by Michelle Lentz
For the last year, in various articles, PowerSet is mentioned here and there as a Google killer. I don’t think that’s accurate. I think PowerSet represents the evolution of search. In a course I teach on Web 2.0, I struggle to explain the concept of Web 3.0 and the Semantic Web. PowerSet has now made that easier for me to demonstrate.
PowerSet is a “natural language” search engine. I don’t really care for that term – I prefer contextual search engine. It indexes not only the key words, as in a typical search engine, but the meaning behind the words – word usage – and is capable of searching for answers within a set language context.
But let’s back up and just play with the tool for a minute. For it’s limited launch, PowerSet has indexed only Wikipedia and Freebase.
I did a search on Jim Henson, creative genius and creator of the Muppets. The first thing I’m shown is a screen with two tabs, which separate out different Henson-related topics: Jim Henson (the person), and The Muppets. Each tab has an excerpt from the main Wikipedia article as well as a brief summary of key facts to the right (albums, birth, death, etc).
Scrolling down there is a list of key words, organized by topic, called Factz. Clicking on a Fact displays a mini-summary of that Wikipedia article. At the bottom of the page is a list of other Wikipedia articles that are relevant to the topic. In this case, there are Wikipedia articles for The Jim Henson Company, The Jim Henson Hour, Fraggle Rock, and more. Clicking on any of these displays the mini-viewer, with a brief overview of the page.
Once I actually enter the Wikipedia article, it feels suddenly like Enhanced Wikipedia. My favorite part, and perhaps the most useful, is the floating Article Outline window to the right. This article outline floats along with you, so that it’s always viewable no matter how far down the document you scroll. The Factz portion shows an outline of the major points of the article, making it easy to skip around and find what you need. You can also search within the Factz outline, further refining the ease of use. Alternatively, in the floating window, you can view the standard outline, which just pulls the major topic headers out of the Wikipedia article.
At the top of the Wikipedia article window, you can click the Explore Factz button. This drops a key word cloud of relevant “Things in this page” at the top of the article. It includes everything from Jim Henson Foundation to Fozzie Bear. A separate cloud gives you a list of verbs (Actions) used within the article.
PowerSet also understands questions. Pointedly asking, “How did Jim Henson die?” gives me the unequivocal answer: Pnuemonia.
So, PowerSet takes Wikipedia to the next level, giving context to your search. Right now, this is a great tool for folks who need to research anything and don’t mind using Wikipedia. (That said, always verify your sources using Wikipedia’s external links.) Imagine this technology applied to Google – to indexing the entire Web. Or take this smaller and imagine it indexing just a single Web site. I would love to have this technology applied to my wine blog – or better yet, some of the larger more comprehensive wine blogs.
Of course, there are down sides. It takes a long time, and a lot of microprocessing power, to index the way PowerSet does, “reading” each sentence and understanding the context. According to an article in the San Jose Mercury News, it takes “a huge amount of computing power – it can take a single microprocessor as much as 20 seconds to analyze a page.” But PowerSet is pretty transparent about its growth models, and has released them to the world.
PowerSet has had the foresight to file all sorts of patents on their technology. If I were Microsoft, or Google, or even poor Yahoo, I’d seriously consider either investing in this little company or trying to buy them. (Rumors already have Microsoft considering a purchase.) I believe, whether through PowerSet or another company, that this is the future of Search.
If you know of any great tech news, apps, or gadgets, shoot me an email at michelle[at]writetech[dot]net or via Twitter.