Slashdot Mirror
Better Search Results Than Google?
Mechanik writes "CNN has an AP article about the next generation of up and coming search tools, which try to cope with the glut of hits that result from 'conventional' search engines such as Google. One tool, Vivisimo, "is like a superfast librarian who can instantly arrange the titles on shelves in a way that makes sense. [...] But unlike libraries, Vivisimo doesn't use predefined categories. Its software determines them on the fly, depending on the search results. The filing is done through a combination of linguistic and statistical analysis." Grokker, another, downloadable program, "not only sorts search results into categories but also "maps" the results in a holistic way, showing each category as a colorful circle. Within each circle, subcategories appear as more circles that can be clicked on and zoomed in on." You have to love the author's use of trying to look for a hotel in France with the terms 'Paris Hilton' as an example of searching gone awry."
I tried this earlier (around noon) when I saw the article. One of my big complaints is that the searches seem to take too long. Google usually is sub-second searches, this seemed to take about 3-5 seconds (this was well before slashdot posted the article, so it wasn't slashdot effect either).
Also, I already do not like the search results showing up in the sidebar with search engines (with mozilla), as that is one of the features I kill as soon as I install mozilla. So, I guess, this search engine has a ways to go before I prefer it.
The searches didn't seem too bad over all, I tried looking for "linux kodak 4530" and its results were not any better or worse than googles. I tried a couple other searches and they seem to be on target about as well as google though.
Norris/Palin 2012
Fact: We deserve leaders who can kick your ass and field dress your carcass.
I'm actually posting this form the browser window of Grokker. Been playing with it for just a few minnutes now, but I can see how something like this can make obscure or broad searches a lot easier. When you enter a search term, Grokker generates a series of circles, each of them representing a subcategory of results for your search term, and each of them in turn filled with subcategories of their own. Searching for "west coast museums", for example, gives me subcategories such as 'travel', 'west coast attractions', and 'history museums'. Once you find your desired subcategory you're presented with a smallish list of matching sites, represented as squares. The categorization seems to make sense most of the time, even if the overall visual effect is remniscent of 70's disco lighting.
I want the fire back.
I think Google's search results are worse after their "Florida" update.
alltheweb.com has pretty decent search results.
John Kerry is a Joke!
You can still find old mirrors of the reverse engineering site, but the only active one I know of is at www.woodmann.com/fravia. The message board is at www.woodmann.net/forum, no crackz, serialz, or warez allowed. Just techniques, tools, etc.
www.kartoo.com - does what the article states without some other applications having to be installed.
I want it to behave as some sort of web-service so I can use a perl module to manage my results. google should have a programming API to extend their service in some way
You mean, like this - Google WebServices?
---- Den ene knappen er powerknapp, den andre er Bender voice knapp "Bite My Shiny Metal Ass"
A search on "to be or not to be" on Google produces 3 erroneous results out of the first 10. Visi-whatever produces 10 out of 10. An improvement.
Searching for info about electronic products is the worst on google.
I use the following along with any thing i want to search and it usually does the trick
-shop -shopping -price -buy -order -shipping".
This no doubts subtracts one or two sites which are good but atleast filters out most of the shopping sites.
for the last time people, I am "frodo from middle eaRTH", not "middle eaST".
Try this search then.
The search phrase was:
"red hair" singer "tom morello"
- - - - - - - - - - -
I am a programmer. I am paid to produce syntax not grammar. Deal with it.
You can get a similar effect in Google by adding a word or two of context to your search. Searching for "paris hilton" gets millions of links to sites claiming to sell the tapes, but searching for "paris hilton hotel" gets hotels in France.
"They redundantly repeated themselves over and over again incessantly without end ad infinitum" -- ibid.
You can write an infinite loop in alot of regexp packages. They would have to have a way of detecting that ( or a very inefficiently written regexp )
Eat at Joe's.
Granted, it requires flash. But it shows you pretty diagrams that help you in refining your search.
the preceding comment is my own and in no way reflects the opinion of the Joint Chiefs of Staff
Google doesn't spam its own listings in return for ad bucks. They do occasionally throw in a "Sponsored Link" but those are always color-coded and usually off to the side of the main list.
What your parent is saying, and I can echo their sentiments, is that there are a million and one crap sites that are keyword-spamming the crawlers. Some really sell the product in question, but most seem to be stealing the review copy from other sites - I saw the exact same listing for a scrolly-mouse on 4 different sites, and none actually sold it.
Google has done an admirable job of staying above the corruption of ad revenue and sponsor pressure. Searching is usually effortless if you can be a little specific.
No, Google is not perfect and sometimes searching can be frustrating, especially when the keywords are necessarily common - "Help" and "Linux" come to mind. But I usually get useful info.
And Google Cache is your friend at work!
GTRacer
- Time to ego-surf!
Defending IP by destroying access to it? That makes sense, RIAA/MPAA. Go to the corner until you can play nice!
If you're up for some maths and some fairly dry reading, check out the paper "Authoritative Sources in a Hyperlinked Environment" by Jon Kleinberg. He describes a search method which takes regular text-based search results and then examines the link structure around those pages. The idea is that pages of comparable content exhibit heavy interlinking. Clusters of such pages can be identified with a recursive algorithm a little like Google's PageRank, and then distinguished with some nifty eigenvector mathematics. This gives you your basic categories, based solely on the link structure.
While the paper doesn't detail how one might label the categories identified, I don't imagine that it's all that difficult to do with some simple correlation algorithms, which wouldn't be language-dependent.
Disclaimer: since vivisimo is down and I've not used it, I could well be talking out of my arse here; this is just one categorisation method with which I'm familiar, and would produce the results mentioned. It may not be how vivisimo actually do it.
Kartoo is a frenchie search engine is kinda different. You sorta need high bandwidth cuz
it uses flash.
http://www.kartoo.com/
link to it here
I still miss the original Alta Vista.
6 out of 10 links on the first page of google are still about "hotel chain heiress Paris Hilton".
Even including the quotes gets you 4 links to pages about the star of the infamous video, and one to an "award winning desktop toolbar with 45 tools!".
This is very difficult to do. Most keyword-based search systems use an inverted index for searches.
The first step involves a hash table that converts keywords into term-ids. Then the inverted index is used: it is a table that holds, for each term-id, a sorted list of document-ids that contain the term. The search process is almost instantaneous, as it only involves operations on sorted lists.
To use regexps, the search engine must convert the regexp into a series of words that match the regexp (a very large set - potentially infinite) and then look them all in the inverted index. This is very slow and, as most users never use the advanced search function, very unlikely to be added to popular search engines until some competitive data structure is discovered.
Google has, among others, a very nice linux filter all ready.
You can get a similar effect in Google by adding a word or two of context to your search. Searching for "paris hilton" gets millions of links to sites claiming to sell the tapes, but searching for "paris hilton hotel" gets hotels in France.
The most under-utilized feature of Google I think has to be excluding keywords. For this query, I would use:
+"paris hilton" +hotel -tape -porn
and probably get much better results. If the word "naked" is never ever going to appear in a legitimate result page, you might as well exclude it.
Same goes for other things. I was looking for information on Microwaves and WiFi the other day... not the ovens, so -oven -food and I got infinitely better results.
--Darren
Man, you guys need to get out more. Google's been synonymous with internet search among a huge swath of non-geek people, and it's been that way for a pretty long time now.
William Safire mentions Google in passing without bothering to define it in the NYT just a few Sundays ago.
Or perhaps you're thinking of the uber-geek reference of the Oxford English Dictionary which now lists Google as a new word for 2003.
They - the people spamming google with this stuff - would just switch to underscores.
Underscores aren't allowed in domain names.
i think he's confusing google with yahoo
I agree. Like they say, "If you don't vote you can't complain about who is elected".
Google has a link at the bottom of each search result screen: "Dissatisfied with your search results? Help us improve".
I think that link needs some Slashdotting.
You've been able to do this for a long time in most search engines. Personally, I find myself often including the words -"buy" or -"search results" in Google to avoid all the spam.