Thursday, 2 September 2010

Enterprise Search – panacea or pain?

Search can be seen by many as the answer to all of our information retrieval problems, but I have always thought that it depends how it is applied – one of my favourite phrases is: “rubbish in, rubbish out”. Search is only any good if it is pointing at good content, and it intelligently utilises what is known about both the documents and the searcher.
An excellent post by Venkatesh Rao this week focused my mind on this issue once again.

There has been much discussion on search over the years, some good points can even be found in the posts advertising search software. Such as this; “The Business Value of Enterprise Search 2009 A review of cost effective solutions for managers” advocating Oracle, which has the line: “A common flaw in the use of information search technologies is overconfidence in the results”.

I have always been enticed by semantic searching, and again there is a lot on this topic out there. Pandia Search Central is a an interesting one to follow, and they review their chosen top 5 semantic search.

An interesting concept that turns traditional search on its head is that of document comparison - computational forensic linguistics. If document comparison was used to search a quality data set, the results are far more accurate and relevant than traditional search technologies.

CFL Software Ltd is one company exploiting this in many different contexts. One great application has been in a music site, Slicethepie. The latest innovation is “SoundOut Search” which enables users to search a music catalogue using what ever snippet of information they have.

No comments: