TDT4 Multilingual Text and Annotations

Item Name: TDT4 Multilingual Text and Annotations
Author(s): Stephanie Strassel, Junbo Kong, David Graff
LDC Catalog No.: LDC2005T16
ISBN: 1-58563-339-9
ISLRN: 114-628-220-295-1
Release Date: May 15, 2005
Member Year(s): 2005
DCMI Type(s): Text
Data Source(s): broadcast news, newswire
Project(s): GALE, TDT, TIDES
Application(s): topic detection and tracking
Language(s): English, Standard Arabic, Mandarin Chinese
Language ID(s): eng, arb, cmn
License(s): LDC User Agreement for Non-Members
Online Documentation: LDC2005T16 Documents
Licensing Instructions: Subscription & Standard Members, and Non-Members
Citation: Strassel, Stephanie, Junbo Kong, and David Graff. TDT4 Multilingual Text and Annotations LDC2005T16. Web Download. Philadelphia: Linguistic Data Consortium, 2005.
Related Works: View


TDT4 Multilingual Text and Annotations was developed by the Linguistic Data Consortium (LDC) with support from the DARPA TIDES (Translingual Information Detection, Extraction, and Summarization) Program. This release contains the complete set of English, Arabic, and Chinese news text (broadcast news transcripts and newswire data, approximately 100,000 documents, 91,000 being news stories) used in the 2002 and 2003 Topic Detection and Tracking (TDT) technology evaluations, along with approximately 114,000 topic annotations created for those evaluations.

The audio corresponding to the broadcast news transcripts in this release can be found in TDT4 Multilingual Broadcast News Speech Corpus (LDC2005S11).

TDT refers to automatic techniques for finding topically related material in streams of data such as newswire and broadcast news. Evaluation tasks in 2002 and 2003 included the segmentation of a news source into stories, the tracking of known topics, the detection of unknown topics, the detection of initial stories on unknown topics, and the detection of pairs of stories on the same topic.


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).

Multiple manual annotations have been applied to the TDT4 data. Briefly, these include:

  • Transcription of audio data
  • Manual segmentation of audio data into individual story units and time
  • Alignment of audio with transcripts
  • Topic selection
  • Topic definition and research
  • Search guided topic relevance annotation
  • Adjudication of relevance judgments against system output

The news stories are presented in multiple formats, including TIPSTER-style SGML formatting, individually separated tokens, and machine translation and automatic speech recognition outputs. There are also files showing topic relevance decisions for each news story.


For an example of the data in this corpus, please view this sample (TXT). This sample is an English translation from an Arabic news broadcast. The translation is the product of the IBM Arabic to English translation engine.


None at this time.

Available Media

View Fees

Login for the applicable fee