- TREC 2016
-
Track Web Page:
trecrts.github.io
The TREC 2016 tasks required participants to monitor the public Twitter API in real time. Thus no document set is available for this task. The interest profiles (i.e., topics) were provided to participants using a REST API, with more profiles than could be evaluated included in the test.
See the guidelines page for the whole set of profiles (and the source of the profiles).
Relevance judgments from NIST assessors for evaluation set (used by batch evaluation)
Clusters of equivalent relevant tweets as created by assessors (format is topic-id cluster-id tweet-id)
Clusters of equivalent relevant tweets in JSON format
Gloss of cluster intent (format is topic-id cluster-id gloss)
Epoch time file
Scenario A mobile assessor judgments
Evaluation script for Scenario A runs using mobile assessors
Evaluation script for Scenario A batch evaluation
Evaluation script for Scenario B batch evaluation
|
|
