PropS offers an output representation designed to explicitly and uniformly express much of the proposition structure which is implied from syntax.
Semantic NLP applications often rely on dependency trees to recognize major elements of the proposition structure of sentences. Yet, while much semantic structure is indeed expressed by syntax, many phenomena are not easily read out of dependency trees, often leading to further ad-hoc heuristic post-processing or to information loss. For that end, PropS post-processes dependency trees to present a compelling representation for downstream tasks.
Find more details, examples, and an online demo at the project page.
Run sudo -E python ./setup.py install from the props root directory.
This will install several python packages and other resources which PropS uses and relies upon (see requirements.txt and install.sh for the complete list).
MacOS users might run into issues installing JPype. An instruction to manually install JPype on MacOS can be found on the berkely parser python interface repository.
vj: To install and run on my MacOs I had to do these things
- First ensure you are working in a Python 2.7. For this, given that i am using
conda 3.6I had to this:
conda create --name python27 python=2.7.13
source activate python27
- Comment out the line
sed -i s/six==1.5.2/six==1.10.0/ requirements.txtin install/instal.sh (apparently MacOs does not support-ion sed?!). Instead execute what it says oninstall/requirements.txt. (The updatedrequirements.txtis committed.) - Download
JPypemanually, (comment out itswgetininstall/install.sh) and edit itssetup.pyto add the following lines
#I added this line below. The folder contains a jni.h
"/System/Library/Frameworks/JavaVM.framework/Versions/A/Headers/"
at the end of setupInclusion(self) in its setup.py. Now rerun sudo -E pythoon ./setup.py install
- I was also forced to install a legacy Java 1.6 when I attempted to run the unit test below.
- python 2.7
- java 7 (make sure to set the JAVA_HOME enviroment variable (e.g., /usr/lib/[your_java_folder])
Run python ./unit_tests/sanity_test.py