Integrated vs. federated search

One of the key decisions to be made when implementing an Enterprise Search solution, including Recommind, is whether each information source is federated or integrated (or ‘indexed’ as Recommind sometimes describe it).

A recent buckeye post does a good job of explaining the difference.  In essence:

  • federated = multiple searches simultaneously; send the search criteria to multiple sources and get their own, independent search engines to return the results they think are the best matches
  • integrated = single search; index all the sources in the one place and get our one, chosen search engine to check them all the and pick the best matches

Note that the choice of federation vs. integration is more at the back-end (how we find matches) than the front-end (how we display results).  Normally federated results are returned separately:  eg a separate tab or listing for each results set.  However there is no technical reason why federated results are not returned in one set of matches.   Or why an integrated search does not return different sources in different results tabs.

You’d imagine that an integrated search should return a better set of matches since it will consistently apply the same criteria and  search logic to all the sources and therefore more consistently pick the winners.

However it is not necessarily the case that you would always choose to integrate rather than federate:

  • not allowed – some sources will not ‘open the bonnet’ enough for you to directly index their content.  This is particularly the case with those that provide (= in competition) search solution themselves
  • not beneficial – for some sources the provided search tool may be preferred to the single Enterprise approach as it will obviously allow both search criteria and results to better target the content.  In such a case federation may be preferable to simply provide pointers for users to then conduct more detailed searches within the source directly.
  • not appropriate – related to the question of benefit, there may be some content sources that users would not expect to be searched in an integrated manner.  And would fully expect/prefer to search independently and separately.  For example, nomatter how well your chosen search tool works, it is unlikely that anyone completing a knowledge search in your firm would expect the entire internet to be included in the scope.
  • not technically possible – there are some sources (the “internet” for example) where it would be impossible to index all the content.

So in putting together an Enterprise Search plan, you need to use the above definitions and criteria to help determine which sources we federate, which we integrate, and which we avoid completely.

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>