« Open.Michigan for Academic Healthcare at UM - How? | Main | New Search Engine - Cuil »

July 24, 2008

eTBLAST - Paragraph search yields journal/author recommendations

eTBLAST is an interesting search tool that allows you to search by either typing freetext keywords or pasting in an abstract of paragraph of a real publication.

eTBLAST: invention.swmed.edu/etblast/etblast.shtml

Right now, it searches MEDLINE (a.k.a PubMed) as well as NASA technical reports, IOP (Institute of Physics?) and CRISP (the government awarded grants database for NIH). Forthcoming (soon, I hope) are the US Patent & Trademark Office (USPTO), PubMed Central (PMC), Online Mendelian Inheritance in Man (OMIM), and DrugBank. My examples will all be in PubMed since health information is my primary context.

eTBLAST

eTBLAST is pretty fast if you search just a couple keywords, but you can do that in PubMed itself just as well. What is different is being able to put in an abstract or paragraph from an article, or to even upload a text file! If you do anything beyond a few keywords, it is not going to be fast - pretty slow actually. Go have a cup of coffee or do something else and come back (or assign it to an assistant). Hey, if you're in research, there are a lot of things for which it is "hurry up and wait", right? So this is a paragraph search.

eTBLAST: Original Search

Notice the link to show what textwords it selected for searching and what MeSH terms they were mapped to. This means you could also use this at the beginning stages of a MEDLINE search to discover possible MeSH terms to include in your own search strategy. So this could be a search tool or a search strategy development support tool.

This is what the initial results look like.

eTBLAST

Now things get interesting. When I'm working with systematic review teams, there is something I recommend to the teams doing hand searching. What I do is take all results, sort by journal, look at what journals account for 80% of the citations, and then we will do the hand searching in those titles. eTBLAST makes this process a whole lot less cumbersome, breaking out and listing the titles of the most frequent journals in the results set. Not that this was a systematic review search, mind you, but still a useful tool. It does the same thing with authors as with journals, so it is easy to break down your search results in a useful and granular fashion.

You can also map the search results in a variety of ways. I will confess, I don't understand all of them, and did not find all of them useful. I did find the publication history graph useful.

eTBLAST: Publication History Graph

Perhaps some of the other methods will make more sense to you? Let me know!

Posted by pfa at July 24, 2008 02:55 PM

Comments

Login to leave a comment. Create a new account.