WWW Text Search Engines
This article briefly summarizes the different text based WWW search user interfaces.
There are two paradigms for the user interface: text keyword search as found in Google and subject guides such as Yahoo!. Some search engines such as Go and Lycos try to cater to both paradigms. I find both of these paradigms to be simple and intuitive, however neither paradigm is easily scalable to other kinds of media. The text keyword search tends to be more general in that you are not limited to the subjects listed on the initial page. However, the subject guides can help in conceptualizing which categories the page might fit into. What are the best user interface features? At the top of my list is AltaVista's ability to customize the search engine to sort the results according to the keywords which I think are most important. For the first time user, I would recommend HotBot's user interface in that it makes the advanced searches as intuitive as possible.

The results pages for the current search engines typically demonstrate a tradeoff found in pattern recognition, that is, false alarms versus misdetections. A false alarm occurs when the search engine indicates a web page is relevant when it is not. A misdetection occurs when the search engine misses a relevant web page. In pattern recognition, you can typically set a parameter to determine whether you will have many false alarms & few misdetections or few false alarms & many misdetections. This relates directly to the results in that the text keyword search engines usually have many false alarms & few misdetections and the subject guides have few false alarms & many misdetections. In my opinion, for the same queries, none of the keyword search engines have more relevant results than the others. One notable feature is that AltaVista groups mirror sites under one heading.

The search engine, Google, claims to have more relevant searches, because it implements a network based weighting system which favors pages which carry more authority. Authority is measured by the analysis of the link structure of the WWW. Roughly speaking, it is assumed that pages which are linked to more frequently are more authoritative.

Why use only one search engine when its possible to search many of the at once? Metacrawler will let you "have your cake and eat it too." It allows users to search multiple search engines from one simple interface.

In summary, all of the WWW search engines are improving. It is yet unclear which method works the best.

Media Lab Overview
LIACS Homepage
MM Conf
ACM Multimedia
Science Direct
IEEE Library
LIACS Publications
ACM Digital Library