The Mosaic search engine

The Mosaic search engine is a demonstrator that was developed by Hedtek as part of the JISC MOSAIC Project. The search engine is a proof of concept investigation of the potential of user attention data (in the form of library circulation data), the ability to find library items that pertain to ones own context, or, browsing around , the ability to adopt other personas and see what pertains to them.

The purpose of the demonstrator was to assess

• If we could construct a search engine that supplied personalisation facilities around a particular kind of attention data, library loan records
• If we could thereby leverage the users context, in this case where they were in the university system (university, course, progression level)
• The feasibility of basing this on the Lucene/Solr, and the way in which the index data would need to be constructed and used.
• Possible query and display interfaces
• Data anonymisation processes
• The sense and value of the intelligence to real users (to be exploited more fully in workshops in Oct / Nov 2009)

Ultimately we would expect the model to support

• Reading lists
• Use records derived from library circulation/loan systems (LMSs)
• Use records derived from virtual learning environments (VLEs)
• OpenURL resolver / ERM derived information
• Search histories
• Searcher location
• Searcher preferences for particular kinds of results

Two kinds of results are supplied by the demonstrator:

• Normal un-personalised search, with no facets selected (results with light blue backgrounds)
• Faceted search (1-4 facets selected) that define points of view, eg personalised for me, for third years at Oxford, for me if I was in the year above, for PhD’s in my subject, this year, last year, etc. (here the relevant results for the persona selected appear with a dark blue background, other results with a light blue background).

Two interfaces are supplied to do the persona adoption

• A Classic faceted shopping interface
• An iTunes browser-like interface

Because we were aware that the faceting might sometimes result in a narrowing of choice, we supplied two mechanisms to broaden choice

• A results like this link for each result, that invokes a search based on the results title (and is surprisingly effective)
• The incorporation of reading list information, supplying the complete reading list if a search result is on that reading list.

The demonstrator contains only LMS Circulation and Reading List data. It combines the COPAC monographs database with three years of usage data from the University of Huddersfield. Not all titles are therefore linked to use data.

The data has been ‘doctored’ in 3 ways

• Split in to 5 fictitious universities
• Example reading lists inserted to illustrate how they might be integrated
• Links made from the Title display to the JISC EIE project ‘compositor’ to display further details and provide opportunity for contribution

The interface implemented here searches only authors and/or titles and then allows the results to be refined using any combination of four facets

• Institution
• Progression level (plus a staff ‘level’)
• Course
• Academic year in which the loan was made

Results are returned for almost any keyword combination search – but not always linked to use data (that’s the real world). To see use data, which is displayed in the four ‘facet’ lists, try searching for

• Accountancy
• Economics
• Medical

To pick up reading list records as well as use data, try searching for

• medical law

and click on the reading list icons.

Try the search engine

Try it!