Posts Tagged ‘web search’

Library catalogue: Site Search analytics

Posted on March 17th, 2011 by Paul Stainthorp

A while ago (and, as with all things Horizon, with the help of Dave Pattern at Huddersfield), we enabled Google Analytics on our library OPAC (sometimes referred to as HiP, the “Horizon Information Portal“). This takes the form of a piece of Google JavaScript which lives in a ‘footer’ document common to all HiP pages.

Chris Leach gave a presentation about using Google Analytics with HiP at the last SirsiDynix Horizon User Group.

Now Nick Jackson has shown me how to enable Google’s Site Search features on our Analytics profile for the library catalogue. Site Search will allow us to ‘tease out’ the search activity within the library catalogue itself, by analysing the URL structure of HiP queries, recognising and extracting the search terms, then tracking the paths users take from those search queries to destination pages (i.e., individual bibliographic record pages on HiP).

For instance: a typical HiP search query ends up looking something like this:

By telling Google Site Search to look in the query parameter “term” for the search keyword(s)—in this case journalism—and to ignore the “session” parameter, Google Analytics can start to group similar queries together and provide us with data about what our users are searching the catalogue for.

Screenshot of the setup page for Google Site Search

It’s been running for less than 24 hours, but already we’re starting to see build up a record of the keywords people are typing into the catalogue:

Screenshot of Google Site Search top search terms on HiP

What could we [and what should we] do with this data? Are there any Google Site Search experts out there who could give me a few tips? If anyone from within the Library at the University of Lincoln would be interested in helping to analyse the search term data, please let me or Chris know.

One thing we’ve already discussed is the idea of using the HiP search term activity as test data to ‘teach’ the Jerome machine intelligence engine about the kind of things Lincoln library users are interested in… this will help us in determining how the Jerome API’s personalisation features might be used to present and relevance-rank results.