MEDRank v0.6.5

© 2011 Jorge.R.Herskovic@uth.tmc.edu

Released under the GPL v2 (see http://www.gnu.org/licenses/gpl-2.0.html or the GPL-2.0.txt file included in this distribution).

MEDRank is a ranking system meant to work on biomedical data, mostly UMLS 
concepts. It also has facilities for performing basic IR evaluation, and one 
semantically-aware vector cosine comparison (see MEDRank/evaluation).

Requires Python 2.6 or 2.7 and BioPython (www.biopython.org). 
Parts work under jython 2.5.2, at least the Node, Link, and Graph classes, but 
it hasn't been tested or used extensively under jython.

Documentation will be forthcoming. MEDRank is meant to be a library you use to
build your own rankers and evaluators. There is one example available in the
examples/ directory that implements summarization as described by Mihalcea in 
2004 (Mihalcea and Tarau. TextRank: Bringing Order into Texts. Proceedings of
the Conference on Empirical Methods in Natural Language Processing (EMNLP 2004))
using MEDRank. I also provide one of my own blog posts in a suitable format to
test the code with (examples/blog_post.txt)

More sample code will be provided eventually. In the meantime, please contact me
if you need assistance.

If you want to experiment with the biomedical literature you'll need to get a 
copy of the MeSH trees (all files in the ASCII format) from 
http://www.nlm.nih.gov/mesh/filelist.html 

CHANGELOG
v0.6.5 
	- Fixed the GraphML import code - now you can roundtrip a file to yED.  
	  MEDRank will read it correctly, even if you modify it in yED.
	Also, small bugfixes.
v0.6.4
    - Small bugfixes
    - New Spreading Activation ranking function
    - Added a very basic networkx export function to the Graph class
v0.6.3
    - Added pickle support for the Node and Link classes
    - Added graphml export and import methods to the Graph class
    - Modified the logger module to remove the process name from the formatter
      under "no-multiprocessing" situations (likely to be jython)
    - Added a couple of utility output functions 
v0.6.2
    - Fixed a bug in naming output columns that could result in a NamedResultSet
      ignoring the column_name property of a subclass.
v0.6.1
    - Fixed a bug in link hashing and adirectional link hashing
    - Logging improvements in default workflows
    
v0.6.0
    - Bug fixes galore
    - Multithreaded locking for SQLite Dictionaries
    - Separate concept and PubMed article server processes (requires web.py from http://webpy.org)

RELEASE NOTES
    Before starting a process that attempts to use the PubMed article cache or the Concept cache, you must start them individually. The scripts are in MEDRank/pubmed/pmid_cache.py and MEDRank/umls/concept_cache.py. They still have many rough edges.