Tuesday, January 16, 2018

When Ovid won't map to a known MeSH term

OK, here is a weird thing. If you search "documentaries and factual films" in Ovid (note the quotes to deal with the "and"), you will receive the notice: Ovid was unable to map your term 'documentaries and factual films' to a subject heading.

However, this term is definitely a MeSH term. You can find it in the MeSH database (it is a Publication Type), and it has been called this since 2008 (it was called something else 1997-2008).

Weird! Since it is a publication type, we tried searching "documentaries and factual films".pt. in Ovid, and got zero results. Which was a clue, but we didn't know it at the time.

We then thought, OK, we'll search it in PubMed, and find an article, and then look at the Complete Reference for that article.

However, when we searched it in PubMed, it turned out that there were zero results. Not a single document has been labelled with this publication type since 1997!

(Which kind of made us wonder why it is a publication type at all. I mean, why is this a publication type and systematic review is not a publication type?!? But I digress...)

And this explains Ovid's behaviour too. Although the exact algorithm that Ovid uses is proprietary, it's generally known that Ovid takes the term you enter, looks for it in an unknown set of fields in the articles in its database, and then returns the most prevalent MeSH terms in the articles containing that term. Since this is a MeSH term that has never been used, and the phrase is stilted enough that apparently it doesn't appear in any titles and abstracts either, no results were found.

Now, I presumed from some of Ovid's other behaviour, that it searched the MeSH database directly as well as looking for the phrase in the titles, abstracts, and indexing of individual articles. I thought that's why a search for public health returned public health but not public health surveillance and other terms containing the words "public health". I presumed that the algorithm found an exact match in the MeSH database and therefore stopped looking. However, if that were true, then this search would have returned exp "Documentaries and Factual Films"/ with zero results.

No, I think this means that a direct search in the MeSH database is not part of Ovid's mapping algorithm. Instead, I guess it does search of articles every single time, even if the entry is exactly the same as a MeSH term. Then, it must generate a list of potential MeSH, and then (get this!) it must delete all of the other found MeSH if there is an exact match. (Whaaat?!?)

Well, I'm sure that it makes sense to someone, somewhere.

Anyway, if you search a known MeSH in the Ovid interface and get an error message, try confirming in PubMed to see if there are zero results. BTW, typing exp "Documentaries and Factual Films"/ did not produce an error message. There were, of course, zero results.

Thanks to my colleague Robert for bringing this to my attention.