In order to provide reproducibility of IR experiments you can run it within docker containers. This repository provides some data sets within single container which you can attach to your own containers.
Now, this image includes following raw datasets:
LISA
, NPL
, CACM
, CISI
, Cranfield
, Time
, Medline
, Adi
.
All DataSets pulled from ir.dcs.gla.ac.uk
All datasets located in /usr/data/
in separate folders.
As a base image used ubuntu:10.04
because it is smallest available.
Result container size is about 210 MB
, ubuntu:10.04
is about 183 MB
.
Build is simple as:
git clone https://github.com/krasnobaev/docker-irdatasets.git
cd docker-irdatasets
docker build -t krasnobaev/irdatasets .
Or, instead you can pull latest automatically builded image from docker.com:
docker pull krasnobaev/irdatasets
Now you should start container as a Data Volume Container:
docker run -d -v /usr/data --name irds -t krasnobaev/irdatasets \
echo DataSets for IR experiments
After that, attach volume to your application:
docker run -i --volumes-from irds -t ubuntu ls -l /usr/data/
- Provide some parser and scripts to convert datasets to some common data formats, like XML or JSON.
- Include some other datasets.