"If people aren't bidding on keywords, and are bidding on concepts, it could completely change the ball game," said Scott Prevost, Powerset's General Manager and Product Director.However, not everyone is a fan of this new natural language search process. This includes Don Reisinger from Mashable. Here is why Don believes natural language search will not work."Keywords may not be fool-proof and may tend to make things more difficult for advertisers, but we can't forget that bidding on them works for one reason: it simplifies a process that is extremely hard to gauge. In Powerset's scheme, companies need to rely upon the intention of those same people and how well it can guess how they interact with a search engine, regardless of the keywords used.Trying to guess what people will say, as opposed to what people will include in their query, is extremely difficult. Why try to guess a whole sentence or a structured query when you can pick one word and hope people use it?"What are your thoughts on this, do you think there is a future in search for natural language processing?
Posted by Courtney Mills at 8:32 AM GMT
Natural language search isn't just about the query or how people express them in a search engine (though that's certainly important). Because Powerset reads and derives meaning out of every sentence, we enable all sorts of new features that weren't possible before. For example, check out our Factz by typing in a simple keyword query (try a famous person, like Henry VII) and look at all of the information we extracted about Henry VIII from across Wikipedia articles. This just isn't possible with keyword search technology. You'll see lots of features like this popping up because of the deep processing we do to documents.Cheers,Mark Johnson, Powerset Product Manager
Trying to work out the intention behind the phrases entered into a search box is incredibly difficult. There's nothing in the AI/NLP/semantic web toolbox that helps much because intentions are either 'keyword-obvious' [eg Boston late availability] or are not fully captured in the structure of a sentence. And even if people could be persuaded to search in well formed sentences it woud still not improve matters because the content being searched is not sufficiently structured to provide positive feedback (ie good results) to sentence-searchers.Congratulations to Barney Pell for spotting the potential for a semantic flip company. However there's no reason to think that the acquisition of Powerset will have any impact on Microsoft's market share of search related advertising
I certainly vote in favor of NLP or semantic based searches, not only with regard to streamlining the right ads, MS could use Powerset to enable a lot more search on the web pages and get customised outputs which could be specific for a field. Whereas google depends mainly on trail and error; many a times results in erroneous or non-essential hits, it may work in favor of Powerset to get the most accurate hits. Something similar works in the application/product that I have created check out :: www.xtractor.in