Arabic Gigaword Third Edition


Item Name: Arabic Gigaword Third Edition
Authors: David Graff
LDC Catalog No.: LDC2007T40
ISBN: 1-58563-460-3
Release Date: Nov 20, 2007
Data Type: text
Data Source(s): newswire
Project(s): GALE
Application(s): information retrieval, language modeling, natural language processing
Language(s): Modern Standard Arabic
Language ID(s): arb
Distribution: 1 DVD
Member fee: $0 for 2007 members
Non-member Fee: US $4000.00
Reduced-License Fee: US $2000.00
Extra-Copy Fee: US $200.00
Non-member License: yes
Online documentation: yes
Licensing Instructions: Subscription & Standard Members, and Non-Members
Citation: David Graff
2007
Arabic Gigaword Third Edition
Linguistic Data Consortium, Philadelphia

Introduction

Arabic Gigaword Third Edition is a comprehensive archive of newswire text data acquired from Arabic news sources by the LDC at the University of Pennsylvania. Arabic Gigaword Third Edition includes all of the content of Arabic Gigaword Second Edition (LDC2006T02) as well as new data collected after the publication of that edition. Also, an archive from a new newswire source -- Assabah -- has been included in the third editon.

The six distinct sources of Arabic newswire represented in the third edition are:

  • Agence France Presse (afp_arb)
  • Assabah (asb_arb)
  • Al Hayat (hyt_arb)
  • An Nahar (nhr_arb)
  • Ummah Press (umh_arb)
  • Xinhua News Agency (xin_arb)

The seven-character codes in the parantheses above consist of the three-character source name IDs and the three-character language code ("arb") separated by an underscore ("_") character.

The epochs and document counts for the data in the third edition are set forth below:

Newly Added Data
Source Date Span Document Count
Agence France Presse 2005.01 - 2006.12 137815
Assabah News Agency 2004.09 - 2006.12 15410 (new source)
Al Hayat News Agency 2005.01 - 2006.1 8799 (no data for 2004)
An Nahar News Agency 2005.01 - 2006.12 104950 (no data for 2004)
Xinhua News Agency 2005.01 - 2006.12 135472

Data

This release contains 547 files, totalling approximately 1.8GB in compressed form (6,673 MB uncompressed) and 1,994,735 K-words.

The table below shows data quantity by source under the following categories: data source (Source); the number of files per source (#Files); compressed file size (Gzip-MB); uncompressed file size (Totl-MB); the number of space-separated words tokens in the text (K-words) and the number of documents per source (#DOCs).

Data Sources and Quanities
Source #Files Gzip-MB Totl-MB K-wrds #DOCs
afp_arb 152 441 1806 147612 798436
asb_arb 28 23 77 6587 15410
hyt_arb 142 559 1932 171502 378353
nhr_arb 134 612 2172 193732 449340
umh_arb 24 4 14 1201 4645
xin_arb 67 171 672 56165 348551
TOTAL 547 1810 6673 576799 1994735

All text files in this corpus have been converted to UTF-8 character encoding.

Certain data and formatting issues observed in previous releases of Arabic Gigaword have been normalized in the third edition:

  • Approximately 15,000 stories from older AFP files (1994 - 2002) contained very brief documents where the text content was not recognized as such; in those cases, the TEXT element appeared empty while the HEADLINE element contained anywhere from three to several lines of text. The content of these documents has been rearranged. The first line remains as the headline and the rest of the lines have been moved into the text segment. All stories of this sort had been originally classified as "other", and that classification has not been changed in this edition.
  • Al Hayat data from 2002 and 2003 contained some Arabic-Indic digits, despite the intention to convert all digit strings to the ASCII digit characters for consistency. The digits have now been converted to the ASCII range. For more details about the encoding challenges presented by this data, see the readme file accompanying this corpus.
  • Some Al Hayat data had stray angle-bracket characters (""), which have been rendered as "". There were also some defective "Doc-ID" strings (the 'id' attribute in the "" tag that begins each news story) in the January 2001 data.
  • Some An Nahar data had "bare" ampersand characters ("&") which have been rendered as "&".
  • Some Xinhua documents included empty sub-elements (HEADLINE, DATELINE and/or TEXT sections containing no data); when HEADLINE or DATELINE were empty, these tags were removed. When the TEXT segment was empty, the document as a whole was removed.
  • In several Xinhua stories, the Doc-ID string, which is supposed to provide the year, month, date and sequence number for the story, had become garbled, yielding an incorrect or impossible date string. A separate data file in the "docs" directory, called "docid_changes.txt", lists the changes in document inventory and Doc-ID strings.
  • Xinhua stories typically end with a formulaic Arabic string (meaning "end-of-story"), which should not have been included as part of the final paragraph in each story.
  • In general, consistent line-wrapping was applied to make the overall text presentation consistent across all sources and with Gigaword releases in other languages. The markup pattern was also applied consistently for all sources without exception.
  • Samples

    For an example of the data contained in this corpues, please view this image of sample text

    Sponsorship

    This work was supported in part by the Defense Advanced Research Projects Agency, GALE Program Grant No. HR0011-06-1-0003. The content of this publication does not necessarily reflect the position or the policy of the Government, and no official endorsement should be inferred.

    Content Copyright

    Portions 1994-2006 Agence France Presse, 2004-2006 Assabah, 1994-2003, 2005-2006 Al Hayat, 1995-2006 An Nahar, 2003-2004 Ummah Press Service, 2001-2006 Xinhua News Agency, 2003, 2005-2007 Trustees of the University of Pennsylvania