Powerset Will Launch In Coming Weeks

Trackbacks/Pings (Trackback URL)

1. Zaid
   

   April 5th, 2008 at 2:28 am

   Is this a delayed April Fools?

   Hopefully not. It’s about time we saw Powerset’s power:)

2. Armand
   

   April 5th, 2008 at 2:33 am

   Hi Mike,

   Thanks for sharing this. Also had a look at Poweset’s site and blog and signed up there as well.

   I like how their searches will be more natural language driven than just keyword based driven. It would be nice to query a more natural question rather than specific keyword(s).

   A very promising company and can’t wait to see when they bring it out to the public….

3. alex
   

   April 5th, 2008 at 3:30 am

   Wasn’t there some search related “startup” that had raised somewhere near the $50m mark?

   Was sure it was Powerset, though Crunchbase says they’ve only raised 12.5 to date. Am I just imagining things?

4. CARversation.com
   

   April 5th, 2008 at 3:42 am

   awesome

5. Marc Fawzi
   

   April 5th, 2008 at 3:53 am

   Mike,

   The “Aha!” feeling you had you would also have from google or any search engine that is accessing structured and semi-structured information as in freebase and wikipedia.

   The challenge in search is in accessing unstructured information on the wild wild web not freebase or wikipedia.
   Having said that I’m sure their NLP is more advanced than what we’ve seen and I do appreciate their approach.

   But until they solve the problem of searching unstructurd information they’re not going to change the world.

6. Andy Gongea
   

   April 5th, 2008 at 3:59 am

   Hey, great news

   I joined the beta testing several months ago. First of all I like the interface. Nice UX implemented and quite good results coming from the queries.

   Hopefully it will mean something in the search engine area.

7. diystartupnews.com
   

   April 5th, 2008 at 4:07 am

   I am looking forward to this but if all it does is index those 2 sources then its a pointless release in my books. Anyone could develop a search engine to give good results from those 2 sources with natural processing or not. The real test of a search engine is lettting it out in the wild.

8. jenkins
   

   April 5th, 2008 at 4:41 am

   Michael,

   Having some background in search I can tell you that:

   (1) They have nothing
   (2) You saw nothing

   Why?

   They’ve built a search engine around two content sites. It’s taken them about 16 months or so. They’ve essentially been “tuning” the site using all kinds of tricks, using all sorts of tools, manual overrides included, for months now. Their team isn’t huge. Their index is small. My guess is that they will cross their fingers when someone searches hoping they saw that query before and have already manually “fixed” it.

   Add these guys to the deadpool, Mike. You’re a bit naive when it comes to hardcore product development. I’d like for this to be real but I seriously doubt that it is.

9. Yakov
   

   April 5th, 2008 at 5:27 am

   Why don’t they want to understand that they can’t beat a regular search engine

10. logosamurai
    

    April 5th, 2008 at 6:02 am

    very interesting.

11. NickeyD
    

    April 5th, 2008 at 6:29 am

    I tried their wikipedia beta last year and it was a total crap. Hope they improved greatly since otherwise it’s going to be a big disaster.

12. John Wesley
    

    April 5th, 2008 at 6:40 am

    Sounds like a total Mahalo killer.

13. damon
    

    April 5th, 2008 at 6:41 am

    “The early version of the service will serve as a showcase for the user interface and engine itself, but it will not have a full web index behind it.”

    I did not go and look this up, but isn’t this exactly what you said about a year ago when they wrote about it?

    Why no discussion about that fantastic set of modeling algorithm’s they “released” a while ago? Hasn’t that been a big hit!

14. Elliott Ng
    

    April 5th, 2008 at 6:42 am

    I agree with Barney that search is still in the “toddler” stage and will grow and learn to better understand the meaning of documents and the intent of search queries. Semantic analysis has promise. Powerset is pursuing a huge dream, of solving this in a generalized, scaleable way across all domains. Barney, I believe Powerset’s best bet is to look for massive user contribution in an open-source project kind of way, in order to solve specific domain spaces really well. At UpTake (formerly Kango) we’ed be interested in doing that, so long as the model allows us to economically benefit. Congrats again for steering the company to this point, and I can’t wait for the beards to come off!!

15. mike
    

    April 5th, 2008 at 7:00 am

    HYPE,HYPE,HYPE.

    When is Techcrunch going to give up on this company and stop promoting them or has techcrunch got some financial incentive at stake ?

16. John Wesley
    

    April 5th, 2008 at 7:12 am

    haha I though Mike might be an investor when I first read this, typical TC

17. Chris
    

    April 5th, 2008 at 7:33 am

    It’ll be interesting

18. Yes
    

    April 5th, 2008 at 7:39 am

    Enough about Powerset until they have a compelling product…

19. Craig J
    

    April 5th, 2008 at 8:08 am

    I don’t understand why investors spend a butt load of money on search companies especially when they know people are not looking for better search. There is no problem in the market (at least for another few years). The majority of population is still trying to grasp and get used to Google and regular search. Only a small group of techie geeks are complaining about search.

    I just don’t understand how companies like this (and Mahalo) justify their product and what obsesses investors to invest on “New” search engines.

20. kenny
    

    April 5th, 2008 at 8:15 am

    “In preparation for the launch, some of the Powerset team have vowed not to shave until the product is released.” I hope they don’t all look like Bin Laden by the time it’s released.

21. John Wesley
    

    April 5th, 2008 at 8:53 am

    Even if the features are better, search is a habit. Still, I often search with natural language questions and the results are lame in GOOG. I’d try it.

22. Tom
    

    April 5th, 2008 at 8:56 am

    Naysayers, let’s wait and see before judging this. We should be cheering them on as they are willing to take this big huge challenge on. One thing that I do have to say is frustrating are posts like this when we are told that a “release is coming in a few week”s coupled with a review by the author. It really makes it impossible to have meaningful conversation in the comments. Lets see the product and then debate.

23. NLP
    

    April 5th, 2008 at 9:02 am

    Am I imagining things or did you guys just delete the screenshot showing an obviously flawed set of results and the related comments posted by your loyal readers?

24. Nat
    

    April 5th, 2008 at 9:10 am

    With the economy going downward and Google stock is hurting and so is the adsense, they should release Powerset a long time ago. all i can say is goodluck to powerset.

    Nat
    http://www.workersinc.com

25. MikeT
    

    April 5th, 2008 at 9:12 am

    Offtopic: Michael, what’s with all these flashing banners on TC? They are really repulsive and are heavy on the eye! If you can control what ads you get on the site, maybe you could filter them out, because they don’t make TC look good…

26. Googlist4ever
    

    April 5th, 2008 at 9:28 am

    Google works fine for me, thanks but no thanks

27. John A
    

    April 5th, 2008 at 9:35 am

    Michael, natural language search engine Cognition.com has already launched a similar service. Check out their Wikipedia search at http://wikipedia.cognition.com/

    Here’s a recent article about them in SoCalTech — http://www.socaltech.com/inter.....08112.html

    Here’s another article in InformationWeek –
    http://www.informationweek.com.....tid=760137

    Thoughts?

28. un.valley
    

    April 5th, 2008 at 9:44 am

    lol.

29. searchengineer
    

    April 5th, 2008 at 10:09 am

    @7 completely agreed. Powerset took a bet that adding a large number of semantic features would be a way to get a step function in relevance. Playing with their demos, however, it seems pretty clear that they are struggling to find any real-life scenarios where their approach is better vs. a generic web search engine. here are their challenges:

    1. The vast number of web searches are 1-3 term navigational or research queries. Not possible to get right unless you have the document in the index, so they need to go to at least a 10-20B document index to be an interesting alternative to Google.

    2. Most of the examples they are showing produce fairly similar results if you run them against Google; any difference in the search results can usually be fixed by doing a single synonym substitution - E.g., “who bought IBM” vs. “who acquired IBM”. Query re-writing is now done by most major search engines to implement stemming and synonyms. For example, google “international business machines” and notice that Google partially rewrote the query to match “IBM” on the page. Unless Powerset’s technology has dramatically improved over what they showed on their demo site, at best their technology seems to end up producing a slight improvement over that approach (which does not need any of their deep semantic analysis) in a real world scenario vs. the dramatic improvement they need to be a viable alternative.

    3. Besides raw index size, they seem to be missing a ton of features that are essential for creating a modern search engine - spell correction, spam fighting, one box, etc, etc. Without those features, it seems difficult to believe that any average search user will switch to their engine, given the huge consumer satisfaction (80%+) for existing web search engines.

    Rumor has it that they are burning through $1MM / month even getting to this point. They have a talented team, so it seems difficult to image that they are not aware of those issues. My guess is that the internal mantra has been “even if we can’t get consumers to use this, we will get acquired for our engineering talent and IP”, but given the search consolidation, it feels like that window is closing, unless they get picked up by someone like Baidu….

30. ElectroGeek
    

    April 5th, 2008 at 10:22 am

    This is an interesting concept however do they really think it’s possible to go up against the Google, Microsoft and Yahoo giants? I will be watching these guys.

31. hannes
    

    April 5th, 2008 at 10:25 am

    Great News - Hope to see some stuff at the conference where Barney is hanging out in France - http://www.iwomc.com - it definitely is time to see something!

32. James
    

    April 5th, 2008 at 10:51 am

    Powerset needs to launch, now. No one should cover their near launch stories anymore.

    Hopefully they bring something totally new.

    Of the new Google ‘challengers’, Mahalo is a big fail due to the thin set of results going stale rapidly and Wiki Search results are complete garbage.

33. steve
    

    April 5th, 2008 at 11:24 am

    I’m curious about NLP’s comment (currently #21). I just saw the post so I have no idea if this is true? Anyone?

    NLP: Am I imagining things or did you guys just delete the screenshot showing an obviously flawed set of results and the related comments posted by your loyal readers?

34. Alex
    

    April 5th, 2008 at 11:30 am

    First off, I don’t believe for a second that Google is the end all, be all. Fact is there was life before Google and there will be life after Google. For those who think Google will always be “it” I have one name for you “Yahoo”

    Second, I’ve used Power Set and I can say it’s just another example of how people assume that by trying to attack Google will find a solid exit strategy. Fact is, Powerset is technology “junk” as far as I am concerned. Can I do better? Absolutely not but the reality is that this is a deadpool in the making. It’s quirky, irrelevant and frankly just doesn’t do what it promises and any response to that from them will simply yield “we are getting better.” Don’t get better, be better and don’t publicize “better” for over a year and then excuse junk.

    Finally, I hope that people get past the “we are better than Google” because Google was a nobody at one point and the one to beat Google will be a nobody at first, fact of technology life, always has been and always will be.

35. Dan Burcaw
    

    April 5th, 2008 at 11:34 am

    <a href=”http://www.doubleencore.com/blog/2008/04/02/powerset-broken-search/”As I mentioned last week, PowerSet has been using Amazon’s Mechanical Turk to assist in their internal testing process. Apparently, they feel comfortable enough with their results to launch. I really wonder why they think they have special sauce that Google doesn’t have. Especially since Search is their bread and butter.

36. Lora
    

    April 5th, 2008 at 12:09 pm

    Great news but..i don’t believe that it will last forever…
    http://funnyhack.blogspot.com

37. Mark Johnson
    

    April 5th, 2008 at 12:38 pm

    I wanted to toss a thought into the ring (although, agreeing with @22 that any discussion is a bit premature). A lot of the comments have speculated whether our results are better than traditional keyword search. Instead of focusing on results, consider that Powerset has a powerful natural language processing engine that extracts information from every sentence in Wikipedia. This allows Powerset to create a unique set of browse and explore features in our product, not just traditional blue-link search results.

    Mark Johnson
    A Very Scruffy Powerset Product Manager

38. Tim Basie
    

    April 5th, 2008 at 1:18 pm

    I don’t get the interest in this site. Natural language research is 50 years old, starting with Eliza. NOTHING with natural language has ever really worked out. Why will Powerset? Natural language understanding is too difficult and computers can’t really do it. The HAL plug-in for Lotus 1-2-3 in the 80s was cool but not really useful. Text adventures that required too much typing (other than “go north”) were not fun to play. Then there were numerous DOS and Unix plug-ins to translate English sentences to OS commands. They didn’t help. It appears as if computers are BEST USED with token or keyword input. Also, there have been reports about how lazy users are getting with search. They get mad if exactly what they’re looking for doesn’t appear in the top 3 links as a result of one or two vague keywords punched into the google search box. People aren’t going to start typing full sentences and conversing with the computer to find what they want; and double this for mobile devices. Powerset will be cool for a week or so, then we’ll all go back to google.

39. Michael Arrington
    

    April 5th, 2008 at 1:27 pm

    for those of you who claim bias, get a life. and then read our full coverage at http://www.techcrunch.com/tag/powerset.

40. Michael Arrington
    

    April 5th, 2008 at 1:29 pm

    agree with Mark, above. The engine allows you to find information that is just not possible in a short amount of time with a traditional search engine. you’ll get it when it launches.

41. Wayne
    

    April 5th, 2008 at 1:30 pm

    After 20 months(!) of “Powerset-is-going-to-be-awesome” articles, marginally-better Wikipedia search is a bit of a letdown.

42. Andy Gongea
    

    April 5th, 2008 at 1:50 pm

    There are many persons that have the one and only purpose: - to complain.
    Better try to build your own startup and you will see how hard it is. It doesn’t matter that the search engine was announced months ago, it matters that the developers have taken their time to build a good product.

    It is easy to criticize and hard to act.

43. Wayne
    

    April 5th, 2008 at 3:22 pm

    @42: Sorry, my expectations were improperly managed.

44. DaveS
    

    April 5th, 2008 at 5:28 pm

    Hey Mike,

    Weren’t Vaporset (excuse me, Powerset) suppose to launch on TC40? What was the deal with them when you accepted them to the conference? Weren’t the part of the deal that they launch there or was there some deception on either your or their part?

    Just curious,
    Dave

45. Scruffless Insider
    

    April 5th, 2008 at 8:44 pm

    @29

    “My guess is that the internal mantra has been ‘even if we can’t get consumers to use this, we will get acquired for our engineering talent and IP’”

    Actually, the internal mantra is “My God, we are really doing it. We really are building a NL search engine. Holy shit!”

    Why?
    Well, we really are doing amazingly well on wikipedia.And that’s not easy, especially when we are not tuning and crossing our fingers and anticipating queries as @8 suggests but we are doing the real work of getting NL search to work. So right now on Wikipedia we are doing better than anyone else can possibly do on many queries and as good as anyone can do on most others. And we’ve only just started. We have the knowhow and the team and the clear sense of how to improve. People think NL is terribly difficult? Boy are they right. NL Search or any other NL application for that matter IS rocket science. But remember, rocket scientists figure out how to build rockers. That’s what they do. We have a team of rocket scientists in NL and search and we know how to do this. For us, launch is just the beginning.

    Something to keep in mind.
    Our product improves everyday. We’ve got incredible headroom. The other guys? With their technology, they are happy with teensy improvements in core search over a year. With our technology, we’re heading for improvements in relevance in the doube digits in the same timespan.

    So. The mantra:
    “Holy shit. We are actually doing this.”
    It’s chanted everyday — even by people who do not sport facial hair.

46. David Kahn
    

    April 5th, 2008 at 10:47 pm

    It’s weird. It has the same problem as Hulu had when it came out. I just kind of entered my email address in the invite me section. and then clicked “forgot my password” and they basically sent me the password “changer.” The format is cool although you have to get through 25 karma points to actually view the thang. The search is awesome. I asked “Who is Blanche Bruce?” and it answered. I could have really used this for my research essay.

47. BuzzCritic
    

    April 6th, 2008 at 9:56 am

    I’m looking forward to seeing what NL search has in store. Not sure what to expect.

48. David Scott Lewis (Zytech Solar, a Going Green 100 Winner)
    

    April 6th, 2008 at 4:39 pm

    It’s perfect for the Live-Yahoo-Ask search crowd.

    If you’re a Google power searcher, it won’t be useful to you.

    Better for Main Street, U.S.A. — but that’s the real market.

    BTW, I’ve been in their beta for several months, so I’m speaking as someone who has used it rather than someone who has just read about it.

49. T. Benson
    

    April 7th, 2008 at 9:59 am

    Compare Cognition Technologies’ Semantic NLP.

    Cognition’s Semantic NLP is technology that focuses on the understanding of word and phrase meanings in modern computer applications. Cognition’s goal is to make its clients’ technologies more “human-like” in the understanding of language and more profitable.

    Cognition isn’t trying to be a full Web search engine, but you can compare Cognition’s Semantic NLP on the same dataset (Wikipedia) here: wikipedia.cognition.com. Do the same searches on both and see what you get.

    You can also try it on other vertical data sets:
    CASELAW.Cognition.com
    MEDLINE.Cognition.com

50. NLP
    

    April 9th, 2008 at 4:58 pm

    Ahem, guys. so, please help me decide if I should continue reading TC by answering my question (that now shows as comment #23).