One of the entries buried away down my ‘To Do’ list is a little cluster grouped under the heading ‘Serendipity.’ Collectively, this would be a set of tools to assist one (principally) during the writing process: Martin and I got some way down the route of defining some assistance systems that currently don’t exist, so far as we can tell. Some of them, on the face of it, don’t seem terribly hard to implement, particularly since Mac OS X already has an excellent full-text retrieval system and a half-decent summarisation tool, which constitute two of the tricky bits.
The first component of Serendipity was called ‘Prompt,’ and it goes like this: Imagine software that watches what you type. It continuously performs keyword extraction on the last sentence or two, and runs proximity searches on a specified folder of reference material. A secondary window displays what the system thinks might be relevant passages.
One can see this working well, for example, if one starts to quote a line of Shakespeare, and has the Complete Works in the reference folder: it should be straightforward to make the appropriate verse appear just around the time one forgets the precise quote, no?
Of course, this isn’t an original idea. There’s an Emacs plugin to do much the same (from MIT or somewhere), and Microsoft have recently been waving flags about this type of tool. But the other day, I stumbled over a reference to a patent on exactly this, dating from 1999 – a little before Serendipity was sketched out. The patent is held by… Apple.
This suggests several things: firstly, that such systems are indeed possible. Secondly, that there’s a huge layer of subtlety in pulling them off cleanly (I suspect one not-obvious problem is presenting sufficient helpful material without it being distracting: smoothly cross-fading the prompt window from one guess to the next might help, for example). Thirdly, one suspect we will see such a system from somebody, someday.
I look forward to that day.
Oh, and this other patent is worth a look, too. It’s a year older, but obviously dates to before the dissolution of the Newton group on Jobs’ return in 1997: it refers to keyword extraction from calendar entries leading to the automated retrieval of relevant information from other document sources. Which is the sort of logical extension of (what I called) Serendipity Prompt that gets me quite excited.