This is the README file for Ice, a rapid information extraction customizer.
Ice is licensed under the Apache 2.0 license.
The easiest way to use Ice is to use the binary distribution.
To use the binary distribution, download the binary distribution and unzip the package. In runice.sh, point both $ICE_HOME and $ICE_LIB_HOME to directory of the binary distribution. Both variables are set to . by default.
Then, from the working directory, run
./runice.sh
Ice bundles a relation tagger based on Jet, which tags mentions of relations in text files, using the models that you build with Ice. Note that before the Ice tagger can find the relations, you have to use Export in Ice to export them to the underlying Jet tagger.
To run the tagger, from the working directory, run
./runtagger.sh propertyFile txtFileList apfFileList
where propertyFile is the Jet properties file. We suggest that you use tagprops that is delivered with this package. If you are familiar with Jet and Ice, you can use your own properties file too. txtFileList are the list of text input files, and apfFileList is the list of output files in Ace apf format. Both file lists assume the "one-file-name-per-line" format and should have the same length. The output file corresponds to the input file at the same line number.
We assume that you have git and maven installed on your system.
Please run:
mvn package
If everything works, you should find ICE-0.2.0-jar-with-dependencies.jar (the fatjar), and ICE-0.2.0.jar in target/
Ice relies on Jet and its models. We provide the Jet binary and necessary models in the binary distribution. However, if you are building from source, you might also want to obtain Jet from: http://cs.nyu.edu/grishman/jet/jet.html
The current version of Ice assumes that it is run from a "working directory", where three
Jet property files are located: parseprops, onomaprops, and tagprops. These three files
tell Ice where models for Jet are located. These files are released together with the
Java source code in the src/props
directory.
In theory, Jet model files can sit anywhere. However, to use the property files directly,
you can copy data/
and acedata/
directories from Jet into the working directory.
In addition, Ice itself uses two configuration files: ice.yml and iceprops, which should be put in the working directory as well.
After these steps, the working directory we have prepared will look like this:
working_dir/
parseprops - Jet property file 1
onomaprops - Jet property file 2
tagprops - Jet property file 3
ice.yml - Ice configuration file 1
iceprops - Ice configuration file 2
data/ - model files, including parseModel.gz
acedata/ - model files
With these files, we should be ready to go.
The easiest way to run Ice is to run from the working directory we prepared in the previous section.
Copy src/scripts/runice.sh to the working directory. In runice.sh, point $ICE_HOME to the directory containing ICE-0.2.0-jar-with-dependencies.jar (target/), and $ICE_LIB_HOME to the directory containing Jet-1.8.0.11-ICE-jar-with-dependencies.jar (lib/).
Then, from the working directory, run
./runice.sh
Please refer to iceman for usage.