On 27 Oct 2006, at 13:06, Stevan Harnad wrote:
> Andy's informal test is a good demonstration that google is
> *already* just about good enough for "known-item searching"
> (i.e., where I know the reference I want, and am just
> looking for an OA version of it on the web).
>
> But the real question is: What proportion of a researcher's
> searching and search-needs consist of known-item searching?
> What about open-ended searches on topics, keywords,
> boolean text items, etc.?
Although I can't determine the true proportion "KI search : OE
search", anyone with a repository can find an approximation to it.
Just filter out all the items in the logs which come from Google (the
string '.google.' appears in the referer field). I have created a file
of 30000 google searches that led to either the ECS or the
Southampton repositories over a 6-week period last year.
The vast majority (>99%) of these appear to be open ended searches
(although it may be difficult to differentiate between a focused OE
search and a known item search).
It should be possible to revisit Andy's experiment and to use these
terms as real OE search queries that historically resolved to an
eprint via Google.
We could then look at the ranking of the eprint that was delivered
via Google and compare it to the ranking within the OpenDOAR CSE.
That should tell us something more about the behaviour of Google with
respect to the literature in repositories.
--
Les
Received on Fri Oct 27 2006 - 21:24:50 BST