Skip to content

Latest commit

 

History

History
44 lines (31 loc) · 1.57 KB

README.md

File metadata and controls

44 lines (31 loc) · 1.57 KB
Raw

This system is specfically designed to prepare data for the 'Science Discourse Tagger' https://github.com/BMKEG/sciDT orignally developed by Ed Hovy's group at CMU. These are UIMA pipelines that process open access articles from PubMed Central to generate split-clause data to be processed by sciDT. We also provide scripts to facilitate gathering training data from human annotations and export functions to our preferred output formats.

Instructions for Mac

  1. Install the necessary dependencies for nxml2txt:
sudo port install texlive-latex texlive-latex-recommended texlive-latex-extra py-lxml
  1. Install nxml2txt and add the binary to your $PATH:
git clone https://github.com/spyysalo/nxml2txt.git
cd nxml2txt
chmod 755 nxml2txt nxml2txt.sh
  1. Clone and build this library
git clone https://github.com/BMKEG/sciDT-pipeline
cd sciDT-pipeline
mvn -DskipTests clean assembly:assembly

This will build a fully assembled jar file here:

  1. Running the system is best performed using the provided shell script that executes the edu.isi.bmkeg.sciDT.bin.SciDT_0_Nxml2SciDT class.
./runPipeline /path/to/folder/ #nThreads /path/to/nxml2txt/executable

Where #nThreads is the number of threads we want to run the preprocessing pipeline on.

This should run to generate a number of files in subfolders. These are:

  • nxml2txt
  • bioc
  • preprocessed_bioc_results
  • scidt
  • tsv

The input files for the main sciDT system are in (A) the scidt folder and (B) the tsv folder