linking pages


Libraries are usually subdivided in the usual hierarchical fashion: fiction over here, non-fiction over there. It might be reasonable, though, to put some non-fiction books in the science fiction collection because some science fiction authors are also scientists, or they are very interested in science and how it's going to develop. The point is that some of the books in a supposed "fiction room" could in fact be non-fiction. Also, some of the books in a non-fiction room might well be fiction (for example, some of Martin Gardner's books---not to mention Whitley Streiber's books).

It seems reasonable to decide what to put where based on the predominance of publications of a particular author. Since, for example, Marvin Minsky co-authored a science fiction novel with Harry Harrison, that novel should be in the science fiction room, yet all his other books should be in the AI room, since that's predominantly what he does.

The way web search engines catalogue things, however, Marvin Minsky would be listed under both sections, and each listing would be equally "strong". Or, rather, neither would have any strength, so there would be no way to distinguish the two. A searcher who knows nothing of Minsky would not see that his novel with Harrison should really be in the fiction section, not lumped in with his non-fiction.

Here's the point: page X has many associations with other pages. Those associations aren't necessarily equally strong. What's important is how strong it is compared to the association between page X and the pages it's otherwise most similar to. If the association between page X and page Y is much weaker than the association between page X and pages A, B, and C, then page X more properly belongs with pages A, B, and C, and not with page Y.

There should be a connection between The Turing Option (the name of the science fiction novel that Minsky wrote with Harrison) and The Society of Mind (one of Minsky's AI books on a related topic). Similarly, travel books should be linked to history books when the travel writer is describing a visit to, or even simply mentioning, a place that a history book discusses.

Keep this up and soon there would be connections everywhere. Usually, however, the user won't want to know that there's a connection between Kim Stanley Robinson and the travel section (Robinson writes novels about Tibet as well as other kinds of novels). What's important is how often the user has been looking at the Kim Stanley Robinson section in the fiction room and desired to go to the Travel room (or conversely).

Consequently: the system needs a stringer or tour guide who puts strings of different strengths between pages. The strengths would be proportional to how many times the user went from page A to page B and the string would itself be an object that could be manipulated inside the system, just like the pages it connects. It should have its own attributes just as the pages do. This actor is watching the user but it's also organizing the data.



last | | to sitemap | | up one level | | next