
What is 'Semantic Interoperability'?
'Semantic Interoperability' simply means that specific words have shared, exact meanings across different systems. The reason this is such an important issue in big data is that it's very difficult and time-consuming to process qualitative data if there is no semantic interoperability.
While numerical data - such as financial data - is relatively easy to process at scale with tools as simple as spreadsheets and basic mathematical formula, health data is completely different. Most health data is qualitative and non-numerical, and our health and medical treatment is based on the words that doctors write on our records. This means we have to continuously manage the relationships between specific clinical terms.
​
A basic example is shown below:

The above example is certainly complicated, but it is also the reason that we have great difficulty in sharing our health data between different sites. As our health records age, terminology continues to change, and these mapping issues become gradually worse over time.
Standards such as HL7 FHIR help greatly with the transport of health data, and - where adopted - OpenEHR will attempt to keep patient data consolidated and transferable. But neither of these approaches solve semantic interoperability at the level of clinical data itself.
Artificial Intelligence is no longer optional in health and life sciences
While there has been a concerted push to adopt formal and standardized clinical terminologies and vocabularies to enable semantic interoperability, this can only ever be a partial solution.
​
While it is absolutely necessary to agree at least a high level vocabulary or ontology, in reality our health data will always be shared between different devices and systems with different naming conventions for health descriptors. Also, medical knowledge keeps expanding, and this means new medical terms.
​
Most health organizations use human clinical coders to convert doctor's notes to formal clinical terminologies. However, this is painstakingly slow and expensive, and it will impossible to achieve broad semantic interoperability this way.
Artificial Intelligence is therefore required and necessary.
​