Tagging Pubmed With Umls

Question

[Christiank]

Tagging with MetaMap seems to be very computational intensive and hard to do for the full corpus of Pubmed. Is there a way to do tagging faster with less accuracy or is there maybe a source where I can download Pubmed tagged with UMLS terms?

[Michael Kuhn]

The UMLS Metathesaurus is actually a collection of dictionaries, and MeSH is just a subset of UMLS. So you have to choose first which dictionary of the UMLS Metathesaurus you want to use.

[Heather Piwowar]

Are you looking for tagged PubMed/Medline abstracts, or article full text? I can't tell from your question and comments.

Answer 1

[Pierre Lindenbaum]

I think that the MESH terms of pubmed are tightly related to UMLS.

So you can use the NCBI eutils to download your papers.

[Michael Kuhn]

yes, but MeSH is just a subset of UMLS. You also have other dictionaries in UMLS, like COSTART, MedDRA, SNOMED, ...

[Pierre Lindenbaum]

you'll get the MESH term in the XML entry of a pubmed article: e.g. in http://www.ncbi.nlm.nih.gov/entrez/eutils/efetch.fcgi?db=pubmed&id=11934909&retmode=xml a mesh term is "Protective Clothing"

[Pierre Lindenbaum]

how are you accessing the article ? Are you using the NCBI-Eutilities ?

[Paulo Nuin]

Yes, that would be my answer too.

[Christiank]

If I understand it right this allows me to ask NCBI to get pubmed article IDs for a given MESH concept. I don't see where that webservice provides me with information about all the concept contained in a single pubmed article and where they appear.

[Christiank]

But then I still don't have a tagged version of the article but the article in plain text.

Answer 2

[Heather Piwowar]

The Medline demo at Lingpipe might help: http://alias-i.com/lingpipe/demos/tutorial/medline/read-me.html and a specialized version, LingMed, at http://alias-i.com/lingpipe/web/sandbox.html

I can't see a Lingpipe example of tagging with the UMLS per se, but you could probably make it work.

Or, you could do a quick and dirty tagging (via NLTK/Lingpipe?/other) with terms and synonyms related to the task you're working on, perhaps derived from the Metathesaurus? http://www.nlm.nih.gov/research/umls/new_users/online_learning/Meta_004.htm

Answer 3

[Joselu]

Try bioLabeler. BioLabeler extracts UMLS concepts from Biomedical texts such as scientific paper abstracts, experiments descriptions or medical notes and can be use to automatically curate and annotate BioMedical Literature or to index large documents databases and improve searches or discover relationships between them.

http://www.biolabeler.com