Results tagged “search” from kwc blog

New Ask.com: which results are paid?

|

The new all-in-one Ask.com search is exciting and all, but it fails one critical test: ads look like search results. They have the exact same appearance and aren't visually separated from the search results. You can't tell the difference between the top search result and the ad because only the very first ad is marked as an ad -- the rest blend in with the top results. Sometimes there's one ad, sometimes there's more, so I personally can't tell without careful reading.

Another knock: search for 'search' and you'll see Ask.com at the top with Google's first mention midway through the second page of results, and no links directly to www.google.com.

A9 drops a bunch of stuff

|

A9 has launched a redesign which features a nice continuous scrolling feature, but their redesign always comes with heavy cutting: bye-bye A9 Maps with street photos, Amazon discount, yellow pages, history, bookmarks, diary, and toolbar. Where other services like Google and Yahoo are racing to add more, A9 has opted for less, which would be admirable if I didn't happen to like what they dropped. The A9 maps and Amazon discount were the only two reasons I ever went to a9.com. The A9 maps were a bit under featured to be killer -- they could never tell me what I was looking at -- but it did make it a useful service distinct from other mapping sites. It's difficult now to look at a9.com and see what they offer that's unique other than their own fancy UI widgets.

A9: What's New

Why relate?

|

Google recently launched Google related links to compete with Yahoo's Y!Q for publishers. Both provide you an automated mechanism for inserting "related" links into your own Web site. I don't get it. Let me start off with an example of a human-authored attempt to incorporate related links (from Wikipedia's April 1, 2006 article, but just about any Wikipedia article will do):

Slashdot incorporated a pink "OMG!!! Ponies!!!" theme [1] at 00:00 UTC. This girlish theme is in stark contrast for a techie website believed to be mostly frequented by male nerds.

The only related links in that sentence important to understanding the subject (link "[1]") were Slashdot and OMG, but there are plenty of other 'related' links to Wikipedia entries about pink, girlish, male, nerds, contrast and Web sites -- great if your goal is to spend three hours on Wikipedia exploring the interconnectedness of entries, but not much more.

A link is a way for someone to leave your page. With either Google Related or Y!Q, you, as a publisher, are saying that you want to insert more ways for a visitor to get the heck out, that the fire marshal has directed visitors to proceed to the nearest related link safety exit in a calm and collected manner.

I use a more profitable system for saying, "Get the heck out": ads. A large percentage of the entries on this site are crap or have only short-term relevance. My entry about Bootcamp won't matter for long, but many more people will read it a year from now because of search engines. I'm too lazy to keep older blog entries relevant and I don't want to delete part of my historical record, so I review older entries from time to time and place my ad exit signs if I think that the entry won't age well... or if I think I can make a buck :).

Ads are just as crappy as related links, but older entries are mostly compost and you can't really crap on crap. At least Google is better incentivized to make ads more relevant. As a publisher, masking the smell with a Google Adsense check helps and there is one good that comes of it all: you can buy fancy new toys to blog about ;).

Behold the awesome knowledge of Google

|

Google had me more than a bit disappointed with GMail + Chat and the whole China censorship issue, but Paul@Icarus Diving gives me warm fuzzy Google feelings once more with this screenshot of all the useful knowledge contained within the Google Suggest search box:

googlesuggest.jpg

No wonder people are trying to wire up their bathrooms with Web terminals -- Google really needs to add "How to use a Japanese toilet" to that list. Try it yourself if you wish to discover what other knowledge may lie within the magical textfield.

via apophenia

Retrievr is awesome

|

(and not just because one of my photos shows up as the default 'nothing' search)

Retrievr lets you search for photos by drawing a sketch. I've seen photo similarity searchs before, but this one's fast and it's tied to Flickr, which means that you might actually be able to use it as more than a toy. I've been debating whether or not to do another 100 photos collage this year. I skipped last year I skipped due to lack of photos. This year I perhaps have too many photos, unless I could have cool search technology like Retrievr built in to my photo management software.

A search with no 'enter'

|

Yahoo has launched Instant Search, it's attempt to one-up Google Suggest. Instead of trying to autocomplete your search terms for you, the top search result appears for you as you type, i.e. you don't hit enter or click 'search.' I can see this approach solving a lot of usability problems for Web search for the average user (where average user is probably not someone reading this blog). According to Nielsen, 40% of the search population is not search literate, which he demonstrated by showing a video of an example user in this 40% attempting to do a search query. The number of different ways other than the correct way the user tried was impressive, though understandable.

The elimination of one the search steps should be beneficial to the 40% crowd. Also, this could be merged with a Google Suggest-like autocomplete -- the autocomplete to speed text entry and reduce spelling errors, and the instant search to eliminate problems with clicking the correct button or typing the correct key.

As currently implemented, though, the Y! Instant Search isn't ready to help that many people. The number of queries for which there is an actually instant search result is very small, so small that you have to adapt their examples searches to figure what works. I see this as a big impediment to my general use because the instant search is just slow enough that it's hard to tell whether the instant search is still being fetched or there is no instant search result. Also, I find the general Yahoo! Search results page (the page you get if you do click 'search') extremely annoying. On my laptop screen the top search result isn't even fully visible without scrolling down due to all the ads at the top. A search engine that places most of its results 'below the fold' is rather worthless.

Google + Blogs

|

This involves Google and blogs, so of course I'm obligated to post: Google Blog Search.

Pros: 1) It's fast. 2) It's really fast. 3) It places a "References" link next to blog entries that are linked to by others. 4) You can get feeds of your search queries (ala IceRocket).

Cons: 1) I can't make heads or tails of their relevance. 2) Their index isn't impressive yet (both in depth and recency). 3) It's pretty easy to get a spam blog or two in the top search results page.

Relevance is a difficult issue. When I search for 'jython' do I want a blog about Jython-related issues, or do I want a recent blog entries mentioning Jython. They try to offer both by having a short list of "related blogs" at the top of each search result listing, but it's far from perfect: I get bp's comment feed instead of his main blog when I search for 'bp.'

You can get off-results even for very prominent blogs. I searched for Scoble, author of "the scobelizer weblog." Presumably, someone who is #30 on Technorati's Top 100 with 6,087 links from 4,003 sites be an easy search result. For related blogs, Google returns "Alex Scoble's IT Notes" (a different Scoble). The top result in the main list of results is "the scobelizer weblog," but the URL listed beneath it is http://www.velveetaland.com, which is a blog that links to Scoble. I get similar problems if I search for "John Gruber" (author of Daring Fireball, #76 on Technorati's list).

Of course, a normal Google result for either Scoble or John Gruber gives the desired result.

Sites like Technorati are all but unusable because of slowness and frequent database outages, so speed is IMHO the competitive advantage here. It's hard to care about feature XYZ when it only works 50% of the time and takes 30 seconds to complete.

BTW: My favorite blog search is still Bloglines. It's slow, and it does have frequent outages, but lets you exclude your subscriptions from the search results and it contains a very broad index. I'm also a fan of IceRocket, which has RSS search subscriptions and tends to catch tons of people who link to my mythbusters entries. It's also pretty zippy, though not Google zippy.

Peter Norvig, Google; Ken Norton, Yahoo!; Mark Fletcher, Bloglines/Ask Jeeves; Udi Manber, A9; Jakob Nielsen, NN Group

I went with bp and Neil to a BayCHI talk on "Recent Innovations in Search." I agree with bp's sentiment -- there were some interesting moments, but the talk was short on revelations or insights. I guess that is to be expected as the title of the talk is past focused ("Recent Innovations") rather than future focused ("Future Innovations"); it's hard to believe that the panelists would give away yet unrevealed technologies they were working on. I'm going to try and save as much effort as possible, given that bp posted his notes. In fact, as I am going to crib from his notes, or just omit what he already has, you should just go read them instead.

Cool bloglines feature

|

The new citations feature that I have just recently noticed on Bloglines makes it even more useful than Technorati in finding entries linking to your blog. It's improvements over Technorati include: * speed (though this is probably due to the # of users on it) * the ability to customize your search to search inside or outside your bloglines subscriptions, which makes it much easier to find the random strangers who are linking to your blog. * includes LiveJournal entries. * seems to preserve entries much longer than Technorati, so you can find much older links.

Of course, it doesn't have the same level of API openness that Technorati has, and despite the inclusion of LiveJournal blogs, Technorati seems to carry some blogs that Bloglines does not, which makes their search results complementary, rather than competitive.

Citations search

Peter Norvig, Google

Notes: Stochastic Local Searches

|

Tutorial session at AAAI

slides

Unintentionally interesting

|

Multiple people had been telling me about Neighbor Search, which lets you look up who contributed to what political campaign. You can either seach by name or you can input your zip code and see what turns up.

I found a particularly devious use of the tool: it happens to be a really good way of looking up the address of famous people. You also find out who they gave money to, as well as what they list their occupation as.

It all started when I started trolling through Los Altos donations and I noticed Andy Grove's name (head of Intel). I then tried the next Bay Area CEO that popped to mind -- Steve Jobs -- and was able to find his wife listed. Silicon Valley CEOs are kinda boring, though so I started typing in the names of famous actors and musicians. At first I was discouraged, until I started typing in the names of known politically active celebrities and came across Susan Sarandon. My first real celebrity, at last! Strangely, I couldn't find Tim Robbins, but perhaps his real name is different.

Typing in the names of various celebrities, as well as looking up their real names to search on, wasn't paying off as fast as much short attention span would allow. I needed to get more matches quickly. I thought to myself, "If I was a famous person, where would I live?" Bel Air, Beverly Hills 90210? Bingo! Between the zipcodes for the two areas the names came rolling in. Here's a partial list (a question mark means it's unclear if it's correct):

- Susan Sarandon
- Andy Grove
- Steve Jobs
- Jon Bon Jovi (?)
- Helen Hunt
- Johnnie Cochran
- Mary Steenbergen
- Mel Brooks
- Ted Danson
- Mike Myers
- Larry David
- Paul Reiser
- Mike McCready (Pearl Jam guitarist)
- Leonard Nimoy
- Melanie Griffith
- Michael Douglas (?)

Also, the job titles you see listed are often interesting. Here are some of the jobs titles from a single zip code search (90200):

- Executive, Warner Brothers
- Producer/CEO, Jim Henson Co
- Chairman/CEO, Warner Bros Entertainment Inc
- Executive Producer, Walt Disney Studios
- Producer, Sony Pictures
- Executive, Castle Rock
- Chairman, BMI Music
- President, Sony Work Group

Clearly, there are easier ways of getting the addresses of famous people. A $5 map to the stars probably gives a thousand times as many names as this, but for some reason I find this indirect method particularly entertaining.

Is this Kaltix?

|

Awhile back, Google acquired Kaltix. Now Google Labs has launched Google Personalized Search, which adjusts your search results based on a profile that you fill out. Each profile consists of a bunch of categories/subcategories that you indicate that you're interested in (there's even a checkbox for birding).

The actual search is the cool part (it also makes me pretty sure this is Kaltix). When you do the search there is a slider bar that lets you select how personalized the search is. It starts off at "min," but as you slide it, the results instantly change to reflect your profile. I tested this with "armstrong." The top search result was "Armstrong Floor, Ceilings." When I started moving the slider, Lance Armstrong and Neil Armstrong moved to the top of the search results (I had checked "cycling" and "astronomy"). I was disappointed to see Satchmo drop from the results, so I went back and clicked "jazz" and Louis Armstrong stayed in the top search results.

The one question I have with this approach, though, is whether or not a high-level, explicitly declared profile is actually going to get the search results people want. I think one of the canonical examples for personalized search is searching for "Java." There's plenty of checkboxes for computing-related interests, but I don't see any for "Indonesia" or "Coffee" that might help out people who don't want results related to the Java programming language. At the same time, if there were checkboxes for topics as specific as "coffee," it might make filling a profile out a bit overwhelming. Perhaps, instead, there needs to be a ternary state to the high-level checkboxes: not specified, interested, NOT interested.

Newsjunkie by keyword

|

PubSub lets you create virtual RSS feeds based on a search term. For example, you can subscribe to Nanotechnology, Stanford, or a search term of your own, and it will return articles from over a million blogs. Here's one I created for the mars rover.

If only Google News could add this feature to their search -- they have news alerts, but those are e-mail based.
(via Brad Choate)

Update: With Brad Choate (maker of fine MovableType plugins) showing up in the comments within 10 minutes of my posting, this made me realize that this is a powerful tool in support of Scalzi's Law of Internet Invocation. Instead of having to actively Google your name to find new mentions of yourself, PubSub will deliver all new occurrences of your name directly to your news reader within minutes of it being posted. Those Internet Spirit Summoning spells will work that much quicker now.

Personalized search

|

jheery seinfeld's roommate's startup Kaltix is starting to rise above the radar:
- Searching for the personal touch | CNET News.com

No one from Kaltix or anybody who actually knows anything about Kaltix would give a real quote to the CNet stalkers, but they did trackdown Jan Pedersen of Altavista for the insightful one-size-fits-all, "[Kaltix] was likely looking to get bought out." (note: my comments on Pedersen's talk have been reposted).

I didn't like this talk enough to actually transcribe the notes. Luckily, heerforce has already transcribed his notes/outline, and they're probably better than I would have done anyway.

In short, I expected a lot more from the chief scientist at Altavista. His talk was so vague compared to Eric Schmidt's Google Forum - most of the stuff Pedersen put up was general knowledge. Also, if you notice from heer's notes, his "Future" section of his talk was very vague (though he was rushed).

Forum: Google

|

Eric Schmidt of Google gave a very interesting talk at PARC. The first half of his talk was about information movements and the second half was anecdotes/information about Google. Schmidt started off with a comparison to electricity: it started off as a big boom, then it became a utility. Similarly, it started off with thousands of companies, and then became very few. Schmidt argues that all big bubbles have followed the model of thousands then few: railroads, auto, dot com. To me at least, this seemed hand-wavy: the auto industry, for example, has a huge parts and support industry around it, and railroad/electricity are inhibited by infrastructure - a company doesn't have to own Internet backbone to produce a product or service.

The funniest anecdote he gave was about the "bias" of Google News. He was giving a talk and someone in the audience asked him what Google's slant was when it displayed articles. Schmidt tried to explain that a computer selected the articles, and thus there was no slant, but the audience member insisted that every news source has a slant. As Schmidt tells the story, he went to the researchers/engineers that were responsible for Google News and asked them if it were possible that Google News had a slant. As it turns out, the researcher that created the program is Indian and put in two biases: (1) International news is favored, and (2) cricket. I had actually noticed (2), because it seemed improbable to me that there would be a cricket sports link everyday on the front page.

Another funny anecdote he had was when he was making a point that cost drives everything. When Google was still at Stanford, they needed to build server casings. What did they use: legos or duplos? Duplos, because they are cheaper.

More stuff below and in the extended comments.

Abstract:
Eric Schmidt, Google Talk
- "Scarcity to Abundance Drives Everthing Tech"