Converts PubMed NXML format into (almost) raw text to be used for NL analysis.

This software assumes a unix-like environment and that you have a working installation of nxml2txt added to your path. Besided that you only need Python.

1. Install MacPorts from: https://www.macports.org
2. Install python 2.7 and set it as the default:

sudo port install python27
sudo port select --set python python27

3. Install the necessary dependencies for nxml2txt:

sudo port install texlive-latex texlive-latex-recommended texlive-latex-extra py-lxml

4. Install nxml2txt and add the binary to your $PATH:

git clone https://github.com/spyysalo/nxml2txt.git
cd nxml2txt
chmod 755 nxml2txt nxml2txt.sh
export PATH=<PATH WHERE nxml2txt IS INSTALLED>:$PATH

5. Install this project:

git clone https://github.com/sistanlp/nxml2fries.git

./nxml2fries [--no-citations] arg1.nxml [... argn.nxml]

• --no-citation: If enabled, this option removes the reference citations from the text and replaces the space they used by white-spaces characters.

• argn.nxml: The nxml file or list of files to operate over.

Each output file is in tab-separated-values format. Its fields are:

• Paragraph ID: A unique id in the document assigned to the corrent paragraph
• Section ID: The id of the section in the paper.*
• Normalized section name: A normalized version of the section to create equivalence classes of sections between papers. For example, Materials/Methods and Materials and methods would have the same normalized name materials-methods.*
• Is title: Wether the text in the line is the title of a section/paper/figure. 1 for true and 0 for false.
• Text: The text of the current paragraph/figure/reference.

* If this field has no information, it's content will be N/A.

The title for section s2f.

A paragraph of section-id s2g, with a normalized section.

One of the references of a paper

A figure's title