minor fixes

The corpus now has 250 publications with open access PDFs.

Changes include:

• fixed bug that caused some dataset references to be duplicated
• added more publications (mostly related to NHANES)

This intended for evaluation, prior to launching the Rich Context leaderboard competition.