Ixquick and Multi search engines
By now I’m sure that we’re all aware that just using one single search engine on its own isn’t a particularly effective way of searching the Web. Even the large search engines which index 200-300,000,000 pages hardly scratch the surface of the one billion or more pages that are currently publicly available. If you do find what you need with a single search engine, all well and good, but in many cases people are very often dissatisfied with that they get, and finish their search wondering if they have done all that they can to get everything possible.
This is understandable, and of course it’s never going to be possible to get everything available, however good the search engine is, and how skilled you happen to be at searching. The most that you can do is to do as much as you can, and remember that you’re dealing with essentially an infinite amount of data, rather than the finite datasets that we’ve all been used to dealing with in the past. In order to get a glow of satisfaction (or at least, to still the nagging doubts in your mind) it’s really necessary to utilise a number of different search engines to get as comprehensive view as possible of those web pages that are out there.
This is where multi (or meta) search engines come into play. If I’m being pedantic, I suppose that I really shouldn’t call them search engines at all, since that’s not what they do. They really act as go-betweens between you and the search engines, by taking your query, passing it onto the engines to execute, gathering the results and displaying them for you on the screen.
Most multi-search engines only use a small handful of search engines to look for material; the Big Hub for example only uses half a dozen. Ixquick  uses a total of 14, as you can see in Figure 1 below.
Figure 1: Ixquick search dialog box
To be specific, those engines are AOL, AltaVista, Direct Hit, Excite, Fast Search, Goto, Infoseek, Live Directory, Lycos, MSN, Open Directory, Snap, Webcrawler and Yahoo. This is a nice wide spread, covering many of the top engines, although HotBot and Northern Light are noticeable by their absence, which is a slight weakness. Of course, it is not necessary to use all the engines and the tick box does give the searcher the option of searching a smaller number, though Ixquick works so quickly the amount of time saved by doing this is negligible.
An advantage of multi search engines is quite clear therefore; they provide a much more comprehensive view of available webpages than you’ll get by using one single engine, and they also execute the searches much faster than it is possible to do manually. However, there are disadvantages to this approach which have perhaps limited their value in the past. Since a search is going to be run across a number of different engines, it’s necessary to be careful about the syntax that is being used - most search engines will understand that ‘+’, ‘-’ and “…” mean include, exclude and phrase, but this cannot always be relied upon. Furthermore, once you start to use slightly more advanced syntax, such as domain:uk many search engines will have no understanding of what they are supposed to be doing, and may actually look for a string of characters ‘domain:uk’ rather than finding websites with .uk in the URL. Consequently, while you have comprehensiveness, you also have to deal with simplicity. Ixquick however does not exhibit this problem; it understands which certain engines will understand which syntax, and translates and forwards the search only to those search engines which will properly understand and respond to them. Consequently, the relevance increases, inappropriate results are excluded and users don’t need to learn the intricacies of the way each engine works.
Multi-search engines display the retrieved data in a variety of different ways. Several multi search engines will simply display lists of web pages in the order that they were retrieved by the search engines, so you will get a page of hits from AltaVista, followed by a page of hits from HotBot and so on. This is seldom useful; a good multi search engine will utilise anything up to 14 search engines, and it can quickly become tiresome wading through the results in this linear fashion. Worse, by the time you’ve got to the end of the list you’ve forgotten what was at the beginning of it. Finally, it can be difficult to see which pages are returned most often, and this is a shame, since it can be a very good indication of how relevant a page is to your search; if lots of search engines return that particular page it’s likely that it’s going to be of more use to you than a page that is only returned by one search engine. Some multi-search engines take a different approach however, and will collate the results, de-duplicate them and then display the results using their own classification scheme, or will provide a single ranking, based on their own algorithms.
In most cases, none of the approaches are terribly helpful. Search engine developers work hard on their relevance ranking systems, and each engine has it’s particular strengths and weaknesses - some rank higher for pages that have lots of links that come into them, others on word positioning and so on. If this is all thrown away by the multi-search engine that engine is essentially claiming that it knows better how to relevance rank.
Ixquick however takes a different approach, working with the strengths of search engines, rather than against them. Ixquick looks at the results that are returned by the various search engines, and it awards each page a star for any search engine that placed that site in its own top ten. It then further ranks by the position of that page in the top ten.
Let’s explore this by using ‘Ariadne’ as the search term. The fourteen search engines used found at least 45,839 matching results, which it was then able to distill down to 48 unique top-ten pages. The top result was our very own “Ariadne - Library and Information Science Journal”, that position because a search done using that search term returned that web page by Direct Hit, Yahoo, Goto, Fast Search, Altavista, Lycos, and AOL. Furthermore, it was in first position with the first 4 of those engines, in second position with Altavista, fourth with Lycos and seventh with AOL.
The second highest ranked web page was “Ariadne: Entrypoint - Version 5.0” which appeared to be some type of computer classification scheme, and it was ranked by Lycos (2), Altavista (3), Goto (6), Fast Search (7) in the positions given in brackets.
The further down the list of the 48 unique top ten pages you go, the smaller the number of search engines that return pages, and the further down the top ten those pages are to be found, until you get to the logical conclusion of one search engine returning one web page in tenth position on its list of hits.
Ixquick also provides an equivalent method of searching for news, though this is rather less impressive, given that it only searches Associated Press, CNN, Los Angeles Times, Reuters, San Francisco Chronicle and the Washington Post. Admittedly some of those resources such as Reuters and AP are right at the top of the news tree, but a search on ‘Tony Blair’ only returned 28 unique pages, the first of which was dated for May 1998! Rather than just rely on that single result I also looked for ‘Bill Clinton’ and Ixquick returned 31 hits, and none of them were particularly impressive or appropriate. I don’t quite understand why Ixquick are using this approach, since I think it slightly weakens their overall success and in the area of news information it just doesn’t seem to work.
In common with everyone else these days Ixquick also offer an MP3 search facility, utilising 5 different engines; 2look4, Astra Web, Lycos, MP3 and Oth.net and both searches that I ran (‘Meatloaf’ and ‘sisters of mercy’) worked well, giving me a variety of appropriate songs very quickly.
When the user goes to view any of the pages that are returned Ixquick will open a second window on top of the first, taking users directly to the page they are interested in. I can understand why they do this, since it does allow users to easily keep their search results but explore new pages. However, having watched a lot of new users to Ixquick searching and then visiting their results I have to say that it does confuse many of them, and many get confused when they try to return to the Ixquick page; it’s only the sharp eyed users that notice the ‘Back’ option is greyed out and who think to close the second browser window. Ixquick might want to consider the AskJeeves approach of displaying returned hits by clearly framing them, while giving the opportunity to remove the frame and go directly to the web page.
Ixquick: a summary
The criticisms are minor however, and I was informed last week that Ixquick is due to revamp the site shortly, so my quibbles may already have been addressed by the time this column is published. In summary, I think that it’s an excellent search engine and has, within a very short space of time, become my preferred multi-search engine. Its method of displaying results uses the power of individual search engines, rather than working against them, which provides a more accurate and comprehensive result. The display is clearly laid out on the screen and the ranking method used is clear and easy to follow, even for novice searchers. I think it certainly deserves a place in any collection of bookmarks, and it’s worthwhile seeing how it compares to your own favourite multisearch engine. I don’t think you’ll be disappointed.
Phil Bradley is an Independant Internet Consultant.