Monthly Archives: December 2004

You are browsing the site archives by month.

Ask Jeeves

Ask Jeeves

SUMMARY
Type of engine:Major search engine.
Relevancy of results:Good for web search.
Features and functionality: Ask offers a lot of features these days, and I think they are good ones.
Quality of help and “about us” pages: Very good. The help pages cover many topics and are easy to understand.
Business model: Sponsored results from Google’s AdSense on ask.com and through distribution to other sites owned by Jeeves, such as iWon and Excite. Here’s a full list of Ask’s properties.

INTRO
Just like the major TV networks, there are now four major search engines: Google, MSN, Yahoo, and Ask Jeeves*. Ask Jeeves? That’s right. With the acquisition and integration of Teoma’s technology, Ask Jeeves has snuck into the top tier. I used to think of them only as a natural language engine that provided mediocre relevance, but things have changed. When was the last time you used Ask? If you’re like me, it’s probably been a while. These days, Ask offers a variety of search features including their own web index search, local search, desktop search, and more.

*This is not intended to slight any of the other high-quality search engines out there. I’m just referring in this case to deep pockets and large audiences.

Check out their Smart Search page to see all the different things the company is up to, including local search, a toolbar, and a host of other things. None of their extra features are amazing and completely different from what other engines are doing, but they’ve done a nice job bundling it all together and the help pages do a nice job of explaining it all.

Advanced Search offers some nice features such as limiting searches to title or URL. In the information retrieval world, limiting searches to fields is core. But it has not taken off so much for web search engines. Yahoo also offers field filters, but Google and MSN do not.

Features
The user preferences page offers a couple features you may want to adjust before searching on Ask. You can set your default location for local search, set the number of results per page up to 100, and you can also uncheck the option for keeping frames on when you click on a search result and leave Ask.com. I hate it when search engines trap me in their frames after I’ve clicked on a site, but maybe other people feel differently since frames are set as the default.
Some other things to point out is that search results have cached, or archived, versions. This is very nice for pages that have gone dead or are temporarily down, but you still want access to their content.

Right now they’re pushing My Jeeves, the personalized version of their services. (Note: they very clearly state that it is still in beta.) Unlike the previous generation of personalization services like stocks, email, news, and so forth, offered by the portals, and still in effect at places like MSN, Yahoo, and AOL, My Ask is focused on personalizing search. My Ask offers the ability to save and organize searches. Users can even add notes to their saved searches. This is similar to what other sites like Furl, Net Snippets and other companies are doing. But with the added advantage of being a search engine, Ask can offer a more fully integrated search and save option.

To see what queries are popular, Ask provides Jeeves IQ. Unfortunately it only shows a few topics, such as news searches and biggest gainers, and they only show the top 5 or 10 queries per topic. I wish they would expand this. I could be wrong, but I seem to recall that a few years ago they published much more about their query stream.

Jeeves also has kids search that has a focus on school subjects like history, science, and math help.

Query Examples
I searched for peanut peanut butter cookie recipe. One of the first things I noticed were the related topics that appear on the right hand side of the page. There were suggested topics such as peanut butter cookies, make peanut butter cookies, free peanut butter cookie recipe, and so forth. None of these refined my query enough to make me rethink my search strategy, but each of them did provide a slightly different tactic with different results. Every time you click on a related topic, new related topics are generated. This is nice because it doesn’t keep you in an infinite loop and you can travel through gradual steps to different topics. My big suggestion here is for Ask to think of a way to mark the path I have traveled. As I kept clicking through related topics and reviewing search results I lost track of what I had clicked on. Maybe there could be a running list that is kept off to the side of the page that shows what I have already looked at, or some other visual mapping clue to help me remember my way.

By changing my search ever so slightly and making cookie into cookies, peanut butter cookies recipe I got totally different related topics and almost entirely different web results. On the first page of results, only one site was returned for both queries. Something seems a bit fishy about their truncations because those two searches should provide nearly identical results; or if not nearly identical, then at least there should be a significant overlap in the result sets.

But OK, so the results are different, but let’s take a look at the relevancy. This is a pretty easy search topic and all the results are relevant in that they have peanut butter cookie recipes on them. Time to try something a bit harder: peanut butter cookie recipes for diabetics. The first thing that stands out, and is not good, is that Ask went from 3 sponsored results to 11. Eleven! I had to scroll below the fold, and then some, just to get to my first web result. That’s bad. So bad that if I were a regular user searching on Ask for diabetic recipes I would leave. But since I’m here, it’s worth checking out the sponsored results. None of them, not a one, are relevant. Two are recipe sites that have regular, non-diabetic cookie recipes and the rest are diabetes sites that do not have peanut butter cookie recipes. None of them have a peanut butter cookie recipe for diabetics. This is an obvious case of the sponsored results matching to one word – in this case diabetes – and providing nine false positive matches. Flipping over to Google I see the same sites with the same display text being shown in their sponsored links section. It seems the relevancy is Google’s problem and the number of results being shown is Ask’s problem.

The relevancy of the web results was pretty good with one exception: there were two sites that were selling a diabetes dietary supplement, but had links to other sites with peanut butter cookie recipes. This is a tough thing for engines to catch because the site provides enough content and link text to match the query, but the goal of these sites is to sell their products. There was also one site with a diabetes recipe for frozen peanut butter mocha parfaits that had as an ingredient vanilla wafer cookie crumbs. So I can understand why it was returned, but it really is not relevant. Also, it is worth noting that the only related topic that came up is sugar free peanut butter cookies. Why did we go from 10 related topics to 1? There should be many related topics for my search, such as other diabetic cookie recipes.

Seeing as how Ask has been known as being a natural language engine, I searched for how many calories do diabetic peanut butter cookies have? The first result, above even the sponsored listings, is a red rectangle with another search box in it that says: Find the nutritional content of… But I also searched for just the word calories and I got the same box. It is not natural language, but it is a hard-coded result based on word matching. I also tried what is the population of Scotland? but there were no natural language type of results returned.

Ask also offers tabs for pictures, news, local, and products, but those reviews will have to wait for another time.

I also want to point out that Ask returns content for some queries. Like for San Francisco weather the current temperature and 7 day forecast are displayed as if they were the first result. Their help section has more info about what they are calling smart search. Other examples are people searches and the now ubiquitous stock searches.

CONCLUSION
Somewhere along the line Ask Jeeves went from a natural language engine to an all around engine. Obviously Teoma was the major factor in their evolution, but I was surprised to see how little placement natural language searches are getting now on their site. Their relevance and features are respectable and deserve more attention from the search industry. I also think they have the nice advantage of being less under the microscope than Google, Yahoo, or MSN, and as such they take more risks without incurring the wraith of the media and the search industry.

Icerocket

Ice Rocket

SUMMARY
Type of engine: Tabs for news, blogs, phone pics, images, and personals.
Relevancy of results: Web results come from Alexa, so I will save that review for another time. Other results, like news and image search, comes from other sources.
Features and functionality: Average. Pretty straightforward.
Quality of help and "about us" pages: Needs Improvement.
There really are no help or about us pages, but that is supposed to be remedied. As Blake Rhodes stated in his Search Lounge interview, "Yes we will have help pages in all sections soon. "
Business model: Sponsored listings. Though they only show one paid listing per query which really helps with their perceived relevance.

INTRO
Icerocket is a new search engine from Dallas, Texas that has the unique goal of being responsive to users. Blake Rhodes, the CEO, maintains a useful blog about the engine and to understand his philosophy about responsiveness this is a good post to read: http://blog.icerocket.com/node/view/14.
He writes, "When people write me with suggestions, what do I do? I LISTEN to them!! Without our users, we are nothing. As you can see, we have made several changes to the site based on the emails we have received." For more information about Icerocket, be sure to check out the Search Lounge interview, with Blake. Icerocket also differentiates itself by promoting two tabs that I haven’t seen at other engines: Phone Pics and Find a Friend. You be the judge if those are useful options or not.
And I should mention that Mark Cuban, the tech entrepreneur and owner of the Dallas Mavericks, is an investor. If nothing else, that fact helps to bring some extra attention to the company.

REVIEW
There are search tabs for:
Web – searches Alexa.com. Blake stated in our interview, "We actually don’t pull Alexa’s results, just the thumbnail images and the traffic numbers. We crawl and index a lot of it ourselves". I have to admit that I can hardly tell the difference. For most of the queries I tried the results seemed to be exactly the same on Alexa and Icerocket. However, for the query Elizabeth Edwards breast cancer, there was the slightest variation in that positions 3, 4, and 5 switched places, but both orderings were equally good. I found a couple other queries with slight changes in ordering, but nothing significant. Rather than review Alexa’s relevancy, I will save that for another article. In terms of Icerocket’s interface, thumbnail snapshots, a la Alexa, are displayed next to each result and there is also a "quick view" option to peek without leaving Icerocket. There is also an "archive" link next to each result that takes you to the Wayback Machine. As I’ve said before, I’m happy any time I see the Internet Archive being used.

Blogs – this is great to see as a choice and I am sure this option will begin to crop up more and more. Blog searches default to "by date", and you sometimes don’t get any hits. I searched for san diego mayoral race and got zero results. When I chose "sort by relevance", I got 10 results, which seems to be the max. You might think that only the very latest results should be returned in a blog search, and for the most part I agree. However, in the sorted by relevance option the posts ranged from 7 hours to a day and a half ago, and that’s still pretty recent. The sorting by relevance is misleading, because all the results were actually sorted by date. Let me clarify: sort by date returned no results. Sort by relevance returned 10 results but they were actually sorted by date. In terms of relevancy, the results were good. Only two of the results were not relevant, and that’s because one was a political blog that mentioned San Diego and the Los Angeles mayor’s race, but was not about the San Diego race. The other result that wasn’t so good was a blog that only posts links to the titles of news stories, so yeah I guess the link was relevant, but the blog itself provided no context.

For another test query, I checked on Yahoo news to see what the lead story was. I grabbed the phrase Saudi attack and plugged it in and sorted by date. The most recent one was "-10037 minute ago", almost as if I was searching faster than time itself! And changing it to sort by relevance returned the exact same results. The results were peculiar. The first result was from Outside the Beltway, OK, that’s good. But the eight other results all seemed to be from the same source, only titled slightly differently. Most were called "Saudi Arabia News .Net – Newspapers on the Net", but there was also "Cairo News .Net – Newspapers on the Net" and "Manila News .Net – Newspapers on the Net", and all the pages were formatted and branded similarly. The articles were all indeed unique, but it raises a couple of points. First of all, should de-duplication occur at the source level so that the user can get a broad range of sources? And if so, how can an engine recognize that these sites are all from the same source? It’s not easy. Also, are these really blog results? They seem more like news results to me.

This is all just a longwinded way of saying it’s great they’re searching blogs, but some advanced features might save the day and make it more useful. And definitely the sort by relevance needs to really be sorting by relevance.

News – n/a. I tried my Saudi attack query, but got no results. San Francisco hotel strike also got no results. It seems the news search must be temporarily down. (Note: I tried it several times on 12/16 and 12/17, all with no luck.)

Phone Pics – this is a feature I haven’t seen on any other engine. I tried a search for Howard Dean and got 10 results, some of which were very relevant, while others were somewhat relevant. The difference being that the very relevant were photos of Dean, while the somewhat relevant were correct in context, but weren’t actually photos of him. For example, there was a photo of a Dean banner. The results are coming from Textamerica with the same exact results being shown in both places. Phone Pics is a nifty thing to play around with, but I’m not sure it deserves a tab at the same level as web, news, etc.

Images – a meta search engine for images. Icerocket aggregates results from a variety of sources. Right below each result it says where the result was grabbed from, such as Webshots, Picsearch, etc. Howard Dean returns 106 results, all of which are relevant. Of course that’s an easy query, but I don’t want to spend too much time on image search. I’ll quote Blake from his Blog: "If you look at our image search, I think you’ll find some images you won’t see on the other engines". I’m guessing he means the combination of meta-results is unique even though each individual result comes from another engine.

Find a Friend – a meta search engine for personals. I have never seen this on a search engine before. It searches from multiple databases based on the criteria you enter, such as desired and age and location. I guess this is a cool feature, but similar to Phone Pics I’m not convinced it’s on the same tier as web search. And it should probably be called personals search.

For other features, Icerocket has an RSS Builder which lets you add RSS feeds to your web site. There is an Advanced Search, but the options are fairly basic. Hopefully they’ll build that up a bit more. There is a Toolbar, which includes an option for Firefox users like myself. And Ice Spy, which shows what other users are searching for.

As I mentioned, Icerocket only displays one sponsored result, and for every query I tried it was a small link titled "Free ipod! ". On the one hand it’s too bad they’re not targeting the ads to my query, but on the other, and probably bigger hand, it’s nice to only have one very small link instead of the three to five common on other engines. And, what’s even more noteworthy is that they are leaving nearly all of the right hand side of their web results’ pages empty. Empty! Can you imagine empty real estate on a search engine these days? It’s like a big gaping hole that’s just waiting for AdWords to slip right in. I really like that they haven’t done that, but I won’t fault them if they do end up doing it. After all, they need to make money.

I hope that for each tab they put up some information about where the results are coming from, particularly since transparency goes along with Icerocket’s goals of being responsive.

CONCLUSION
In my opinion the most compelling thing Icerocket is offering is its Blog search, although I never did figure out for sure if they’re powering it themselves or not. Other interesting things that Icerocket offers are the tabs for searching phone pics and for searching personal ads. I haven’t seen these two filters called out by other search engines, but maybe there’s a reason for that. I hope in the future they make web search more distinctive and unique. Even if they don’t get to the point of doing their own web crawl, they can distinguish themselves by customizing Alexa’s results, customizing their interface, or mixing in results from other engines.

FindArticles – Interview with Chris Broekhoff, Director of Content

The Search Lounge is very pleased to feature an exclusive interview with Chris Broekhoff, Director of Content for FindArticles.com. FindArticles is a free online search engine for magazines and articles. Its database has “5.5 million articles from over 900 publications”. Google Scholar recently went live with a similar product. These valuable services are changing the way people use article databases. In the past these expensive resources have only been available through libraries or other institutions that subscribe to them. Although the end-user gets free content either way, FindArticles is breaking the barrier between proprietary databases and free online content.

This interview was conducted exclusively via email, though in true Search Lounge style it was conceived of over pints of beer.
As a disclosure, Chris and I worked together from 1999-2003 at LookSmart, FindArticle’s parent company.

Chris, thanks for joining us at the Lounge. Can you give an introduction to FindArticles?
FindArticles is a targeted search product that gives users access to over 5 million research-quality articles from a wide variety of publications. The majority of articles in the product are free to view in their full text, and many of them can’t be found anywhere else on the web for free.

What is your role?
I oversee our content acquisition and management strategy. Basically I figure out what types of content our users want and then try to acquire it from publishers and other content owners.

How long has FA been around?
FindArticles has been around for about 4 years. The product had a pretty low profile until we relaunched it in late 2003 with expanded content and new features. Since then we’ve been focusing heavily on expanding our content and delivering features our users ask for.

What types of articles are available in FindArticles? And who supplies them?
Our goal with FindArticles is to create a one-stop shop for research-quality content on the widest possible range of topics. Right now you can find everything from art criticism from a publication like Art Journal, business and finance news and analysis from journals like Business Economics and Money Digest, political reporting from Harper’s, health and fitness information from specialized medical journals and consumer health magazines, and the latest on technology trends from titles like PC Magazine and eWeek. You can see a complete list of publications in FindArticles here: http://www.findarticles.com/p/articles/an_1

Our biggest supplier of content is Thomson Gale, with whom we’ve had a strong relationship since we launched the product in 2000. Beyond Gale, we have a few direct relationships with publishers, and some partnerships with premium content services for our paid content.

How have users responded to FindArticles?
Users have responded very positively to the product, especially since we re-launched in late ’03 and began developing features that people had been asking us to build for a while, such as advanced searching and sorting options, and RSS feeds.

Can you tell us about the current state of searching and browsing on FindArticles? How about some useful tips and tricks for our readers?
Not surprisingly, most users who come to FindArticles enter simple keyword phrases in the search box. To these queries we return up to 2000 articles ranked according to relevance, using such things as word frequency and proximity. However, we offer a number of refinement features directly on the search results page if you don’t find what you want right away. Next to the search box you can limit your search to publications in a specific topic, and above the search results you can sort the articles by date, article length, and publication name. One of my favorite features is the ability to search within a publication or exclude it entirely from the search results. You can do this by clicking the “options” link next to each result in the list. Finally, a popular feature with users is the ability to exclude articles that you have to pay to view by clicking the “free articles only” box. And these are just the features that are available on the results page, by choosing the advanced search option, you can do more sophisticated things like search within certain database fields, exclude terms from your search, select specific publications, and limit your search to publication date ranges and article lengths. While we’ve put more emphasis on our search functionality, users can also browse topic categories and explore publications by issue.

How does your service differ from the databases available through my local public library? And as a follow-up, how do you think libraries should view services like FindArticles?
Unlike the databases you’ll find in your local library, FindArticles is developed as a consumer search product, so it’s generally more user-friendly and easier to search than the mix of databases that libraries offer. Through our experience with other search products, and our understanding of what users want and how they react to our products, we’re able to iterate on features and design in a way that continually improves FindArticles value to users. I don’t think you see that kind of focus on end users in the library databases. On the other hand, if you have the time to dig through them, libraries offer access to significantly more published content than we do, and if you have the expertise to search them, they have more advanced and complex features than FindArticles.

I think libraries should view FindArticles as complementary to what they offer their patrons. We offer casual to moderately serious researchers a convenient and comprehensive product, but we don’t really compete with the expertise that librarians can provide their customers, or the vast collection of published material, both in digital and print form, that libraries offer.

What is the business model for FindArticles? How do the database distributors make money from it?
FindArticles primary revenue is advertising, both contextual and search targeted. We share a percentage of the ad revenue with our content partners.

How many of the articles are free versus premium content that requires users to pay?
We don’t disclose exact numbers, but the majority are free to view.

What do you see as the future of free online article databases?
I’m optimistic that we’re going to see more and more published content available in free web product like FindArticles. There’s been a definite trend in usage towards web-based content, and as search-targeted and contextual advertising products continue to improve I think we’re going to see more publishers realizing that it’s in their best interest to get their content in front of as many web users as possible. At FindArticles we want to be on the forefront of this trend, but we also feel that the way to do this is to have a diverse offering of free and premium content. While we see a lot of value in a completely ad-supported model, it doesn’t necessarily work for all types of content. Our goal is to give our users access to as much content as possible, and give our content partners a variety of options to meet their goals.

Lastly, what is your favorite drink?
Vodka martini

Thanks Chris. Are there any other comments you would like to add?
Just that we are always looking for ways to improve the content and features of FindArticles, and I’d love to hear what your readers would like to see us do.

Welcome to The Search Lounge’s Permanent Home

Welcome to www.searchlounge.org, the new and permanent home for the Search Lounge.

Along with migrating to the new domain, I am also changing my back-end blog software, so please be patient while I get everything up and running.

Thanks,
Chris.

Blingo

Blingo

–This “review” is my way of helping Search Lounge readers win free prizes. The review is definitely on the light side. Good luck. In fact, don’t even read this review, just go to Blingo and win yourself a prize–

Type of engine: Sweepstakes, oh and general web search.
Overall: Put it this way: it’s all about winning the prizes. It’s not that the search is bad, but the results are directly from Gigablast so I don’t know that it’s worthwhile rating Blingo’s relevancy. As far as I could tell there’s no difference at all from the Gigablast results.

SUMMARY
Relevancy of results:
Same as Gigablast, and I need to do a review of them.
Features and functionality: Average. There’s not much to it other than a search box and pictures of the prizes you can win.
Quality of help and “about us” pages: The “about us” pages are all about winning prizes. I couldn’t find anything search related.
Business model: Lure users in with free prizes, and then show them Google ads.

INTRO
Blingo just released a press release today so I thought I’d be right on top of things by doing a brief review for the Lounge. Blingo uses Gigablast’s results so I won’t spend too much time evaluating the relevancy. That’s best left to a Gigablast review.

The prizes are a one year Netflix subscription, a digital camera, a digital video recorder, an iPod, a $250 Amazon gift certificate, movie passes, and a Sony music gift certificate. Just go to their site, start searching, and hope you’re one of the random winners. (I hope I win an iPod.)

UI & FEATURES
No registration is required unless you win something. So you can just go to Blingo and search. That’s different, if I remember correctly, from iWon, where you had to register. They provide “see also” links that are taken from Gigablast and formatted differently. And there’s a “did you mean” suggestion taken directly from Gigablast.

QUERY EXAMPLES
I searched for Oakland public library and there were four Google ads that went more than halfway down the page. I tried coniferous trees of North America and got the same thing. And both of those queries are non-commercial in my book. But it’s not exactly surprising for a site like this. I should say that the results for both of these queries were good enough that it’s not a waste of time to search on Blingo.

CONCLUSION
If you like Gigablast, you might as well use Blingo. Same results, and you can win prizes. So stop reading this and go win a prize.

Post Navigation