFRIEND, CALLHOME, and Switchboard-2 corpora. Each file contains one test segment. The test segments are divided into three-second, 10-second, and 30-second tests, each in its own directory.

46

hours

Low

LDC

nan

Arab

No

LDC

With-Fee

500.00 $

No

language identification

nan

Zaid Alyafeai

ACE 2005 Multilingual Training Corpus

[]

nan

https://catalog.ldc.upenn.edu/LDC2006T06

LDC User Agreement for Non-Members

2,006

multilingual

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

other

text

other

Below is information about the amount of data in this release and its annotation status. Further information such as breakdown of genres and formats can be found in the associated README file.

433

documents

Low

LDC

nan

Arab

No

LDC

With-Fee

4,000.00 $

No

information retrieval

nan

Zaid Alyafeai

TDT5 Multilingual Text

[]

nan

https://catalog.ldc.upenn.edu/LDC2006T18

LDC User Agreement for Non-Members

2,006

multilingual

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

other

text

other

The TDT5 corpus spans collections from April-September 2003 of English, Chinese, and Arabic news text. A total of 15 distinct news "sources" are included (where a "source" comprises data from a given news agency in a particular language; when an agency publishes in multiple languages, each language is considered a different "source"). In contrast to earlier TDT corpora, TDT5 has no broadcast/audio content, only printed news from wire and web sources.

407,503

documents

Low

LDC

nan

Arab

No

LDC

With-Fee

1,500.00 $

No

topic classification, language modeling, generation

nan

Zaid Alyafeai

TDT5 Topics and Annotations

[]

nan

https://catalog.ldc.upenn.edu/LDC2006T19

LDC User Agreement for Non-Members

2,006

multilingual

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

news articles

text

other

A total of 250 topics, numbered 55001 - 55250, were annotated by LDC using a search guided annotation technique. Details of the annotation process are described in the annotation task definition.

104

documents

Low

LDC

nan

Arab

No

LDC

With-Fee

500.00 $

No

information detection,information extraction,language modeling,machine translation,topic classification

nan

Zaid Alyafeai

KUNUZ

[]

nan

http://jarir.tn/kunuzcorpus

CC BY-NC-ND 4.0

2,019

multilingual

ar-CLS: (Arabic (Classic))

other

text

manual curation

KUNUZ is an XMLized version of Sahih Albukhari, the most authentic hadith book.

7,563

documents

Low

Joint group for Artificial Reasoning and Information Retrieval

nan

KUNUZ: a Multi-purpose Reusable Test Collection for Classical Arabic Document Engineering

https://ieeexplore.ieee.org/document/9035212

Arab

No

other

Free

nan

No

machine translation, cross-lingual information retrieval, named entity recognition, information retrieval, language identification, document classification, information extraction

AICCSA

2.0

conference

International Conference on Computer Systems and Applications

Ibrahim Bounhas, Souheila Ben Guirat

LISI: Laboratory of Computer science for industrial systems, Carthage University, Tunisia

Corpora are important resources for several applications in Information Retrieval (IR) and Knowledge Extraction (KE). Arabic is a low resourced language characterized by its complex morphology. Furthermore, most existent Arabic language resources focus on Modern Standard Arabic (MSA). This paper describes KUNUZ a multi-purpose test collection composed of voweled and structured classical Arabic documents. Its goal is to provide a unique benchmark for assessing applications in several areas of document engineering including IR, document classification and information extraction. The documents are also translated in English to allow Arabic-English cross-lingual IR and machine translation. As far as IR is concerned, we follow the standard topic development and results sampling used in international campaigns. The paper, describes the process of topic development, results pooling and relevance judgment. It also analyses the results of some processing tools and IR models used in the runs. In order to enhance the results of our experiments, we also proposed to combine the results based on a meta-search approach using Support Vector Machines (SVM) classification.

Jezia Zakraoui

DZDC12

[]

nan

https://github.com/xprogramer/DZDC12

unknown

2,020

ar

ar-DZ: (Arabic (Algeria))

social media

text

crawling

DZDC12 is a multi-purpose parallel corpus crawled from facebook

2,400

sentences

High

Department of Electronics and Telecommunications, Université 8 Mai 1945 Guelma,24000,Guelma,Algeria

nan

DZDC12: a new multipurpose parallel Algerian Arabizi–French code-switched corpus

https://link.springer.com/article/10.1007/s10579-019-09454-8

Arab

No

GitHub

Free

nan

No

machine translation, language modeling, dialect identification, information retrieval, offensive language detection, gender identification, natural language inference, user behavior

LREC

2.0

journal

International Conference on Language Resources and Evaluation

Kheireddine Abainia

PIMIS Laboratory, Department of Electronics and Telecommunications, Université 8 Mai 1945 Guelma,24000,Guelma,Algeria

Algeria’s socio-linguistic situation is known as a complex phenomenon involving several historical, cultural and technological factors. However, there are three languages that are mainly spoken in Algeria (Arabic, Tamazight and French) and they can be mixed in the same sentence (code-switching). Moreover, there are several varieties of dialects that differ from one region to another and sometimes within the same region. This paper aims to provide a new multi-purpose parallel corpus (i.e., DZDC12 corpus), which will serve as a testbed for various natural language processing and information retrieval applications. In particular, it can be a useful tool to study Arabic–French code-switching phenomenon, Algerian Romanized Arabic (Arabizi), different Algerian sub-dialects, sentiment analysis, gender writing style, machine translation, abuse detection, etc. To the best of our knowledge, the proposed corpus is the first of its kind, where the texts are written in Latin script and crawled from Facebook. More specifically, this corpus is organised by gender, region and city, and is transliterated into Arabic script and translated into Modern Standard Arabic. In addition, it is annotated for emotion detection and abuse detection, and annotated at the word level. This article focuses in particular on Algeria’s socio-linguistic situation and the effect of social media networks. Furthermore, the general guidelines for the design of DZDC12 corpus are described as well as the dialects clustering over the map.

Jezia Zakraoui

ADCC

[]

nan

https://adccorpus.wixsite.com/site/single-post/2016/11/30/what-is-adcc

unknown

2,017

ar

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

other

text

other

Arabic Daily Comunication Corpus (ADCC) is daily conversations written text in Modern Standard Arabic which was collected from different resources.

4,000,000

tokens

Medium

nan

Towards Intelligent Arabic Text-to-Speech Application for Disabled People

https://ieeexplore.ieee.org/document/7899133

Arab

No

other

Upon-Request

nan

No

language modeling, topic classification, information retrieval, natural language inference, text-to-speech translation

ICIHT

1.0

conference

Proceeding of International Conference on Informatics, Health & Technology

Amal Alsaif, Njoud Albadrani, Ashwag Alamro, Reham Alsaif

Kingdom of Saudi Arabia Ministry of Education Al-Imam Muhammad Ibn Saud Islamic University

Assistive technology customizes speech technology to offer a new communication channel for disabled people such as blind or having speech difficulties. Converting written text into natural speech has been addressed in the last decades for some languages such as English, hence, used in many applications such as voice answering machines, reading articles and exploring software for blind people. Other languages such as Arabic are still not fully served to have high quality Text-To-Speech applications. This paper describes our effort in developing an intelligent Text-To-Speech mobile application for Arabic. We use a set of statistical language models n-gram for word prediction and auto-completion for easy typing. A large new Arabic corpus for daily communication in different domains is constructed which could be used for other purposes. A serious of normalization processing, including spelling correction, is applied to the corpus to maintain the consistency and unify the occurrence of the same words. We use outsource Sakhr Arabic Text-To-Speeh voices as one of the best speech synthesizer exist for Arabic. To ensure a high usability of the application, we use simple graphical user interface and easy access libraries to favorite phrases with an ability of adding pictures with recorded speech. Our experiments shows that word prediction using global and local corpus decries 50% of keystroke of typing desired sentences with a high prediction of 84% of bigram model.

Jezia Zakraoui

KACST

[]

nan

http://www.kacstac.org.sa

unknown

2,015

ar

mixed

other

text

crawling

The KACST Arabic corpus comprises more than 700 million words from the pre-Islamic era to the present day (a period covering more than 1,500 years), collected from 10 diverse mediums.

7,000,000

tokens

Medium

King Abdulaziz City for Science and Technology University

http://shamela.ws/ http://saaid.net/ http://www.awu.sy/ https://uqu.edu.sa/page/ar/518 http://www.kfu.edu.sa/ar/departments/sjournal/Pages/Home.aspx http://www.boe.gov.sa/MainLaws.aspx?lang=en http://www.arablegalportal.org/ http://www.alwatan.com.sa http://rosa-magazine.com/ http://www.spa.gov.sa/

A 700M+ Arabic corpus: KACST Arabic corpus design and construction

https://link.springer.com/article/10.1007/s10579-014-9284-1

Arab

Yes

other

Free

nan

No

natural language inference

LREC

20.0

journal

International Conference on Language Resources and Evaluation

Al-Thubaity, A.O

King Abdulaziz City for Science & Technology (KACST) university

Compared with English, Arabic is a poorly-resourced language within the field of corpus linguistics. A lack of sufficient data and research has negatively affected Arabic corpus-based researchers and natural language processing practitioners. Although a number of Arabic corpora have been developed in recent years, the overall situation has improved little. The aim of this paper is twofold. First, it reviews 14 Arabic corpora categorized by their designated purpose, target language, mode of text, size, text date, location, text type/medium, text domain, representativeness, and balance. The review also describes the availability of the reviewed corpora, the presence of tokenization, lemmatization and tagging, and whether there are any tools available to search and explore them. Second, it introduces the King Abdulaziz City for Science and Technology (KACST) Arabic corpus, which was designed and created to overcome the limitations of existing Arabic corpora. The KACST Arabic corpus is a large and diverse Arabic corpus with clearly defined design criteria. It is carefully sampled, and its contents are classified based on time, region, medium, domain, and topic, and it can be searched and explored using these classifications. The KACST Arabic corpus comprises more than 700 million words from the pre-Islamic era to the present day (a period covering more than 1,500 years), collected from 10 diverse mediums. Each text has been further classified more specifically into domains and topics. The KACST Arabic corpus is freely available to explore on the Internet (http://www.kacstac.org.sa) using a variety of tools.

Jezia Zakraoui

ArSenTD-LEV

[]

https://huggingface.co/datasets/arsentd_lev

http://oma-project.com/ArSenL/ArSenTD_Lev_Intro

custom

2,019

ar

ar-LEV: (Arabic(Levant))

social media

text

crawling and annotation(translation)

ArSentD-LEV is a multi-topic corpus for target-based sentiment analysis in Arabic Levantine tweets

4,000

sentences

Medium

American University of Beirut

nan

ArSentD-LEV: A Multi-Topic Corpus for Target-based Sentiment Analysis in Arabic Levantine Tweets

https://paperswithcode.com/paper/190601830

Arab

No

other

Free

nan

No

sentiment analysis

LREC

45.0

conference

International Conference on Language Resources and Evaluation

Ramy Baly, Alaa Khaddaj, Hazem Hajj, Wassim El-Hajj, Khaled Bashir Shaban

(1) MIT Computer Science and Artificial Intelligence Laboratory, Cambridge, MA 02139, USA (2) American University of Beirut, Electrical and Computer Engineering Department, Beirut, Lebanon (3) American University of Beirut, Computer Science Department, Beirut, Lebanon (4) Qatar University, Computer Science and Engineering Department, Doha, Qatar

Sentiment analysis is a highly subjective and challenging task. Its complexity further increases when applied to the Arabic language, mainly because of the large variety of dialects that are unstandardized and widely used in the Web, especially in social media. While many datasets have been released to train sentiment classifiers in Arabic, most of these datasets contain shallow annotation, only marking the sentiment of the text unit, as a word, a sentence or a document. In this paper, we present the Arabic Sentiment Twitter Dataset for the Levantine dialect (ArSenTD-LEV). Based on findings from analyzing tweets from the Levant region, we created a dataset of 4,000 tweets with the following annotations: the overall sentiment of the tweet, the target to which the sentiment was expressed, how the sentiment was expressed, and the topic of the tweet. Results confirm the importance of these annotations at improving the performance of a baseline sentiment classifier. They also confirm the gap of training in a certain domain, and testing in another domain

Jezia Zakraoui

CamelTB: Camel Treebank 1.0

[]

nan

http://treebank.camel-lab.com/

custom

2,022

ar

ar-CLS: (Arabic (Classic))

books

text

manual curation

the Camel Treebank (CAMELTB) is a 188K word open-source dependency treebank of Modern Standard and Classical Arabic. It includes 13 sub-corpora comprising selections of texts from pre-Islamic poetry to social media online commentaries, and covering a range of genres from religious and philosophical texts to news, novels, and student essays.

242,000

tokens

Low

NYU Abu Dhabi

nan

Camel Treebank: An Open Multi-genre Arabic Dependency Treebank

http://www.lrec-conf.org/proceedings/lrec2022/pdf/2022.lrec-1.286.pdf

Arab

Yes

CAMeL Resources

Upon-Request

nan

Yes

part of speech tagging, morphological analysis, dependency parsing

LREC

nan

conference

Language Resources and Evaluation Conference

Nizar Habash, Muhammed AbuOdeh, Dima Taji, Reem Faraj, Jamila El Gizuli, Omar Kallas

New York University Abu Dhabi

We present the Camel Treebank (CAMELTB), a 188K word open-source dependency treebank of Modern Standard and Classical Arabic. CAMELTB 1.0 includes 13 sub-corpora comprising selections of texts from pre-Islamic poetry to social media online commentaries, and covering a range of genres from religious and philosophical texts to news, novels, and student essays. The texts are all publicly available (out of copyright, creative commons, or under open licenses). The texts were morphologically tokenized and syntactically parsed automatically, and then manually corrected by a team of trained annotators. The annotations follow the guidelines of the Columbia Arabic Treebank (CATiB) dependency representation. We discuss our annotation process and guideline extensions, and we present some initial observations on lexical and syntactic differences among the annotated sub-corpora. This corpus will be publicly available to support and encourage research on Arabic NLP in general and on new, previously unexplored genres that are of interest to a wider spectrum of researchers, from historical linguistics and digital humanities to computer-assisted language pedagogy.

Nizar Habash

MADAR Twitter Corpus

[ { "Name": "Saudi Arabia", "Dialect": "ar-SA: (Arabic (Saudi Arabia))", "Volume": "1,070", "Unit": "sentences" }, { "Name": "Kuwait", "Dialect": "ar-KW: (Arabic (Kuwait))", "Volume": "213", "Unit": "sentences" }, { "Name": "Egypt", "Dialect": "ar-EG: (Arabic (Egypt))", "Volume": "173", "Unit": "sentences" }, { "Name": "UAE", "Dialect": "ar-AE: (Arabic (United Arab Emirates))", "Volume": "152", "Unit": "sentences" }, { "Name": "Oman", "Dialect": "ar-OM: (Arabic (Oman))", "Volume": "138", "Unit": "sentences" }, { "Name": "Yemen", "Dialect": "ar-YE: (Arabic (Yemen))", "Volume": "136", "Unit": "sentences" }, { "Name": "Qatar", "Dialect": "ar-QA: (Arabic (Qatar))", "Volume": "126", "Unit": "sentences" }, { "Name": "Bahrain", "Dialect": "ar-BH: (Arabic (Bahrain))", "Volume": "113", "Unit": "sentences" }, { "Name": "Jordan", "Dialect": "ar-JO: (Arabic (Jordan))", "Volume": "107", "Unit": "sentences" }, { "Name": "Sudan", "Dialect": "ar-SD: (Arabic (Sudan))", "Volume": "100", "Unit": "sentences" }, { "Name": "Iraq", "Dialect": "ar-IQ: (Arabic (Iraq))", "Volume": "99", "Unit": "sentences" }, { "Name": "Algeria", "Dialect": "ar-DZ: (Arabic (Algeria))", "Volume": "92", "Unit": "sentences" }, { "Name": "Libya", "Dialect": "ar-LY: (Arabic (Libya))", "Volume": "87", "Unit": "sentences" }, { "Name": "Palestine", "Dialect": "ar-PS: (Arabic (Palestinian Territories))", "Volume": "74", "Unit": "sentences" }, { "Name": "Lebanon", "Dialect": "ar-LB: (Arabic (Lebanon))", "Volume": "66", "Unit": "sentences" }, { "Name": "Somalia", "Dialect": "ar-SO: (Arabic (Somalia))", "Volume": "60", "Unit": "sentences" }, { "Name": "Tunisia", "Dialect": "ar-TN: (Arabic (Tunisia))", "Volume": "51", "Unit": "sentences" }, { "Name": "Syria", "Dialect": "ar-SY: (Arabic (Syria))", "Volume": "48", "Unit": "sentences" }, { "Name": "Morocco", "Dialect": "ar-MA: (Arabic (Morocco))", "Volume": "45", "Unit": "sentences" }, { "Name": "Mauritania", "Dialect": "ar-MR: (Arabic (Mauritania))", "Volume": "37", "Unit": "sentences" }, { "Name": "Djibouti", "Dialect": "ar-DJ: (Arabic (Djibouti))", "Volume": "2", "Unit": "sentences" } ]

nan

https://github.com/CAMeL-Lab/CAMeLBERT

unknown

2,018

ar

mixed

social media

text

crawling

A large-scale collection of parallel sentences built to cover the dialects of 25 cities from the Arab World

2,980

sentences

Medium

NYU Abu Dhabi

nan

Fine-Grained Arabic Dialect Identification

https://aclanthology.org/C18-1113.pdf

Arab

No

other

Upon-Request

nan

Yes

machine translation, dialect identification

nan

conference

International Conference on Computational Linguistics

Mohammad Salameh, Houda Bouamor, Nizar Habash

Carnegie Mellon University in Qatar;Carnegie Mellon University in Qatar;New York University Abu Dhabi

Previous work on the problem of Arabic Dialect Identification typically targeted coarse-grained five dialect classes plus Standard Arabic (6-way classification). This paper presents the first results on a fine-grained dialect classification task covering 25 specific cities from across the Arab World, in addition to Standard Arabic – a very challenging task. We build several classification systems and explore a large space of features. Our results show that we can identify the exact city of a speaker at an accuracy of 67.9% for sentences with an average length of 7 words (a 9% relative error reduction over the state-of-the-art technique for Arabic dialect identification) and reach more than 90% when we consider 16 words. We also report on additional insights from a data analysis of similarity and difference across Arabic dialects

Raed Alharbi

ZAEBUC

[]

nan

https://sites.google.com/view/zaebuc/home

CC-BY-NC 4.0

2,022

multilingual

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

other

text

manual curation

the corpus is an annotated Arabic-English bilingual writer corpus comprising short essays by first-year university students at Zayed University in the United Arab Emirates.

33,300

tokens

Medium

New York University Abu Dhabi, Zayed University, UAE

nan

ZAEBUC: An annotated Arabic-English bilingual writer corpus

http://www.lrec-conf.org/proceedings/lrec2022/pdf/2022.lrec-1.9.pdf

Latn

Yes

other

Upon-Request

nan

No

language modelling, language identification, morphological analysis, text error analysis

LREC

nan

conference

European Language Resources Association (ELRA)

Nizar Habash, David Palfreyman

New York University Abu Dhabi, Abu Dhabi, UAE, Zayed University, Abu Dhabi, UAE

We present ZAEBUC, an annotated Arabic-English bilingual writer corpus comprising short essays by first-year university students at Zayed University in the United Arab Emirates. We describe and discuss the various guidelines and pipeline processes we followed to create the annotations and quality check them. The annotations include spelling and grammar correction, morphological tokenization, Part-of-Speech tagging, lemmatization, and Common European Framework of Reference (CEFR) ratings. All of the annotations are done on Arabic and English texts using consistent guidelines as much as possible, with tracked alignments among the different annotations, and to the original raw texts. For morphological tokenization, POS tagging, and lemmatization, we use existing automatic annotation tools followed by manual correction. We also present various measurements and correlations with preliminary insights drawn from the data and annotations. The publicly available ZAEBUC corpus and its annotations are intended to be the stepping stones for additional annotations.

Jezia Zakraoui

AraNPCC

[]

nan

https://archive.org/details/AraNPCC

CC-BY-NC 4.0

2,022

ar

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

news articles

text

crawling

the corpus is a large Arabic newspaper COVID-19 corpus, automatically collected from 88 Arabic newspapers from 12 Arab countries.

7,277,525

documents

Medium

KACST

nan

AraNPCC: The Arabic Newspaper COVID-19 Corpus

http://www.lrec-conf.org/proceedings/lrec2022/workshops/OSACT/pdf/2022.osact-1.4.pdf

Latn

No

other

Free

nan

No

language modelling, topic classification, corpus spatio-temporal analysis

LREC

nan

conference

European Language Resources Association (ELRA)

Abdulmohsen Al-Thubaity, Sakhar Alkhereyf, Alia Bahanshal

The National Center for Data Analytics and Artificial Intelligence King Abdulaziz City for Science and Technology (KACST)

This paper introduces a corpus for Arabic newspapers during COVID-19: AraNPCC. The AraNPCC corpus covers 2019 until 2021 via automatically-collected data from 12 Arab countries. It comprises more than 2 billion words and 7.2 million texts alongside their metadata. AraNPCC can be used for several natural language processing tasks, such as updating available Arabic language models or corpus linguistics tasks, including language change over time. We utilized the corpus in two case studies. In the first case study, we investigate the correlation between the number of officially reported infected cases and the collective word frequency of “COVID” and “Corona.” The data shows a positive correlation that varies among Arab countries. For the second case study, we extract and compare the top 50 keywords in 2020 and 2021 to study the impact of the COVID-19 pandemic on two Arab countries, namely Algeria and Saudi Arabia. For 2020, the data shows that the two countries’ newspapers strongly interacted with the pandemic, emphasizing its spread and dangerousness, and in 2021 the data suggests that the two countries coped with the pandemic.

Jezia Zakraoui

Arabic-ENglish named entities dataset

[]

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

https://github.com/Hkiri-Emna/Named_Entities_Lexicon_Project

CC BY 4.0

2,017

multilingual

mixed

news articles

text

crawling and annotation(translation)

Arabic-ENglish named entities dataset is created using DBpedia Linked datasets and parallel corpus. For annotating NE in monolingual English corpus we used Gate tool. Our approach is based on linked data entities by mapping them to Gate Gazetteers, and then constructing a type-oriented NE base covering person, Location and organization classes. The second task consists of the use of machine translation to translate these entities and then finally, generating our NE lexicon that encloses the list of Arabic entities that match to the English lists.

48,753

tokens

Low

Monastir University - Tunisia, Umm Al-Qura University - Saudi Arabia

nan

Constructing a Lexicon of Arabic-English Named Entity using SMT and Semantic Linked Data

http://iajit.org/PDF//vol.%2014,%20no%206/10491.pdf

Arab-Latn

No

GitHub

Free

nan

No

machine translation, named entity recognition

IAJIT

nan

journal

The International Arab Journal of Information Technology

Emna Hkiri, Souheyl Mallat, Mounir Zrigui and Mourad Mars

nan

Named Entity Recognition (NER) is the problem of locating and categorizing atomic entities in a given text. In this work, we used DBpedia Linked datasets and combined existing open source tools to generate from a parallel corpus a bilingual lexicon of Named Entities (NE). To annotate NE in the monolingual English corpus, we used linked data entities by mapping them to Gate Gazetteers. In order to translate entities identified by the gate tool from the English corpus, we used moses, a Statistical Machine Translation (SMT) system. The construction of the Arabic-English NE lexicon is based on the results of moses translation. Our method is fully automatic and aims to help Natural Language Processing (NLP) tasks such as, Machine Translation (MT) information retrieval, text mining and question answering. Our lexicon contains 48753 pairs of Arabic-English NE, it is freely available for use by other researchers.

Mourad Mars

CIAD: Corpus of Iraqi Arabic Dialect

[]

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

https://github.com/ebady/Iraqi-Arabic-Dialect-Dataset

CC0

2,022

ar

ar-IQ: (Arabic (Iraq))

social media

text

crawling

The corpus has been collected, annotated and made publicly accessible to other researchers for sentiment analysis research.

1,170

sentences

Medium

University of Kerbala, University of Babylon, and Southern Technical University - Iraq

nan

Constructing twitter corpus of Iraqi Arabic Dialect (CIAD) for sentiment analysis

https://ntv.ifmo.ru/file/article/21138.pdf

Arab

No

GitHub

Free

nan

No

sentiment analysis

STJITMO

nan

journal

Scientific and Technical Journal of Information Technologies, Mechanics and Optics

Mohammed M. Hassoun Al-Jawad, Hasanein Alharbi, Ahmed F. Almukhtar, Anwar Adnan Alnawas

University of Kerbala, University of Babylon, and Southern Technical University - Iraq

The number of Twitter users in Iraq has increased significantly in recent years. Major events, the political situation in the country, had a significant impact on the content of Twitter and affected the tweets of Iraqi users. Creating an Iraqi Arabic Dialect corpus is crucial for sentiment analysis to study such behaviors. Since no such corpus existed, this paper introduces the Corpus of Iraqi Arabic Dialect (CIAD). The corpus has been collected, annotated and made publicly accessible to other researchers for further investigation. Furthermore, the created corpus has been validated using eight different combinations of four feature-selections approaches and two versions of Support Vector Machine (SVM) algorithm. Various performance measures were calculated. The obtained accuracy, 78 %, indicates a promising potential application.

Mourad Mars

Negation and Speculation in Arabic Review (NSAR)

[]

nan

https://github.com/amahany/NSAR

unknown

2,022

ar

ar-EG: (Arabic (Egypt))

reviews

text

manual curation

The Negation and Speculation Arabic Review (NSAR) corpus consists of 3K randomly selected review sentences from three well-known and benchmarked Arabic corpora: Large Scale Arabic Book Review (LABR), Large Arabic Multi-domain Resources (LAMR), and Multi-domain Arabic Sentiment Corpus (MASC). It contains reviews from different categories, including books, hotels, restaurants, and other products written in various Arabic dialects. The negation and speculation keywords have been annotated along with their linguistic scope based on the annotation guidelines reviewed by an expert linguist.

3,011

sentences

Low

Ain Shams University

Large Scale Arabic Book Review (LABR), Large Arabic Multi-domain Resources (LAMR), and Multi-domain Arabic Sentiment Corpus (MASC)

Annotated Corpus with Negation and Speculation in Arabic Review Domain: NSAR

https://github.com/amahany/NSAR/blob/main/NSAR_Paper.pdf

Arab-Latn

No

GitHub

Free

nan

No

sentiment analysis, information retrieval, review classification, Negation and Speculation Detection

IJACSA

nan

journal

International Journal of Advanced Computer Science and Applications

Ahmed Mahany, Heba Khaled, Nouh Sabri Elmitwally, Naif Aljohani, Said Ghoniemy

nan

Negation and speculation detection are critical for Natural Language Processing (NLP) tasks, such as sentiment analysis, information retrieval, and machine translation. This paper presents the first Arabic corpus in the review domain annotated with negation and speculation. The Negation and Speculation Arabic Review (NSAR) corpus consists of 3K randomly selected review sentences from three well-known and benchmarked Arabic corpora. It contains reviews from different categories, including books, hotels, restaurants, and other products written in various Arabic dialects. The negation and speculation keywords have been annotated along with their linguistic scope based on the annotation guidelines reviewed by an expert linguist. The inter-annotator agreement between two independent annotators, Arabic native speakers, is measured using the Cohen’s Kappa coefficients with values of 95 and 80 for negation and speculation, respectively. Furthermore, 29% of this corpus includes at least one negation instance, while only 4% of this corpus contains speculative content. Therefore, the Arabic reviews focus more on negation structures rather than speculation. This corpus will be available for the Arabic research community to handle these critical phenomena.

Ahmed Mahany

ArMI: Arabic Misogynistic Dataset

[]

nan

https://github.com/bilalghanem/armi

unknown

2,022

ar

mixed

social media

text

crawling and annotation(other)

Arabic multidialectal dataset for misogynistic language

9,833

sentences

Medium

The ORSAM Center for Middle Eastern Studies, University of Alberta

nan

ArMI at FIRE 2021: Overview of the First Shared Task on Arabic Misogyny Identification

http://ceur-ws.org/Vol-3159/T5-1.pdf

Arab

No

GitHub

Upon-Request

nan

Yes

misogyny identification

FIRE

6.0

conference

Forum for Information Retrieval Evaluation

Hala Mulki, Bilal Ghanem

ORSAM Center for Middle Eastern Studies, Turkey; University of Alberta, Canada

This paper provides an overview of the organization, results and main findings of the first shared task on misogyny identification in Arabic tweets. Arabic Misogyny Identification task (ArMI) is introduced within the Hate Speech and Offensive Content detection (HASOC) track at FIRE-2021. The ArMI task combines two related classification subtasks: a main binary classification subtask for detecting the presence of misogynistic language, and a finegrained multi-class classification subtask for identifying seven misogynistic behaviors found in misogynistic contents. The data provided for this task is a Twitter dataset composed of 9,833 tweets written in modern standard Arabic (MSA) and several Arabic dialects including Levantine, Egyptian and Gulf. ArMI at FIRE-2021 has got a total of 15 submitted runs for Sub-task A and 13 runs for Sub-task B provided by six different teams. The systems introduced by the participants employed various methods including feature-based, neural networks using either classical machine learning techniques, ensemble methods or transformers. The best performing system achieved an F-measure of 91.4% and 66.5% for subtask A and subtask B, respectively. This indicates that misogynistic language detection and misogynistic behaviors identification in Arabic textual contents can be, effectively, addressed using transformer-based approaches.

Hala Mulki

Sanadset 650K: Data on Hadith Narrators

[]

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

https://data.mendeley.com/datasets/5xth87zwb5/4

CC BY 4.0

2,022

ar

ar-CLS: (Arabic (Classic))

books

text

crawling

Sanadset is a full hadith dataset that contains over 650,986 records collected from 926 historical Arabic books of hadith. This dataset can be used for further investigation and classification of hadiths (Strong/Weak), and narrators (trustworthy/not) using AI techniques, and also it can be used as a linguistic resource tool for Arabic Natural Language Processing.

650,000

documents

Low

Abdelmalek Essaâdi University

nan

Sanadset 650K: Data on Hadith Narrators

https://doi.org/10.1016/j.dib.2022.108540

Arab

No

Mendeley Data

Free

nan

No

machine translation, topic classification, text generation, named entity recognition, question answering, information retrieval, natural language inference

Data in Brief

nan

journal

Data in Brief

Mohammed Mghari,Omar Bouras,Abdelaaziz El Hibaoui

nan

The chain of narrators (Sanad) plays a vital role in deciding the authenticity of Islamic hadiths. However, the investigation and validation of such Sanad fully depend on scientists (Hadith Scholars). They ordinarily utilize their acquired knowledge, which in this manner needs a critical sum of exertion and time. Automated Sanad evaluation using machine learning algorithms is the best way to solve this problem. Therefore, a representative Sanad dataset is required. This paper presents a full hadith dataset which is named Sanadset and is made openly accessible for researchers. Sanadset corpus contains over 650,986 records collected from 926 historical Arabic books of hadith. This dataset can be used for further investigation and classification of hadiths (Strong/Weak), and narrators (trustworthy/not) using AI techniques, and also it can be used as a linguistic resource tool for Arabic Natural Language Processing. Our dataset is collected from online Hadith sources using data scraping and web crawling. The main contribution of this dataset is the extraction of narrator chains that were originally present in textual form within Hadith books. Each observation in the dataset contains complete information about a specific hadith, such as (original book, number, Hadith text, Matn, list of narrators, and the number of narrators).

Mohammed Mghari

Arabic news articles

[]

nan

https://webz.io/free-datasets/arabic-news-articles/

unknown

2,016

ar

mixed

news articles

text

crawling

Arabic news articles dataset crawled from the Webz.io API Language category

236,383

documents

High

Webz.io API

nan

Arab-Latn

No

other

Upon-Request

nan

No

Financial Analysis

nan

Webz.io API

nan

Jezia Zakraoui

Arabic Named Entity Gazetteer

[]

nan

https://sourceforge.net/projects/arabic-named-entity-gazetteer/

CC BY 3.0

2,013

ar

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

wikipedia

text

crawling and annotation(other)

A gazetteer of entities curated from Wikipedia.

68,355

tokens

nan

Automatically Developing a Fine-grained Arabic Named Entity Corpus and Gazetteer by utilizing Wikipedia

https://aclanthology.org/I13-1045/

Arab

No

sourceforge

Free

nan

No

named entity recognition, information retrieval

nan

Amr Keleg