Metadata Attributes?

Hi all-

I wanted to throw this out to see what kind of comments folks had. There’s no implied criticism of DTPO, so I’d appreciate any response to uphold a thoughtful perspective.

DTPO has its “AI” approach to classification and search, but another way is by adding metadata attributes to each piece of data in an unstructured body of documents. To do this, one needs to be able to add custom metadata fields to a particular file.

RTF files have this feature, but I have yet to find an application that supports the \userprops control word. HTML allows for custom metadata, but again, I don’t know I want to edit raw HTML or use Dreamweaver as a writing tool.

So how could DTPO support metadata attributes? Well, in some sense it already does through groups and replicants. One important aspect of metadata attributes for me is the “key:value” pair. A DTPO group would have to say more to be a key:value pair. “Green” could refer to pants or shirts, so you want a group perhaps that was “pants=green” or “shirts=green.”

This approach could work pretty well in DTPO, until you wanted something more than a single value pointed to by a key. For example, you might want “game=phillies;mets” and "game=mets;nationals.’ But in DTPO you’d be hard put to write a search term that would look for something in all games that the Mets played. (A later thought: obviously, nested groups provide some, if not all, of this functionality.)

So has anyone else thought about this stuff with respect to DTPO? I’ve just started so maybe there’s some huge body of info that I’m unaware of.

Anyway, if you’d like I’d love to hear your two cents…

Best, Charles

i don’t speak for the developers here. i am an end-user.
personally i don’t think I need every app to do everything. that’s the microsoft bloatware issue all over. devonthink does the AI paradigm well.
in short i use another app for metatag searching using the spotlight technology. i use houdahspot for that. and i examine metatags with hamsoft’s
one of the biggest problems with tags is consistent spelling and use. the use of a dedicated taggng application is also appropriate. i have a free copy of fresh via macheist which supports the open metadata standard. nevertheless i use the app ‘tags’ after my tags are set-up because it remembers what these are and i can click on them like buttons.
no one needs to start tagging the way i do it. one can search in spotlight using “tag:” and manually time tags in file comment areas using get info of the finder but this is highly ineffecient and prone to errors.
part of the attraction of ‘tags’ and houdahspot are the capacity to script when their competitors are not interested in power users.
this is a new paradigm of course. devonthinkpro is very well thought out and there’s scripts and resources in place.
i don’t see tagging as a replacement for dtp- particularly for information workers. tagging can be very useful with mediafiles where info like artist is already there. I can see art departments using tagging definitely.


Are you already aware that a future 2.x release will support custom metadata and posting ideas for its implementation?

No, I wasn’t, that thread must have slipped by. Very good news. I’ve been using Tinderbox quite a lot, although I wouldn’t suggest DEVON adopt that paradigm uncritically. BTW, what’s up with the [tags:] in your posts?

To respond to the other poster, I don’t think the current mania for “tagging” satisfies my interest in metadata. One thing to keep in mind is that lots of existing data (library records, legal documentation) already have metadata attached, and so support in DTPO is in some sense necessary to keep from truncating the information content of a document.

Also, as I mentioned, the current “tags” systems are single valued, and have no key, so they’re stuck in folksonomy mode. I’m thinking of much more precise uses, and, of course acting upon classifications.

(A second read of mrglsmrc’s post tells me he’s doing something more sophisticated than the kind of tagging I described above. Mrglsmrc, can you tell how your create custom (multivalued?) tags with your system? Also, is your use simply to find things, or are you trying to create structures (hierarchies, maps) out of what you find?)

Best, Charles

You know, it’s interesting how conversation ends up clarifying your thinking on a topic:

Perhaps many on this list are focussed on finding things, searching for something in the haystack.

I should say that I’m interested in metadata attributes so that I can create structures with that data; for example a visual map, where author and date become the X and Y coordinates, and the genre becomes the shape of each datum.

So metadata is very complementary to the features in DTPO. But let me repeat that this isn’t an argument for adding features to DTPO, but just an inquiry into how DTPO can support this type of work.

Best, Charles

My half-baked effort to mark related topics for quicker future reference, mostly in reaction to noticing a lot of uncorrelated repetition that I’d prefer be kept together, e.g. under a FAQ subforum.