What Metadata?

By Michael Wells posted 10-28-2010 16:16


I have a couple of philosophies I've used or seen in different departments with regards to metadata.  Our ECM product uses "templates" which are sets of "fields" which contain metadata.  (I don't know how many other products use similar models - I can only speak about what I know.)

Each Template has a set of fields.  Each of the fields has a "Type" (text, date, integer, etc.), a length, and maybe a default or some constraining expressions.  Each field only exists in one Template.

So, in order to meet different needs I use different models.


In some departments we use one template (set of fields) for each type of document.  These Templates are used at capture and again for searching and retrieving the doucments.  This is the simplest and least confusing layout.

On the other side of things, I've seen some metadata designs which use one large template for ALL documents, with the understanding that not ALL documents will have ALL fields filled in.  Sometimes these documents have metadata linking across all documents, sometimes they don't.  (For instance, every document will have an account number no matter what type.)  This does allow you to share key pieces of metadata across document types (Like Name or SSN, for instance) but this model has a number of drawbacks.  These shared templates can be very large and cumbersome for scan operators and search users.  The shared model also invites confusion - "Why aren't all the fields filled in???"  This model also limits administrators in the field constraints they can use, for usually there are very few "key" fields which are always filled in.

So, what do we do?  In many cases now I use a hybrid approach, were templates are shared only among similar document types.  If it is important for the "Address" field to be searchable across 3 different document types, then those document types all go on one template.


How do you design your metadata templates?


One side note - the latest update of our product introduces a more flexible metadata model, where fields exist independantly of templates and templates are non-exclusive sets of those fields.  This lets you build templates for whatever purpose you wish from any of those fields.  Mix, match, and blend the fields with different document types while keeping the templates as small as possible.  I look forward to upgrading soon!

#metadata #fields #ScanningandCapture #template