All models, whether simple or complex, take the form of a semantic graph, which uses semantic triples to encode information. These semantic graphs have varying complexity that is driven by the complexity of the scientific domain that is being modeled. For many domains, a simpler set of triples which represent key/value pairs may be sufficient to faithfully represent the data. In other cases, the need to represent data about multiple objects necessitates connections between these key/value pairs which creates a connected graph. When starting, the suggested approach is to choose the simplest model which can represent the information. As all graphs live on a continuum, it’s possible to enrich a simple model over time to capture more complex relationships. Thus, it’s not really appropriate to say a Tabular or Connected Model approach is an either-or proposition; they are both semantic graph models of varying complexity.

​

• Tabular Model: For domains where there is a single business object which is being measured and all of the measurements are directly related to this object, a tabular approach can be used. The tabular simplifies modeling by capturing just the individual measurements and associated measurement units and attaches them to a single entity being measured using a leaf node data pattern and simple aggregation patterns.

  • Another way to think of tabular models is as a set of key/value pairs, where the keys can be clearly distinguished without ambiguity and no references are required between keys (they can stand alone and accurately describe the data without relating to anything else.) Note that the context of the key/value pairs is represented only by the ADF file they are defined within; there may be no central node which connects the different keys. Tabular models are quick to assemble but lack the flexibility of full graph models. Nevertheless, tabular models, if constructed correctly, can be enhanced into full graph models as desired; many domains can be completely modeled using a tabular approach and no further expansion into a full graph is necessary.

​​

• ​Graph Model: For domains where there are multiple business objects being modeled or there are complex relationships such that it is insufficient to capture just the simple outputs of a single entity, a connected graph model richly models the different relationships in a complete, connected graph. An example of a connected graph model is the Allotrope LC-UV ADM.

​​