I'm working to import a medical taxonomy. Prior to going this route, I've had a medical group manually generate a categorization scheme using 'hierarchical select.' Unfortunately, they don't really do it ... which is understandable, but which also REALLY undermines some of the power built in w/ other drupal modules (in particular notification system and faceted search, etc).

I've been so excited about leveraging some of these tools like opencalais to map taxonomy terms that they naturally use on to an existing taxonomy. I've looked at MESH, NLMS, and OpenGalen as sources for that taxonomy library. The problem I'm running into as an intermediate (probably a novice user by many of your standards who have been doing this for awhile) is that UMLS: http://www.nlm.nih.gov/research/umls/index.html has its own organizational schema and format. Dman - I am so excited about what you're doing, especially what you did with MESH, but in this case, because MESH is flat and doesn't correspond with how doctors actually talk, it unfortunately didn't suit my purposes in this case. OpenGalen: http://www.opengalen.org/ seems like a good solution, but I'm having trouble importing a schema. It's structured in OWL 2.0 and although I'm able to browse it with Protege, I cannot find a format with which to export it and import it in w/ Taxonomy XML importer.

I'm sorry to clog up this forum this way, but should anyone have perspective, thoughts, experience or whatever on my course of action, I would sincerely appreciate it.

Comments

Equinger’s picture

Equinger commented
Title: Ideas about importing/how to import a medical taxonomy » Ideas about importing a medical taxonomy
dman’s picture

dman commented

UMLS is indeed its own little format. It suits a certain need and all, but it's silos of syntax like this that made me decide to put my weight behind RDF. If someone has a syntax, it can probably be ported to RDF. The Galen implementation seems to bear that out
If it can be ported to RDF, I can probably ingest it.

When I looked at MeSH, it was anything but flat, but I'm sure you're right that it may not apply to how doctors talk - It was developed for the research journals, a taxonomy for pubmed.

Looking at the description of Galen, It probably covers a range of elements that taxonomy_xml should be able to take - owl:class, rdfs:subClassOf etc, although a large amount of the OWL detail and logic will be left out, you should at least be able to extract a glossary out of it.

The problem - as with Mesh - is that at several hundred Megs big, it's a bit too huge to process with a web process without some magic.

Equinger’s picture

Equinger commented

Thanks a lot dman! Yeah, I'm plugging away at this - exported OpenGalen as RDF/XML through Protege. Because of its size, Protege has to work locally w/ 1gb Java heap.

I have the RDF here w/ OpenGalen's OWL files referenced for import from the web root:

http://wiki.cepcolorado.com/beastly_med.rdf

The taxonomy import stops almost immediately w/out any errors or any import unfortunately.

Let me know if you have any other thoughts/suggestions. I utilized your Mesh server and script yesterday and it worked great(Thanks!)...but I'm assuming that I won't be able to utilize it for this though, huh?

With regard to MESH ... I only mean flat w/ respect to whatever levels I was concerned with, so yeah, not really fair.

Equinger’s picture

Equinger commented

quick fyi, 84 items are referenced from that rdf through one another and, as you mentioned, the aggregate of those files is large totalling 521 MB.

izmeez’s picture

izmeez commented

subscribing

dman’s picture

dman commented

OMG, you mean my instructions were enough to actually replicate a MeSH import? Cool!

Well I imagine that PHP will die badly if you point it at a file like that. That's why I had to do the piecemeal snippet thing with MeSH. The same thing could be done to Galen ... I imagine.

Even better would be to find one of the rpc galen services that the Galen website talks about. See, that's the next step after getting the dump - being able to respond to smaller queries. That's the next step that my MeSH server takes it to So if we could find an endpoint like that then we could talk to it sensibly.

groovehunter’s picture

groovehunter commented

subscribing

allio_froggio’s picture

allio_froggio commented

I was wondering if there'd been any progress with this, say with Drupal 7? To get really fancy, I'm investigating a project wherein we'd link to video (probably using the Media Module) and would use the taxonomy to describe the video content (we wouldn't actually be hosting the video on the Drupal site). If anyone has worked on something like this, I would love to share notes!