Common Ontology for Personal Data Interoperability – (Part 2) The What and How

Common Ontology for Personal Data Interoperability, Part 2

Wednesday 5G Convener: Julian Ranger

Notes-taker(s): Julian Ranger

Tags for the session - technology discussed/ideas considered:

Discussion notes, key understandings, outstanding questions, observations, and, if appropriate to this discussion: action items, next steps:   Day 2

1. In Part 1 of this session the previous day it was agreed should have a further Part 2 session at this IIW to explore what it would take to create a single ontology (or minimum set) - the What & How.  Points to note:
   1. Look at use cases and solve for those first
   2. Look at other ontology standardisation examples and why they failed or succeeded
   3. Discuss "Data to Value" maps
2. The first question was what is the economic model for companies to come together to create a common ontology?  (Usually companies do as secret sauce.)  Two answers
   1. Philosophical alignment - marketing point
   2. Businesses use data faster with less work <- the main point
3. Talking about data elements only for the ontology - fields and values.
4. Marc Davis introduced the concept of Data to Value map to work out which areas should be worked on first for a common ontology
   1. Data fields follow a power law, i.e. some fields are used a lot, and most others used very little
   2. If you then map in a classic 2 by 2 matrix with Y axis being data frequency across ontologies to be normalised, and X axis being value of data element to the business on Y axis, you can plot where each data element / element sits.  Clearly those in top right quadrant (high data frequency & high value) are the ones to normalise to a common ontology first.
5. Question of what are the first steps on the path of getting to the 'Golden Goal' of a single normalised ontology (or minimum set)?
   1. Create a small manageable group in first instance to start the process of creating an open ontology
   2. Wikipedia style site to manage process
   3. Suggested follow Microformats process
   4. 1st publish examples of normalised data with traceability
   5. Review commonality & discuss
   6. Agree version to go forward
      1.  (Note: keep a & b for history of why have end result)
   7. Agreed will include an example representation to show usage - will use JSON
   8. Probable graph structure - unique strings (though discussed option of non unique if contextual) - may map to GUIDs (though discussion on whether necessary at this level)
6. Agreed that following group of 4 would create strawman of 5 above, based on an initial example of a normalised social media post to be provided by digi.me, with others to contribute if interested (contact group via julian@digi.me):
   1. Julian Ranger
   2. Drummond Reed
   3. Kevin Marks
   4. Marc Davis
7. Final discussion on economic model for contribution:
   1. Open, with contributor licence - look at OWFA as an example
   2. Or could create a CIC (can't then sell), but contributors have shares if any value accrues