I attended Joseph Busch's excellent talk on taxonomies during the last presentation session of the day (had to miss the IA slam, that sounded like a lot of fun as well). One of his key points was that content producers are going to have to create metadata, because it is not possible to hire enough librarians to do it all. Information architects can provide simple rules to the content creators for categorizing information. This led to a great dialogue about how to incentivize content producers so that they create metadata.
Here is a URL for his presentation (in PowerPoint): http://www.asis.org/Chapters/asispvc/feb_10_2004/ASISTPVC021004.ppt
Please share your thoughts!
Posted by jeffrey.linwood at February 29, 2004 06:53 AMHere are my notes from the preso. I have tried to capture Joseph's not-on-the-slide comments (forgive me for my mistakes -- it is 2 in the morning):
First topic: The Dublin Core is Descriptive Catalogues
Myth 1 - Taxonomies are monolithic hierarchies
- Facet Categories (FC) are mututally exclusive
- Each FC has a controlled vocabulary (CV) within it
- It is easier to maintain Faceted Taxonomies than pre-specifying all the possible interactions between the facets in a monolithic taxonomy (4 facets x 10 items per is 44 items to maintain but with 10^4 interactions)
- Joe doesn't subscribe to the notion of formalized taxonomy (non-polyhierarchical or a concept can only appear once in a taxonomy)
Myth 2: People retreive content by topical subject
- Products & Services and Organization are the most important taxonomy facets, but they are 't harder than one might think to arrive at (most corporations donhave a definitive list -- differences between marketing and support, versioning, etc.)
- Content Types are also very important for retreival
- If these facets are well-articulated, one might not need much the in the Subject field.
Myth 3: No one else can do the content indexing other than librarians
- There are not enough librarians in the world to index all the content out there
- We have to make indexing simple and easy so that errors and frustrations and time-suck are minimized
- We need specific guidelines to aid indexing to make it faster and less open-ended
- The topic of contextual metadata and content-parsing-based suggested metadata came up here - more in the Tools discussion on Sunday
Myth 4: All Search Engines can retrieve is a list
- This is a problem of information space design within a website
- Show the extent of the available content
- Filter content into reviewable groups
- Find more like this
- Show latest
- Feature selected
- Siderean demos are relevant here
Wrap-up:
Areas for growth and applicability:
Outside the website:
- Contribute content to larger portals
- Have to make it findable from within the new and larger context
- Map to other taxonomies
- Integrate into web search engines
- How can one collection become part of another
- XML + RSS, etc.
Overall I thought it was a very good talk. It is great to have someone speak so clearly and calmly on this topic and help move us forward to the point where we all will have the option of using Taxonmy tools to build our websites (when they are appropriate).
Posted by: Brett Lider at February 29, 2004 07:42 AM