Dear All,
the message below from MedLib-L may be of interest to the list: a very lucid
/ knowledgeable explanation what happens when searching 'innocent' phrases
on PubMed and why e.g. 'heart failure' produces only one hit.
Regards,
Reinhard
R. Wentz, ICSTM, London
------------------------------
Date: Wed, 9 Jan 2002 11:10:23 -0600
From: Donna Berryman <[log in to unmask]>
Subject: Re: Translated search : PubMed (long)
There have been several postings the last few days about how PubMed
translates what's typed into the query box into an actual search. I think
that knowing how PubMed does this makes searching a lot easier....at least
it has helped me a lot.
PubMed uses a process called Automatic Term Mapping. Once a word(s) is typed
into the query box and the search is initiated, PubMed checks the word(s)
agains the MeSH Table -- if it finds a match, it slaps the MeSH term into
the search, adds the appropriate text words (in order to catch those
"in-process" or "supplied by publisher" citations that are not yet indexed
with MeSH terms) and runs the search. It figures it's work here is done.
If it does not find a match, the next place PubMed looks is the Journal
Table. If it finds a match (as in yesterday's example of the search for
heart failure), it runs the search, thinking the job is done. If there is no
match in the Journal Table, it will check the Phrase Index. If no match
there, it will check the Author Table. If no match there, it will do an "all
fields" search.
PubMed repeats this process every time we do a search UNLESS we do something
that will turn off the Automatic Term Mapping, such as: using quotation
marks around terms (""); using truncation (*); or, using command searching
([tags]). Hence, the explanation for the differences in translation that
were noted in the posting below.
Just to keep life interesting, the Gateway searches differently. In the
Gateway, quotation marks are required in order to search for a multi-word
term. The Gateway treats the words in quotes as a unit. It treats the words
without quotes as individual, unrelated entities which it will then relate
using the term "AND." Hence, the differences in translation that were noted
in the post below between "pain therapy" and pain therapy.
Our jobs are certainly not easy. But, perhaps they'll seem a bit more
logical when we understand how a given database's search function works.
Donna Berryman, MLIS
Outreach/Education Coordinator
New England Region, National Network of Libraries of Medicine
University of Massachusetts Medical School
Worcester, Massachusetts
email: [log in to unmask]
or, email: [log in to unmask]
|