Having engaged several Fortune 500 companies with projects to develop Semantic Technology solutions we have identified several consistent requirements that have become the foundation for successful deployments of Semantic Technologies.
The overarching pattern that we see in these companies can best be described as real time entity tracking in order to perform real time business analytics. Typical entities are students, telephone customers, credit cards or insurance policies.
We identified and built out four components as the basis of our Semantic Technology Projects. Component one is an ETL system that takes data from various input streams and transforms the data into events, encoded as RDF triples, that go into a publish subscribe queue. To facilitate this we created a number of plugins for the open source ETL tool Talend to provide an R2RML mapping from data into triples. The second component is a forward chaining/backward chaining rule system that takes events out of the queue and combines it with the already existing knowledge about a particular entity and generates new knowledge. For some applications we see more than 10,000 triples per entity. Rules need to be able to deal with a new event in a fraction of a second. The third component is a machine learning component that is trained to generate predictions based on the features of a particular entity (for example: what is the customer going to call about when calling the call center). These predictions are again coded as individual triples. Finally, the fourth component is a reporting system that allows us to do real time analysis over all existing entities.
|Copyright © 2016 Franz Inc., All Rights Reserved | Privacy Statement||