This repo holds a collection of tools for the TREC Microblog tracks, which officially ended in 2015. The track mailing list can be found at firstname.lastname@example.org.
- TREC 2013 API Specifications
- TREC 2013 Track Guidelines
- TREC 2014 Track Guidelines
- TREC 2015 Track Guidelines
The Microblog tracks in 2013 and 2014 used the "evaluation as a service" (EaaS) model, where teams interact with the official corpus via a common API. Although the evaluation has ended, the API is still available for researcher use.
To request access to the API, follow these steps:
- Fill out the API usage agreement.
- Email the usage agreement to
- After NIST receives your request, you will receive an access token from NIST.
- The code for accessing the API can be found in this repository. The endpoint of API itself (i.e., hostname, port) will be provided by NIST.
The main Maven artifact for the TREC Microblog API is
twitter-tools-core. The latest releases of Maven artifacts are available at Maven Central.
You can clone the repo with the following command:
$ git clone git://github.com/lintool/twitter-tools.git
Once you've cloned the repository, change directory into
twitter-tools-core and build the package with Maven:
$ cd twitter-tools-core $ mvn clean package appassembler:assemble
For more information, see the project wiki.
Replicating TREC Baselines
One advantage of the TREC Microblog API is that it is possible to deploy a community baseline whose results are replicable by anyone. The
raw results are simply the output of the API unmodified. The
baseline results are the
raw results that have been post-processed to remove retweets and break score ties by reverse chronological order (earliest first).
To run the
raw results for TREC 2011, issue the following command:
sh target/appassembler/bin/RunQueriesThrift \ -host [host] -port [port] -group [group] -token [token] \ -queries ../data/topics.microblog2011.txt > run.microblog2011.raw.txt
And to run the
baseline results for TREC 2011, issue the following command:
sh target/appassembler/bin/RunQueriesBaselineThrift \ -host [host] -port [port] -group [group] -token [token] \ -queries ../data/topics.microblog2011.txt > run.microblog2011.baseline.txt
trec_eval is included in
twitter-tools/etc (just needs to be compiled), and the qrels are stored in
twitter-tools/data (just needs to be uncompressed), so you can evaluate as follows:
../etc/trec_eval.9.0/trec_eval ../data/qrels.microblog2011.txt run.microblog2011.raw.txt
Similar commands will allow you to replicate runs for TREC 2012 and TREC 2013. With
trec_eval, you should get exactly the following results:
Licensed under the Apache License, Version 2.0.
This work is supported in part by the National Science Foundation under award IIS-1218043. Any opinions, findings, and conclusions or recommendations expressed are those of the researchers and do not necessarily reflect the views of the National Science Foundation.