Tuesday, August 04, 2015

MeSH on Demand

Does MeSH on Demand have possibilities for identifying thesaurus terms for use in a search, as I wondered if in a previous post?

MeSH on Demand is here: http://www.nlm.nih.gov/mesh/MeSHonDemand.html.  There is information about it in this NLM Technical Bulletin article

As a first test, I entered this text:

The use of antibiotics to treat otitis media in children under 5

MeSH on Demand suggests:

Anti-Bacterial Agents
Antibiotics, Antitubercular 
Child
Dermatologic Agents
Humans
Otitis Media

There is a disclaimer that says that humans might have come up with different terms, and that MeSH on Demand might suggest terms that are not in the text, or miss ones that are.  It does not assume any prior knowledge of MeSH.

I would have come up with the first and last terms, and humans, although I think that would be included implicitly if you include an age group.  I would have chosen something more specific than Child.

If you explode the first term, Anti-Bacterial Agents, you would include Antibiotics, Antitubercular.  There is no advice that I can see from MeSH on Demand about exploding.  If you use PubMed, exploding would happen by default.  

I would not have come up (at least, until I had examined some search results) Dermatologic Agents, so that is interesting and I would explore that.

There is no guidance from MeSH on Demand about how to combine terms.   There is also nothing said about freetext terms that you might use.

As a second test, I input the abstract from an reference from PubMed, and the resulting list of suggested MeSH terms had a very small overlap with the actual terms applied to that reference.  Indexers do look at more than the abstract, so there may be ideas in the article headings, conclusions and tables and figures that are not in the abstract.   Interestingly, Philippines/ is suggested (it is mentioned in the abstract) but is not actually applied to the reference. 

And as a third test, I used a search involving a drug and a condition that I had used as an example before.  MeSH on Demand identified the disease MeSH term I had chosen, and a drug MeSH term I knew about but had rejected (as too broad), plus a MeSH Supplementary Concept (the idea of those was new to me, I have to say, and I am not yet sure how you search it in Ovid!) and a MeSH term for the part of the search I had decided to deal with by means of a publication type limit or filter.   So in this case, MeSH on Demand was useful as a source of ideas.


No comments: