Generating event definitions based on spatial and relational relationships

1. A dynamic event detection architecture configured as one or more hypernodes and comprising:

(a) a hardware workflow engine that runs one or more workflows employed for processing input data by facilitating logic and state manipulation of the input data based on requirements of a specified task;

(b) a semantic database defined by an ontological model that is related to the specified task and which includes a meaning, rules, and data elements based on the ontological model;

(c) a hyperfragmenter system that processes the input data in real-time, in a workflow, to produce fragments of input data that are self-contained and discrete;

(d) a hyperasset file system joined with the semantic database such that both operate automatically, the hyperasset file system enabling the fragments of input data to be stored and retrieved based on specified criteria;

(e) a plurality of distributed experts that use an application program interface to facilitate review of the fragments of input data by the distributed experts at any point in the workflow and to provide additional information to the fragments of input data;

(f) a plurality of event definitions used by a first rule-based language engine to define events based on the fragments of input data in the workflow or based on relationships defined in the semantic database;

(g) a plurality of situation definitions used by a second rule-based engine to define complex situations to create situational rules; and

(h) a filtration system that applies the situational rules to the workflow to determines if an output of the workflow requires further analysis.

2. The dynamic event detection architecture of claim 1 , wherein the filtration system determines if the output should be exported to an external system for further analysis to determine if the output represents a normal activity or an abnormal activity.

3. The dynamic event detection architecture of claim 1 , wherein the filtration system determines whether the output represents a normal activity or an abnormal activity, the determination being based on a comparison of the output to historical training data input by the filtration system, the filtration system further refining definition criteria used to identify the abnormal activity by further analysis of related activities associated with events in the historical training data.

4. The dynamic event detection architecture of claim 1 , further comprising a plurality of path algorithms that can be triggered by the workflow and can be dependent on previously processed fragments of input data becoming available, the plurality of path algorithms being responsible for running and maintaining a state engine or statistical model system.

5. The dynamic event detection architecture of claim 1 , wherein the plurality of distributed experts comprises a plurality of humans.

6. The dynamic event detection architecture of claim 1 , wherein the plurality of distributed experts comprises a plurality of computing devices that execute artificial intelligence algorithms.

7. The dynamic event detection architecture of claim 1 , wherein the plurality of distributed experts perform at least one of the following functions in regard to a fragment of input data:

(a) determine if the fragment of input data is interesting or not interesting;

(b) determine if a fragment of input data that has been determined to be interesting includes at least one of a plurality of predefined specified elements; or

(c) adding additional details regarding a type of the at least one of the plurality of predefined specified elements that was determined to be included within the fragment of input data.

8. The dynamic event detection architecture of claim 7 , wherein a tagonomy is employed to specify tags applied to the plurality of predefined specified elements included in the fragments of input data.

9. The dynamic event detection architecture of claim 7 , wherein the plurality of distributed experts include computing devices that execute artificial intelligence algorithms that perform at least one of the functions and wherein the plurality of distributed experts further include humans who review results of the functions performed by the computing devices that execute the artificial intelligence algorithms to either confirm or modify the results.

10. The dynamic event detection architecture of claim 9 , wherein the plurality of event definitions and the plurality of situation definitions employ neural algorithms to refine definitions of events related to the specified task based on ground truth data, historical training data, and on input provided by the humans comprising the plurality of distributed experts who review the results of the functions performed.

11. The dynamic event detection architecture of claim 10 , wherein the filtration system uses the definitions of the events related to the specified task that have been refined as the definition criteria, and the fragments of input data that have been tagged and provided with the additional details to identify any abnormal activity evident in the fragments of input data that have been tagged.

12. The dynamic event detection architecture of claim 11 , wherein when identifying the abnormal activity, the filtration system applies rules included in the semantic database in analyzing the fragments of input data that have tagged.

13. The dynamic event detection architecture of claim 1 , wherein the filtration system employs the situational rules created in the plurality of situation definitions in producing the output of the workflow engine.