Posted May 12, 2008 8:18 am by with 12 comments

Tweet about this on TwitterShare on LinkedInShare on Google+Share on FacebookBuffer this page

There’s a buzz in the blogosphere over the launch of Powerset, a natural language processing search engine that uses Wikipedia for its index. Either I’ve not had enough coffee this morning or I’m the only one that’s asking these pointed questions.

  1. Didn’t try natural language search? Didn’t it fail?
  2. Didn’t Google spend the last 10 years conditioning search engine users to use a handful of keywords–not natural language?
  3. Isn’t Wikipedia made up of just 2.3 million pages, while Google’s index is likely 40+ billion? Even I could build a search engine that scales to 2.5 million edited and organized web pages.
  4. If Powerset is licensing its natural language technology from Xerox PARC and its index from Wikipedia, where’s the value? What’s to stop Google or Microsoft from licensing the same technology?
  5. Powerset’s co-founder predicts "2008 is the year that semantic and linguistic technologies cross over into widespread consumer use." So, in the next 7 months, you’ll do what no one has done in the past ten years?

Don’t get me wrong, you have to give Powerset kudos for finding a niche in the search engine industry, but I’ve seen this type of technology–yes, as good as Powerset–many times in the past five years and none of it has ever scaled or entered "widespread consumer use."

Have you tested out Powerset? Did it handle all of your natural language queries?

  • I spent about 30 minutes with Powerset last night and was not all that impressed. It seems to me to just be a fancy way to search Wikipedia and not much else.

    Kevin Bondelli’s last blog post..Crystal Strait Pledges for Barack Obama

  • Pingback: Crenk » Blog Archive » Powerset Launches a Wikipedia Search Tool()

  • Geo

    I can’t agree more than you Andy. Typing natural language queries is long and boring. Not only that – its COSTLY.

    Yes, costly. As we know, humans are the lasiest beings on planet earth and our brain is also seeking the lasiest least energy consuming approach to a certain goal. And when the goal is finding information about something it’s not at all formulated like a sentence in the brain. It’s in the form of keywords that need be assembled into a meaningful sentence (that’s a costy process, look at children who are just learning to speak). So why bother with that and the longer writing, when you can just type in “Golden Bridge name” and get your result…

    As I was looking at the presentation video this morning I started wondering – who is going to have time to go through all those nav menus and to click on all those nifty buttons when he can get the same with just typing it into Google…

  • Pingback: » Will Semantic Search Work For Local Search? Local Advertising Journal()

  • Cant I create the same thing (probably better) in like 15 minutes using Google custom search, only searching Wikipedia?

    Jaan Kanellis’s last blog post..SEO Analogies Help Us Understand

  • Q: “What’s to stop Google or Microsoft from licensing the same technology?”

    A: Powerset has an “exclusive license” to PARC’s NLP technology:

  • 1) Ask was not processing every sentence from every document. The amount of processing power required for Powerset to analyze each sentence just wasn’t available back in the days when Ask first launched.
    2) Powerset adds value across the query spectrum. For a topic keyword search like “henry viii”, we show Factz gathered about Henry VIII from across Wikipedia.
    3) Wikipedia is very high value content and Powerset includes lots of tools for discovery and summarization. Note the features we have on our enhanced Wikipedia articles, like “Show Factz”
    4) Powerset has built on the IP licensed from PARC and now has a healthy portfolio of our own IP
    5) It’s not just about the query: it’s about features enabled by our understanding of the text.

    The best way to experience Powerset is to try out the product yourself. Have fun and I look forward to your reaction.

    -mark, powerset product manager

    Mark Johnson’s last blog post..Powerset crosses the Uncanny Valley

  • @Mark Johnson – thanks for your input. If you guys prove me wrong, feel free to come back and rub it in my face. 😉

    The best of luck!

  • PS3

    The way teenagers are abusing the English language these days, it would be a better idea to develop an “text talk language” search Engine!

  • With the MicroHoo stuff going on right now I think another search engine can come in right now – the door is open. But saddly I think PowerSet is not the one to make it happen. The timing is there, just not the search engine.

  • This article shows a lack of understanding of the subject matter.

    “Isn’t Wikipedia made up of just 2.3 million pages, while Google’s index is likely 40+ billion?” Wikipedia’s content is significant because of it’s semantic value. Beyond what they have shown with Factz, Wikipedia holds the references to metadata — statements about web resources are linked to these references.

    “Didn’t Google spend the last 10 years conditioning search engine users to use a handful of keywords–not natural language”. That has little to do with what semantic search is. Semantic has more to do with grepping the words in the page and interpreting that “Ajax” in this page relates to the web tech, not the dish soap. It’s transparent to the user, but the search results presented are more accurate based upon their meaning, not just the textual match.

    Found this article to be far off on insightful content (like the other more recent article which these bullet points you present were copied and pasted).

    Those who actually understand the technology, understand that what Powerset has done is unique, non-trivial, and has a great deal of use and potential.

    Mark Holton’s last blog post..Microsoft buying Powerset?

  • @Mark – you’re welcome to your opinion, but I found your rebuttal lacking anything of substance.

    Andy Beal’s last blog post..Google Testing New iGoogle, No OpenSocial Support (Yet)