Powerset was originally promoted by investors as a potential “Google killer.” As you’ve probably determined by now it’s not. But it is extremely interesting and useful.
I’m not a linguist or an engineer but I spent a couple of hours with the company in their offices and later with the engine and was impressed by some of the things that Powerset could do; it’s clear there’s a “there there.”
Specifically, Powerset was able to retrieve pages with relevant content that didn’t prominently feature keywords used in the queries. In other words, the engine understood concepts rather than just mined the pages for relevant keywords. It also organizes relevant (and related) information in a way that makes it quicker and easier to find. In addition, there are a range of interesting things Powerset does in presenting information and making it more “navigable.”
What the company has achieved, though long delayed, is significant. Microsoft is rumored to be interested in acquiring Powerset, which makes sense for both sides in many respects. Microsoft could integrate and scale the technology and Powerset could seek to apply its algorithm and approach to the entire Internet without having to go out and raise millions more. If the company “goes it alone,” it faces a long slow climb up the mountain (which is certainly not impossible, but difficult).
What’s also interesting to me is that this is familiar search, but with enough visually and technologically to differentiate it — such that it could, if broadly applied to the open Internet, capture usage over time. However, initially, Powerset is launching as a better search engine for Wiikipedia with some additional data from Freebase.
Here’s an example search for “New York.” The top of the page is nicely organized with lots of structured content:
You’re able to click any result and preview content within the results page:
Powerset organizes Wikipedia pages and creates an outline (right column) that is “clickable” to easily navigate through content:
The “outline” allows you to jump to any heading on the page:
The structured content presentation is thematically consistent with the whole direction of search. But the content preview feature and article outlines, as well as the summaries and related links at the top of the page, are nicely done and different. There’s enormous potential here.
When I saw it and discovered it was limited to Wikipedia, it struck me like a proof of concept for some version of “semantic” Web search on a massive scale. (Wikipedia is a micro version of the Internet in some ways.) The massive question now facing management and investors, however, is where to go from here?
For much more discussion, see Techmeme.
As my New York query above suggests, there are some fairly interesting travel and local search applications and implications to all this. We’ll have to see whether and how they play out.
Update: I had an email exchange with an executive at a general search engine who was very critical of Powerset and its capabilities. He pointed to the fact that the company licensed much of its core search technology as a red flag.