Powerset Launches. Verdict: Meh.
20bits.com
20bits.com
- The UI design is visually distracting, bulky, complicated. There's too much going on at the top, 3 differently colored bars with the powerset logo, tab(s) and huge search bar. The search bar is omnipresent, even in the individual Wikipedia page you chose. Ironically, Google has a similar structure in its result pages, but feels completely different because the background is all white and less invasive.
- The Factz are often distracting and presenting useless stuff. E.g.: http://www.powerset.com/explore/semhtml/Simple_DirectMedia_L... presents around 25 factz. I counted 3 or 4 factz that are actually useful. The meaning of the other factz in the context is either obscure ("parameters": which ones?), deceiving or incomplete ("functionality", "ports", "operations", etc), useless ("needs" (??), "SDL", "The Video"), or all of the above.
- The way you actually use the Factz is terrible. First you clik on one fact, then after being flashed with the fade-out bling on the right column you actually have to move the mouse pointer all the way there, then the animation scroll, then probably move the pointer to the left to click on something actually useful or go back up to select another fact because the one you chose was wrong anyway (why? see above). After 5 minutes of this I feel dizzy. Who designed this stuff?
- massive waste of screen real estate with large white space on the right column, either entirely (results page) or partially (actual wikipedia result page you selected).
I do think there are some cool things, but so far it seems hard to use IMO.
Try searching for "Who was the tenth President of the US?" on both Powerset and Google to see the power of "natural language search" in action.
(posting this b/c your comment doesn't make it clear that google wins that search)
If you want a more fair comparison (you may not), site restrict your search to wikipedia and then compare. Both powerset and google don't do well here, the closest they get (and they both get nearly the same answer for the first few hits) is a mention of the tenth Vice President. I'll tell you why:
Tables. The data for this is in "List of Presidents of the United States" in a purely tabular format. Tables suck for everyone right now! They're not really NL, they're not keywordable, they're basically a nightmare because they require a pretty surreal amount of context to grok.
This isn't really an excuse, I admit. Powerset is working on it, but we just couldn't have it ready and still put out something when we wanted. At some point you have to put out something to show the world, and most powersetters feel the current tool is interesting and helpful. It's not perfect, or even necessarily complete, but I don't think anyone with an informed opinion expected it to be a google killer.
But seriously, if you're going to do a comparison at least go apples-to-apples and site-restrict to wikipedia. And yes, today at the lunch table the talk was all about what new data sources to grow into next, and it's publicly available information that we want to scale to the entire web.
Half the point of a pre-product like this (and lots of sites start out with reduced-power versions of their final product, even Google) is to just get the infrastructure made. Anyone who's got half a brain can tell you that getting infrastructure to handle non-trivial search is a huge task. You have query analysis, you have indexing, you have parallelized index servers, you have failover, these days you have EC2, you have databases and backups and monitoring and performance analysis and... arg. It's a lot of work.
And that's _done_ now. Powerset has a real infrastructure serving real queries to real people off a real index which is updating in real time (as fast as we can afford to do it right now). This isn't the end of Powerset's effort, it's the beginning. I wish people would take it as such.
Expectations management is important. That said, if you guys have the resources to persevere, this might not matter. Ask Microsoft about the Zune....
Powerset didn't apply this moniker. In fact, the PR has been trying to deflate it. The media hungers for the fall of the Big Guy, and any story that can play the angle of "The Next Big Thing" gets more hits than, "Check out this interesting startup!"
Maybe we could have managed it better, maybe we couldn't have. My department is engineering, so I don't know. But I'd like to pretend at least a few people are interested in the more moderate but still fascinating reality. :)
The executives talk about "revolutionizing search" and "changing the way people interact with information."
Gosh, that's just what Google did. :P
Perhaps you'd be better off listening to your critics than antagonizing them.
2. Even if I restrict the Google queries to site:wikipedia.org Powerset does no better. That's why I said, even at face value, I just don't buy the product.
3. Excuses. You guys have been parsing Wikipedia for over a year now. I remember the days of "Who did PeopleSoft acquire?" versus "Who acquired PeopleSoft?" This current product is still on that same level.
I have a background in computational linguistics so I understand the scale of what you all trying to do. But the best the web developer in me can muster is a big shrug.
Doesn't matter. In this game it's all perception, and if you're launching a consumer facing product your first major PR volley could be your last.
The caveat above doesn't come through in a headline. And it's little help that PowerSet is solving a problem no one seems to have quite yet. The theme of the day is Google comparisons, but in my mind unless PowerSet is blowing your mind on it's own without comparison, it's dead in the water.
That might seem logical from a technical point of view, but your users will want to find answers, not search wikipedia articles.
It's (just about?) an impossible task to get anyone to switch from Google, so the _only_ chance they have is to imply that they are in fact different, much better, and yes, Google-killers. So, while I agree with the article in that the results aren't there yet, the only way they're going to get any traction is if they're simultaneously making the product better and screaming very loudly that they're awesome.
So, while they may have under-delivered, I don't think they've over-promised, because I still haven't heard any message, and still don't care much. I'd be more than happy to give'em another 30 seconds in 6 months to see if it's better than "key words wikipedia".
Maybe the way I speak isn't natural enough.
This offering just doesn't stack up versus Google, even when you use both solely to search Wikipedia.
My feeling isn't annoyance, but disappointment. Where's my revolutionary new search engine?!?
I think I'd want context for all sites, or not at all. I prefer consistency over occasional glitz.
But you have to be extra-careful about publishing those aims two years before a product launches. It will force you to deliver at 100%.
Risky risky.
i mean this is essentially air.
like norvig said "capital of france" is the important part, not "what is the capital of france?".
oh well i wish them good luck and hope the succeed. competition is a good thing.
If this is the case, then they might be in trouble. They built up too much hype to just be a search tool for wikipedia. If they can upgrade their technology and search much more of the web then we'll have a better idea. For right now? Meh.
Is this the best they have to offer after all that time? I hope not.
I was talking with a friend about the name for a blog. My first conception was a website where I posted lots of tiny bits of code, centered around Web 2.0.
So, lots of little bits of Web 2.0. Hmm, "bit," like bits and bytes. Then I thought of Web 2.0 tropes and came up with 20bits since that was the trend (37signals, 30boxes, 43things, etc.)
And now you know.
I can’t believe people actually finance such projects.
They are probably running out of money and need something to show off to get more funding! :)
Pretty cool, although it's the results that matter, not the technology.
Powerset's been doing this for more than a year and Google is already better at it, at least when I compare the results side-by-side.
http://search.live.com/results.aspx?q=president+of+france
When I was at MSN Search 3 summers ago, they were objectively ahead of Google in that area.
EDIT: I didn't realize that Google's search result answers came on a OneBox. I just thought their first search result had been dead-on.
If this is the case then users would require the least training and get the maximum benefit if Powerset had a pull-down menu with {who, what, when, why} before the familiar search box.
I don't see a single reason to use Powerset over adding "wikipedia" to my query if I want to search Wikipedia only.
They've marketed themselves into a corner -- now it's Google-killer or bust.
I know lots of people inside Powerset and they're all wicked smart. This isn't a critique of them. It's an expression of my disappointment in what has been a massive buildup over the last year and a half.