Figure 1. Current Semanticle support for Visualisation Pipe based Delivery Services
The Visualization Pipe is a powerful architectural pattern supporting versatile multi format, multi target content delivery which can be modelled and which maximises requirement specification re-usability.
It separates the following logically distinct aspects of content generation:
Figure 1 above illustrates this with current Semanticle support for each of the phases shown in green.
Querying embraces selection and filtering which are sometimes shown as seperate phases of the Visualisation Pipe. Selection covers match criteria, in the case of the Semantic Web these may be thought of as similtaneous equations expressed in triples optionally correlated with discriptive logic. Filtering supports further non-semantic extraction criteria including:
- uniqueness filters - removes duplicate results (implemented).
- evaluated filters - typical examples might include geospatial proximity functions for local search, financial or time functions.
- work-flow or contextual filters - for example sorting criteria or pagination.
Clustering is about what we are interested in from what what is selected and these are combined to form discreet results. As noted below the left hand side rules are expressed as clusters which are triples.
Rendering concerns how the results set is to be presented. examples might be as a list, a table, a graph or within a map.
Delivery is about where to ship the results. Typical examples are files, interactive outputs like HTML, and stores. Results might might shipped to multiple destinations. Certain destinations will have requirements about what they can accept. For example a triple store will expect to receive results expressed as triples.
Seperating things which are logically distinct is key to expanding re-usability. A query which can be expressed distictly from the data to be extracted and how it is to be delivered may be re-used for deferent data requirements and delivery to different targets.
As it happens the re-use advantages don't end there.
Expressed in these terms, queries are identical to the right hand sides of rules and the clustering the same as the left hand side of rules. Semanticle can store queries for use within rules and re-use the same software components to solve either.
Moreover, our delivery engine is capable of making multiple selections from a single query, rendering these in various ways and delivering the results to multiple targets - all from executing a single query and without any additional application programming.
Query results delivered to files create export outputs, to stores cause results to be instantiated therein and to outputs cause these to be presented as varous forms of media.
There remains much still to do in terms of extending Semanticle's current Visualization Pipe capabilities to support additional clustering and rendering types.