Conversational Search

The magic in early instances of what we now call 'enterprise search' was being able to find content by typing in a few keywords. It wasn't as cool as the HAL 9000 computer featured in "2001 - A Space Odyssey", but it was good enough to draw a large number of people - myself included - into the business.

Along the way, Google perfected a search platform based on the theory that, at scale, just about any query you could think of had already been used by thousands, if not millions. of other humans. All Google needed to do is keep track of what pages other humans viewed following a query and promoting the page to the top. Essentially, they created a 'crowd-sourced search'. 

The bad news for those of us who work on search designed for use within the enterprise is that there just isn't sufficient content - or query activity - to deliver results as accurate as those we experience on the public web. Consider: Google marketed the Google Search Appliance for the enterprise. It didn't deliver the kinds of results public-facing Google does, and Google pulled the product from the market. For great search, size matters.

Nonetheless, some of the companies that market enterprise search products are now adding elements of machine learning with their products; and while perhaps not as accurate as web-based Google, they do deliver results that start out pretty well and get better with age, as the platforms learn what documents humans view following queries.

And if you've not noticed, some leading vendors are now integrating - and encouraging - what is known as 'conversational search'. Think about it: when you need to find a document in your organization, you may ask a colleague. But you don't simply say "sales'. Chances are you'll ask "where is the new sales report".

It's encouraging to see an increasing number of vendors delivering these capabilities in their commercial products.  The most recent to announce conversational search is Algolia, although I have to say I'm quite disappointed in the Wikipedia write-up on them. In my spare time, should I ever find any,  I should go do some edits, but this 'spare time' thing is rare for me.

Nonetheless, I'm happy to see an increasing number of commercial search vendors beginning to integrate these advanced capabilities into their products. Search in the enterprise has challenges: but hang in there: it's getting better! 

Note: How has your experience been with machine learning and AI integrated with your enterprise search? I'd love to hear your experiences - even if under NDA!