These are the impressions I got after 40 minutes of tests:
- 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.
Summary: It's an expectations game and the combination of all the pre-launch hype and the execution of this launch did them in. A classic case of over-promising and under-delivering.
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.
No one in a position of importance said this demo was a google killer. It's not, it's basically a technology demo and hopefully a usable tool.
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.
1. I don't think the demo is a Google killer, but you can't deny that that's how Powerset-the-company has positioned itself in the market. Any product they launch is going to be viewed in that light.
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.
Ya know...I feel for you, but right now you're a victim of your company's own PR. If Powerset hadn't been blasted through everyone's media filter as a Google-killer, I don't think anyone would care that your first product is tiny and incomplete.
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. :)
"No one in a position of importance said this demo was a google killer. It's not, it's basically a technology demo and hopefully a usable tool."
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.
As a completely anecdotal data point, I read a bunch of tech news sites (preddit, newsyc, engadget, etc), but I had never actually clicked through to read anything about Powerset, and had no idea that they'd been working for a year or two.
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".
As of now, it looks like the only thing they're offering is a fresh new redesign of Wikipedia, from a user's point of view. I'm sure there's a lot going on in the background, but none of that technology really shines through, especially after asking it several questions, to which the results were pretty average.
Exactly. And I know from an engineer's perspective this is tough. The stuff behind the scenes is really amazing, I'm sure, but the thing that truly matters is the product.
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 get the impression their plan is so grandiose that it probably won't ever reach a completed form... The context they've added to wikipedia is neat, but what would they do to present a site I made a week ago? Are they adding this extra context only to the popular sites? Framework based, perhaps?
I think I'd want context for all sites, or not at all. I prefer consistency over occasional glitz.
Hmm... point taken. Maybe this deserves an Ask YC, but what is the general opinion of naming companies in an quasi-arrogant fasion. Powerset seems like a cocky name compared to Google, which is even mispeled making me think they're humble.
I think the question for Powerset now is how quickly they can make their technology apply to other websites. My guess it they choose wikipedia because the html is fairly tame...it's much easier to semantically parse wikipedia than it is the wild wild west that is html formatting on the web.
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.
Well, I'll tell you the story, but it's not that interesting.
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.)
Yeah, I mean this is what Im so surprised at. This kind of capital invested in a university research environment would have produced so much more, and I should know.
MSN's is nice (I'm not a huge fan of Google's, though I like its being integrated throughout search results), but I think Yahoo!'s is the most broad-minded. They really take advantage of their various hundreds of sites to give you relevant stuff. It's one of the times when having many half-baked sites gives you an edge.
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.
Yet "who is the president of france?" lists Mitterand, Putin, the various Presidents of Ireland, prior to listing the current correct president of france.
"who president france" also gives worse results. It seems to work best when you type: {who, what, when, why} <sequence of words in page>
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.
Ok. The results are almost the same, except that Google's results have a much better UI. While Sarkozy was fourth on Google, and third on Powerset, it was both much higher on the Google page thanks to the ridiculously huge Powerset "facz" box, and it was easy to scan because there wasn't so much visual clutter vying for my attention (like yellow highlighting everywhere).
The reality is that the average person cares nothing for what happens behind the scenes, they care about the results. If you are going to spend 15 million dollars on a Wikipedia search tool and it doesn't offer any significantly better search results than anyone else you wasted money, lots of it.
In that light, how are they going to get better if they don't have their flaws pointed out?
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.
- 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.