Skip to content

Latest commit

 

History

History
56 lines (39 loc) · 1.87 KB

PIPELINE.md

File metadata and controls

56 lines (39 loc) · 1.87 KB
Raw

Table of Contents generated with DocToc

Running PropS as part of a pipeline

PropS now supports pipelining from a dependency parser which produces output conforming with Stanford's CORENLP json output format1. This allows for an independent, light-weight, and hopefully easier to install version of PropS.

Installation

The prerequisites are the python packages in requirements.txt.

Run

pip install -r ./props/install/requirements.txt

Usage

Use the new command line flag --corenlp-json-input and supply the json input either from STDIN or in file name as the first argument.

For example, try:

python props/applications/parse_props.py sample.json -t --oie --corenlp-json-input

See all command line options in parse_props.py

Piping with CORENLP

We supply a pipeline script to interact with the CORENLP parser, assuming it is already installed, and that the CORENLP_HOME environment variable points to the CORENLP home folder containing all its jars.

To use it on a sample raw input file, try:

./run_pipeline.sh sample.txt

NOTE:

  • We use the Stanford dependency format (not Universal Dependencies)
  • We use the makeCopulaHead flag

1Specifically, we read the following keys from the json:

  • tokens
  • basicDependencies