Demo for Elasticsearch v.7.4.0 (might run on subsequent 6.x releases with some modification to the setting json file; does not work on Elasticsearch 0.x, 1.x, 2.x, and 5.x).
All documents and qrels are property of the National Institute of Standards and Technology and have been released from research purposes as part of TREC.
- Python 3 (if you're still using Python 2, I strongly encourage you to switch. Check out this page for why I suggest the switch).
- A fairly recent JVM version. At the time of writing, "fairly recent" equals to Oracle JVM v.1.7u55 or above, or OpenJDK v.1.7.55 or above. Check this page for more infos about requirements for Elasticsearch.
- Elasticsearch binaries. You can get them from the official website.
- Optional: a C compiler. I tried to include pre-compiled versions of
trec_evalin thebinfolder (macOS, Debian/Ubunutu, and Windows); This program will try to use the appropriate one based on the platform your running it. If it fails, please download the source code fortrec_evalfrom the NIST website and compile it by yourself (it should be as easy as navigating to the deflated directory and typingmake). Then copy the compiled binary to the bin folder.
- Install requirements in
requirements.txt; that is, runpip3 install -r requirements.txt. - Download the data files (documents, queries, and qrels) from this page. Unzip them in the root of this project (i.e., where this file is).
- Start Eilasticsearch. Assuming that you have unzipped Elasticsearch to the folder where this file is located, you can execute
./elasticsearch-6.2.1/bin/elasticsearchif you are on a UNIX system, or./elasticsearch-6.2.1/bin/elasticsearch.batif you are on Windows. For more information on how to install and run Elasticsearch, please visit this page. - Execute
index.py. This will index the collection. - Execute
search.py. This will search the collection and evaluate the results.