auto-tags

Update on auto-tags

Software

As a follow-up on my previous musings and site development updates on automatic tags and tag links embedded in journal entries, I have an admission to make: maybe auto-tags and auto-links aren't such a brilliant idea after all.

Sure, the system I have developed works wonderfully for tags which are narrowly defined terms and names. But that's not true for all words. I have a tag called "work" in my photo album, but it is simply undesirable to have every occurance of the word "work" link to the tag page when the noun (or verb) is used outside the contect of what the tag is supposed to represent. It simply wouldn't work (manually simulated example of why not). Some example of my current "dictWords" code which do work (hehe): Cheap Thrills, my photography, HorrorPops.

Not being able to use generic nouns and verbs narrows down the number of possible options for auto-tagging significantly:

  1. Don't auto-tag. I'd rather overcome the problems I encounter, but sometimes it's best not to try the impossible.
  2. Use capitalisation to identify text which acts as a tag. An easy solution, but not water-proof due to capitalisation rules embedded in the English language and it requires some consistency from the author, but that could actually be a benefit when it stimulates such consistency.
  3. Use tags to identify tags. Similar to capitalisation, but more robust and somewhat akin to Wikipedia's internal linking system. Somewhat undesirable as manually marking tags can barely be considered "auto-tagging".
  4. Manually ensure that tags are specific terms and names, and not generic words. This would basically mean that tags can not be nouns, verbs or other phrases that do not ensure a unique context.

I'm not sure yet which of these I will choose, although I'm most fond of the latter option as it limits user responsibility to defining proper tags while leaving alone the process of actually writing text. It's not without limitations of its own, but somehow I feel like it is the limitation I could best live with.

I insist that a proper auto-tagging engine can remove the border between fluid content (blog/journal entries) and static content (more permanent viewpoints and collections of information on any given topic) without resorting to hacks such as manually linked or sticky summary posts.

Feedback from people in the web and digital content industry is very welcome. So consider this a casual Request for Comments (RFC).

Considering the efforts of Microsoft (delayed filesystem originally intended for Longhorn), Google (desktop search), Beagle (desktop search) the concept of "personal information space" seems to catch on and it does then make sense to me that web sites also offer a way to integrate various types of content, such as journals, albums, poetry, external links, and so on.

Photo tags

Software

A few weeks ago I added tags to my photography, but I never uploaded the new code. Well, this morning I did. Therefore it is now possible to quickly see all photos tagged cheap thrills, horrorpops, rob, rotterdam.. well, and so on, I'm sure you get the picture. (No pun intended.)

It's not unlikely that as a next step, I will add tags to journal entries (update: done). And then remove the category system I'm not entirely fond of anymore. And then possibly combine tags with my auto-linking which would basically mean not just a revival of the auto-tagging system I had on "Blog of Rob" in the past (called dynamic keywords), but indeed an improvement of it.

© Copyright 1995-2009 Robert John Kaper. All rights reserved.

Tom has more friends but mine are prettier! (#1/1)