Introduction

People with little experience on the Internet are usually overwhelmed by the amount of information what can be found on it. They start learning to use search-engines to find interesting sites. Typical users type `games', and they get millions of locations which point to pages, which point to pages, which may actually contain some games. After a few clicks, they find a game they like.

For professional use of the Internet, the requirements are harder. Professionals do not just want some answer on their query, but preferably the answer. This calls for extended ways of asking questions which are not provided by the usual interfaces. If using the search-engine becomes a bit harder is less important for professional users, because they can spend time on learning to work with the more complex interface. An investment in learning is returned by the speed gained when searching.

What kind of questions will a professional searcher -- a librarian, a scientist, a journalist -- ask?

"Give me information about keeping monkeys at home. I like to find small sites which are specialized in the subject. But, I'm not interested in Zoo's."
"I am looking for a page which list links to sites about monkeys." an indexing-site for the subject.
"I want the most popular site about monkeys."

When I say `monkey', then pages about Gorilla's, Chimpanzees, Baboons, etc. should be found, too.

Can you put any of these questions to existing search-engines? To be able to ask such specialized question, the search-engine must produce more data, and the interface must be able to handle the various results. The human must have tools to play-around with the initial results for some time to find the answer, and be continuously assisted by the engine.

When we look at modern search engines, it is obvious that getting the thing to work is hard. When a question is asked to the machine, it responds very quickly. You hear the developers cry: "hey! it works, at last!", and then go on a well deserved four weeks holiday. There is no energy or money left to develop an intelligent interface. Of course, there are some experimental interfaces on Internet, like the `refine' option which was available for some time from AltaVista, but they are extremely rare. Presentation of the results is usually based on pages which are hit, showing a part of that page. Huge lists of answers are return, and you have to filter-out the data yourself... with little to no assistance by the search-engine (no after-care). Effort is put in ranking the results, but not in reducing the amount of results.

Search Engines should spend more time on after-care (and also pre-care) to help people find the best answer. Only at the end, when enough details about the question are known to give a small set of good answers, real pages or sites should be shown. Whilst all current spiders focus on simplicity of their interface, to attract as many users as possible, the interface presented in this paper concentrates on the requests of experienced and trained users. This interface is designed to handle huge quantities of information, and is not designed for indexing one single site, even if it is a very large site.

Next The Selection Process.
Up Main page.