Ontologies are forms of classifications for terminology. In other words, the idea is as follows:
This is just one way (one ontology) of terms that are used to describe vehicle. There are many other views of Vehicle (axis) on which to pivot this data set, depending on the classification strategy you want to use (enter: Taxonomy).
Business users often fight over what a term means, why? Because they apply the term in a different context (along a different ontology or categorization). The problem is: they are both right at the same time, even with different definitions of the Metadata.
What does this mean to my Master Data Projects?
Well if you're focused on creating a successful BI deployment, and creating value in the business, then you MUST allow these multiple ontologies to exist. However, good governance practices will state that you SHOULD manage them through a taxonomy (a classification of groups of ontologies). I guess you can think of it as a CUBE for metadata with the elements in the cells, and along the axis points. (Someone should visualize metadata this way, it would be really interesting).
In other words, you MUST be capable of creating "master" metadata classifications IF you are to be successful at creating, understanding, and deploying Master Data Management solutions across your company or organization.
Unfortunately there are hundreds of "spread out" tool sets that manage "ontologies" and sets of meta-data, but don't come close to hooking it in to a BI/EDW or even Master Data tool.
How does this impact my project?
If you have embarked on MDM project, or something akin to that, and you have NOT looked at your metadata or ontological layout, you are already behind the 8-ball. You probably are experiencing cost overrun, missed requirements, non-met time frames for delivery. You probably have a disgruntled staff, and disallusioned business users. If on the other hand you have a successful project, I'd like to hear how you did it without Master Metadata.
It's high time to put together the Metadata needed to apply, understand, and decipher Master Data at the business level. Oh - yea, and one more shocker (probably not news to you)... but if you're ever going to get value out of Unstructured data, or you plan on mining unstructured data, then you will ABSOLUTELY need the ontological breakdown of the different languages that your documents are written in. It is a MUST to tying any of that information into a structural world where the context can SHIFT at any given time.
I hope this is helpful, I'd like to hear your experiences on the subject.
By the way, we've just launched a library of technical documents (free) to which we will be expanding these subjects at http://www.DataVaultInstitute.com
Posted July 24, 2009 4:27 AM
Permalink | 1 Comment |