Slashdot Mirror
Computer Scientist Calls For Web Search Shake-Up
alphadogg writes "Given the seemingly non-stop battle between Google, Microsoft and others in Web search, you might think this is a pretty fertile area for new ideas. But a University of Washington computer science professor thinks otherwise and is calling on academia and industry to get way more creative. Timed to coincide with this year's 20th anniversary of Tim Berners-Lee springing the World Wide Web upon us, Oren Etzioni Thursday will have a commentary titled 'Search needs a shake-up' published in the journal Nature. The main obstacle to progress 'seems to be a curious lack of ambition and imagination,' Etzioni writes in the piece, which he acknowledges 'is meant to be provocative.'"
I think this idea is DOA. A startup vs Bing vs Google. Enough said.
C'mon.
This dude's name is just a scrabble draw of the 6 most common letters in the English language...and z.
Yeah, it sucked.
And it's over web search?
That's odd, I thought it was just Google.
I'm sorry, but I Bing doesn't really count as a serious search engine, at least for me.
Search engines aren't about finding stuff, they are about shoving stuff into you in a way that maximizes ad-revenue. And as far as I can tell, they couldn't do that any better than they currently do.
But I need access to a data center with thousands of servers, petabytes of storage, and gigabits/s of bandwidth to demonstrate it.
make imaginary.friends COUNT=100 VISIBLE=false
Maybe Search needs a Problem shake up. Innovation is great, but when I search now, more than ever, I quickly find what I am looking for. Spam results remain an issue, but for the most part, I have what I need in seconds.
So, what's the problem? Why does search need a shake up? Do we need to manufacture new problems? The drinking straw has been around for a little over 100 years. Get creative beverage engineers, make a better straw. One that doesn't suck. ... Oh.
I understand that the piece is admittedly supposed to be provocative and start a dialog. So, what search problems do you have today? Is the fact that most search is text based something to overcome? Is there a frontier of a search engine that helps us ask better questions? How about analytical engines like wolfram alpha? Where can these technologies go? What else IS there that we need?
We'd better start googling for some new solutions.
Shenanigans!
http://www.prnewswire.com/news-releases/decidecom-launches---helps-consumers-purchase-electronics-with-no-regrets-124179959.html
"No other team has the technology, talent and experience in predictive systems to solve this problem," says Oren Etzioni, Decide co-founder and computer science professor at the University of Washington. "We've built the only broad-scale model lineage, text and data mining systems that predict future price and model releases to address this complex consumer problem."
The dude is just plugging his shopping-search engine, and astroturfing a computing conference as part of his marketing campaign.
What a cock.
Innovate. That's an order.
Perhaps the lack of creativity and ambition is from the threat of lawsuits....patents, trademarks, and copyright.
Google; left useless menu, result page right and top ads, corrupted results(ads), auto-complete, g. instant, etc.. yuk.. It's about as bad as it can get... Alternatives needed, like webcrawler in the mid 90's or google in 2000. Ad free good results no bloat.. Thats a START.
All the search engines now generally available use tag-based methods. Among librarians -- who are the real professionals at information searches -- that's a method for quick superficial results.
Promoting deeper research and understanding is best done with subject-based methods. E.g. the way libraries are organized, Library of Congress cataloguing system, etc.
Problem is, AI is nowhere near good enough to do that yet. So you need to hire humans, lots of humans, to actually think about the information. Which is way too expensive.
Another tag-based search engine, even if it's Etzioni's own, is just one more look-alike in the crowd.
Web search has stagnated badly. Or rather it has regressed actually. The databases are getting larger, but the search possibilities are significantly worse than they were in the past. For example, AltaVista had a "near" keyword that could be used to specify that two search terms need to be within 40 words of each other. It also had full boolean logic. What Google offers is pathetic in comparison and is a massive step backwards. Especially with the far larger number of web-pages today, I find myself regularly wading through pages and pages of irrelevant results. (Yes, I am using all that Google offers in semantics, thank you.)
I know this is due to Googles technical architecture. But that itself is outdated. They could do AltaVista (classic) like search. Instead they basically have the same design as 10 years ago and PageRank now fails regularly to performm not that it was that good before. Maybe this is also because Google has very low incentives to do better, as they are a) King of the hill and b) make their money with advertising, i.e. as long as they are noticeably better than the competition, they are fine. And other seem to have trouble reaching even the (relatively low) quality of Google search results.
Truly a sad state of affairs.
Most ACs are not even worth the keystrokes to insult them. Be generically insulted by this and ignored otherwise.
find a way to innovate UW FTP?
Google is doing an outstanding job getting information collected. But at the end of the day it's just a "dumb" search that matches keywords with content trying to hide results that are just junk. Google is all about quantity of information with some basic garbage filters. There is no real relational information being gathered between content sources.
The next step in search should be to start analyzing information based on themes and returning results that may not have the same keywords, but are related in themes. Also, ranking results by reading level, length, etc would probably go a long way to getting rid of junk and help in research. Real text analysis. That would help meaningful content get to the top of results.
Work Safe Porn
I've been steadily using Blekko more and more as an alternative to Google. Its back to simple, with the addition of slashtags to refine searching. The complete banning of content farms caught my attention. its still not ideal but it is growing.
... who's hoping to monetize his public-funded research and make even more money for himself.
Seriously, have any decent and successful academic search engines NOT been sold off to the highest bidder? I remember Metacrawler came out of UW as well - it became popular and was "commercialized". As was WebCrawler, for that matter. And we all know about Google, of course - those Stanford guys did pretty well for themselves...
#DeleteChrome
Viva Gopher Space!
Cuil aka The Next Google Killer aka BIG FAIL
If you can't do, teach, then call for others to do.
"Promoting deeper research and understanding is best done with subject-based methods. E.g. the way libraries are organized, Library of Congress cataloguing system, etc.
Problem is, AI is nowhere near good enough to do that yet. So you need to hire humans, lots of humans, to actually think about the information. "
This is not true. Automatic document-topic induction techniques (look up Latent Dirichlet Allocation, for example) are fairly sophisticated these days. It's not strong AI, but it does human-useful soft-clustering from text alone in a way quite useful for search. Roughly it attempts to figure out the intrinsic subject(s) of your search.
It means that you can search for documents using keywords A,B,C and it can return documents which have none of those words and yet are likely to be related because they share many words with the set of documents which tend to have A,B,C frequently. A search engine can also remember recent searches in a session whose desired topics are likely to be correlated.
All the search companies use close to state of the art versions of these methods.
(Google is finding that good weak AI combined with Big Data can do remarkable things --- statistical language translation is much better than I expected, and that's substantially harder than document clustering).
Go DDG!
http://ddg.gg
First of all, there's an opportunity for other search systems even within Google - as TFA says, Classic Google Search isn't really designed for the constraints of a cellphone screen (much less for voice-based searches, where the keyword model might not even be the right engine to put underneath the user interface, unlike mobile-phone search where it probably is.) A good mobile-phone search UI would be a real improvement, whether or not you end up selling your startup to Google, or marketing your search tool as an Android App..
Second, any time you're talking about Design as a strategic business tool in an even vaguely computer-related area, you've got to think about Apple. They may not want to eat your lunch today, but if they ever do, they'll come out with a product that's insanely great, paradigm-shifting, and shiny, and you'll have to deal with them. Maybe it'll look a better product in your current space, like iPods taking over the MP3 market, but then you find out that iTunes is at least as important as the hardware itself, or maybe it'll be something even more subversive, like an App Store.
And then there are companies like Tellme or Genesys, which have been in the IVR / dumb-phone voice search space for decades, and probably have cool things to do in the smartphone space as well, or Nuance (who own Dragon Dictate and a few other technologies.) (And Tellme got bought by Microsoft a few years back, so there are some hooks into other large product sets.)
Bill Stewart
New Fast-Compression-only CPR http://preview.tinyurl.com/dy575ks
Yahoo may have been in more things, but Altavista was really the search engine to beat, and Google beat them. And of course there were other search/portal companies (like Excite, which @Home unfortunately decided to buy/merge for $Nbillion.)
Bill Stewart
New Fast-Compression-only CPR http://preview.tinyurl.com/dy575ks
Science fair gold medalist, 17, invents better way to search Internet
You can find the actual paper online if you'd like. Being as it is through Nature, and he is not writing it as part of an NIH-sponsored project, the paper is behind a paywall.
Fortunately it is in the main Nature journal, which is possibly the most subscribed-to journal in science; hence if you don't work for a place that subscribes, you can probably get it at your local library.
And no that is not an endorsement of putting academic research papers behind paywalls.
Damn_registrars has no butt-hole. Damn_registrars has no use for a butt-hole.
There is no shortage of software products many of us use online that are rooted in university CS projects. Instead of a generic call to arms, perhaps this CS professor could do something about it. He has the means and the resources at his disposal, after all.
It means to take a bribe to ruin a good product by integrating an inferior search function. "Verizon binged my Droid".
Help stamp out iliturcy.
You have quite an unusual name, Mr. Thursday.
Yeah, but if I started coming up with content that Google couldn't read, it'd be bad for SEO. My rankings would drop and nobody would come and click on my AdSense ads!
ha ha ha - you know what else has not been innovated on in a long time? the fork! this is because once it got figured out people were done with that and moved on.
instead of giving me a ticket why arent you out there solving real crime!!?
Perhaps the lack of creativity and ambition is from the threat of lawsuits....patents, trademarks, and copyright.
I agree this is the number one OBSTACLE in any endeavour to inovate.
With services like DuckDuckGo ( http://duckduckgo.com ) trying to improve on the presentation, privacy and relevance of search results, services like TinEye ( http://tineye.com ) or even Google Images itself ( http://images.google.com ) allowing directly image-based searches and tons of ways to do voice- and image-based search on mobile phones, do we really need even more companies trying stuff like this? On my phone alone I have access to Google Goggles, SnapTell, TinEye, Google Voice Search (in English and my own language, as well as various others), DuckDuckGo and a bunch of searches across specialized areas like movies (combining the likes of IMDb and Rotten Tomatoes) or in specific collections of data like WikiPedia.
Sure, for specific domains, Google or Bing (or any of the others) won't do, because researchers and people with specific demographics will have specialized needs. Well, as the San Diego State Univertsity puts it "there are approximately zillions of specialized search engines" ( http://webquest.sdsu.edu/searching/specialized.html ).
You can complain about how Natural Language has failed to deliver, or at least how it appears that way, but plenty of clever stuff is finding its way into products of the main search engines. Frankly I wonder what people expect exactly, considering the incredible rate at which the amount of available data is growing and how search engines have been improving steadily in my experience, in spite of that trend. I get the feeling people confuse their own getting used to the objective quality and expecting or hoping for more with an actual decrease in quality.
I guess we will see complaints like this until the singularity is here and you can just have some machine deal with the pesky business of searching (and thinking about the right question) for you. And depending on your point of view, that may coincide with the freezing over of Hell.
How about a checkbox that gets rid of all the bot sites in the search. Sites that just copy shit from everyone else. Sites that don't have any real information.
Not since Multics has academia showed much ambition and imagination in computer science.
Peer review is a convergent process viewed over decades and centuries. It's not a particularly good local filter. For my taste, it gets far too much credit for being an effective gatekeeper (most of the time) against errant nonsense. Our pathetic current generation search technology supplements the stem "brog" with a potent example, almost as if it could read my mind.
We don't think that peer review in software amounts to accreditation of bug-free perfection. In a good setting, it might be a fairly effective defense against architectural howlers. Is scientific peer review somehow magically better? As Einstein once said, "the secret to creativity is knowing how to hide your sources". His corollary is less often quoted: "the secret to authority is to hide your scowlers".
In the rare case when we peek behind the scientific curtain, it's an unseemly mess of data collusion, warranting special commissions to assure us that nothing was actually as bad as it looked on virgin grep: political cabals to render suspect data in the best light are business as usual. (We did learn that a paper making sophisticated statistical claims doesn't necessarily number a statistician among the secretive panel; as per established custom, the rubber stamp is equally valid either way.) The commission is right in their verdict as viewed against actual scientific norms if you shed the idea that peer review is worth a pot to piss in divorced from it's established track record for convergence to scientific sanity after a decade or two or ten.
This new-age search guy missed the entire AI memo. The field of AI had all the creativity in the known universe (and then some) back in the 1960s when pronouncing imminent breakthroughs.
And besides, Wolfram Alpha has already staked a huge position in search as-we-don't-yet-know-it, with a whole lot more credibility behind it than anything I expect to see delivered from the hands of this puff-piece pendulum.