Generating the Big Mechanisms Evaluation Corpus

1. BigMech Wiki Instructions + Redundancy

   This link on the Big Mechanisms wiki provides a list of 1,741 PMC id values. Since some of the articles occur in more than one query, only 840 of these are unique. Here is a list of these unique PMC id values.

2. Preliminary Corpora

   Consistent with this list, we have attempted to download these documents to provide to the community as a shared resource. Pending additional bug checking, we now provide this as a resource for the community.

Given the latest lists of the open access xml from PMC, we were able to download 812 of these 840 documents .

1. Corpus Organization

   • We organize directories of the corpus Journal/Year/Volume
   • space characters are replaced with _ in Journal and Volume names.
   • Each article’s files are named according to it’s PubMed ID (pmid)
   • [pmid].pdf - The pdf file
   • [pmid]_pmc.xml - The xml full text

We host the files for this on Amazon, so that they may only be downloaded from links on this site.

• Corpus 1: PDF + XML Files (955.8MB)
• Corpus 2: XML Files (19.9MB)