Topic Detection and Tracking

TDT Phase 2

• Information on the current release


• Description of data sources

Provides details on each of the six news sources contained in the TDT2 Corpus.
• Corpus components and data formats

Explains the various types of data files comprising the corpus, how they are organized and formatted.
• Evaluation Plan and Software (@NIST)

Links to information and resources for the TDT2 Benchmark Test.
• Latest Version of TDT2-Multilangage Topic Tables (v3.3)
• Topic Lists

Note that these pages were developed for use by LDC annotators.  Topic ID numbers and Rules of Interpretation on these pages reflect internal annotation practices in place at the time the topics were developed. Also note that the concept of "Rules of Interpretation" was introduced to TDT after the TDT2 training topics (first 37 topics) had already been selected, defined and annotated.

Mapping from original Rules of Interpretation to current rules of interpretation is presented here. Original rules of interpretation are presented in the TDT2 annotation guide. Revised rules of interpretation are presented in the TDT4 annotation guide.

•  Training Set
•  Development Set
•  Evaluation Set
• TDT2 Email archive of tdt-distrib at ldc.upenn.edu

All the announcements and discussions regarding TDT2 that were posted to the project's email distribution list.

The current mail listserv for the TDT project is hosted at the University of Massachussetts and moderated by James Allan. See http://ciir.cs.umass.edu/mailman/listinfo/tdt-list for general information about the list; there is a form on that web page for subscribing to the list.

• Data Production Guide

Guidelines developed and used by LDC staff for segmentation and annotation of TDT-2 data.