Wednesday, October 11, 2006

Folders, Shortcuts, Aliases, Tags, Taxonomy and Ontology: Google Docs and Spreadsheets

Google spreadsheets has become Google Docs & Spreadsheets, it now includes Writely. Some annoyances remain, some have been fixed. What caught my eye, however is that you can now 'tag' your files -- just as you can 'tag' Gmail messages.

A 'tag' is a string associated with the file. A file can have many tags, there's some UI support for tag reuse, but it's inconsistent. You can filter views by tags. Microsoft Outlook categories are the same sort of thing, though Microsoft's implementation of categories is a baroque and buggy mess. The Google UI for adding single and multiple tags, then removing them, is awkward. They'd do well to study Keyword Assistant, a free tagging plugin written by Ken Ferry. Maybe Ken could sell KA to Google for a million or so.

This is the 'new age' approach to file organization: tag metadata and full text search. No folders - or at most (as Gmail) a few fixed folders.

There's no direct "ontology" (organization), you don't put a folder called "chairs" (note the plural) inside a folder called "furniture". If you attach the tag "chair" (note the singular) to a file, then a search on "furniture" (search within folder furniture) won't find "chair". Of course you could apply the two tags, "chair" and "furniture", but clearly this gets ridiculous. Of course one could have an external ontology (furniture:chair, etc), and it could even by an acyclic directed graph with multiple inheritance (see SNOMED), but of course that's a bit futuristic. Tags, for now, are not drawn from an external ontology, they're invented.

Tags are good. I like tags. One day we'll have the option of taking tags from external machine-useable ontologies as well as free-texting them (students, compare this to free text vs. coded diagnoses on patient problem lists) -- then they'll be even more useful. There's nothing wrong with tags, and nothing wrong with full-text search [1] -- but it's dumb to ignore folders.

Folders are a handy UI tool for creating and modifying flexible real-world subsumption (containment) relationships. Folders with aliases/shortcuts allow participation in multiple hierarchies (and with cycles too!). Sure Folders have defects, but that's no reason to toss them out. iPhoto manages to do pretty well with the combination of folders and 'tags' (keywords). Google should study that too.

Interesting stuff for the industrial ontologists among us.

[1] In theory. I've used every full text search app written for XP. Even the best of them, Yahoo Desktop Search, is awful. XP, especially XP with antiviral s/w, is hostile to this class of apps. OS X Spotlight, for all of its many flaws, is far more satisfactory. In any case, full-text search works much better on rich metadata than on traditional documents, which is why it works so well in Outlook (Lookout for Outlook) and relatively poorly on the file system.

No comments: