Bottom-up approach to taxonomy development

In my previous post, I brought up a topic on the implementation challenges of taxonomy and suggested a few points on overcoming pitfalls for multi-faceted taxonomy implementation. This time round, my reflection is based on ground 0, where there is no corporate taxonomy design to start with.

This idea requires incremental and continuous investment and its not a short project period affair. Like the Chinese saying goes, "Cast a long line to catch a bigger fish".

Step 1 - Implement a simple DMS
If there aren't any Document Management System (DMS) to start with, work with your IT department to invest and setup a really affordable one. There are good open source CMS solutions in the market. Get one that supports tagging and have the ability to export these tags at the document or content chunk level in a standard format such as XML.

Step 2 - Document migration and housekeeping
Get everyone to migrate their folders and document to the DMS and perform housekeeping at the same time. At the department level, tell them to create, organise and label folders as they like. But give these guidelines:

• each folder should not contain more than 15 sub-folders or documents (human beings are generally comfortable with 8 - 15 items in a cluster)
  
• the number of folder levels should not exceed 7 (this prevents document from gaining too much depth making it hard for others to find or discover)

Step 3 - Describe documents
Create 2 manually entered metadata fields - Title (of document) and My tags (comma delimited keywords describing document). New documents or document versions captured in the system must be profiled with these 2 fields. The Title can be defaulted to the filename without the file extension. With My tags, the process of self-describing documents begins. Make the tag clouds of individual contribution available at all times so that they can be mindful to reuse their previous labels. Encourage the tagging of important or useful documents that have been migrated.

Step 4 - Create relationships
If the current CMS doesn't support this step, export the document metadata (including title, my tags and other automatically captured data such as date modified, author and etc). At this point, take each document as a central focus with metadata wrapped around it. Consider 3 documents with the following My tags and title

1. Title (Lifestyle of a diabetic), My tags (lifestyle, diabetic, diabetes, food, exercise)
2. Title (Medication for diabetes), My tags(diabetes, medicine, tablets, insulin)
3. Title (Diabetes resource), My tags (diabetes, blogs, advice)

The common tag of the above example is diabetes. We can now draw relationships using diabetes to make documents on lifestyle, diabetic, food, exercise, medicine, tablets, insulin, blogs and advice to appear upfront. And in this, we are helping the diabetic make discovery of what he may want to know. When we link food to another set of documents with food in My tags, we form another cluster of information around food at the next level (which is now becoming less relevant to diabetes). The title can be split into words and each word to simulate a tag, useful especially for older documents that are not tagged.

These relationships form a dynamic taxonomy with many facets depending on the number of labels that matches your search term(s) (food, diabetes, lifestyle or etc).

Step 5 - Making the search results manageable
We may end up with huge number of document entries for diabetes and the levels below it. This is where the tags and other metadata can be helpful. We can filter the documents by tag labels, author, date modified and etc to find the latest and relevant documents.


Benefits

• Surfaces information around a topic very early in a tree. Unlike traditional navigation systems that help you find the exact document by traversing trees, all documents relating to the domain of interest is presented upfront and gets less relevant as you traverse down the tree. So you traverse to make sure you don't miss out anything important and you are unlikely to go deep.
• Less likely to run into the problem of missing a document located in another branch of a conventional taxonomy.
• More meaning and context to diabetes are added as more documents are tagged with diabetes. Within an organisation, this means the knowledge base around a particular domain.
• You choose what you want/don't want to know about the topic by filtering, and get exposed to what you didn't thought could be of use in the first place.

Documents are made miscellaneous in this manner and not limited by the conventional view of a taxonomy. The next powerful thing, instead of providing comments and rating after reading a document, social tag it.