Accelerated content analytics based on a hierarchical data-flow-graph representation

1. A computer-implemented method for partitioning an Unstructured Information Management Architecture (UIMA)-Processing Engine ARchive (PEAR) file into software-based and hardware-accelerated components comprising:

creating a data-flow graph representation of the UIMA-PEAR file;

evaluating the software and hardware complexity of each component of the data-flow graph representation;

flattening hierarchies of the data-flow graph representation; and

selecting the components to be hardware accelerated from the flattened hierarchies of the data-flow graph representation based on data dependencies of data types produced and consumed by each component of the flattened data-flow graph, wherein selecting the components to be hardware accelerated comprises:

removing off-loadable software components from analysis engines, wherein the highest level of hierarchy of the data-flow graph representation is composed of the analysis engines and wherein each analysis engine is a composition of a set of library-based or user-defined software components;

removing analysis engines that have no more off-loadable software components:

merging remaining analysis engines to form new hardware-based components while preserving the data dependencies of the data types, wherein at least some of the merged analysis engines are from different levels of hierarchy of the data-flow graph representation of the UIMA-PEAR file; and

adding the new hardware-based components to be implemented in hardware.

2. The method of claim 1 , wherein the data-flow graph representation comprises:

a plurality of nodes, each node representing a component that either produces data or consumes data; and

a plurality of edges, each edge representing a data type and connecting a component that produces the data type with a component that consumes the data type.

3. The method of claim 2 , wherein evaluating software and hardware complexity of each component comprises:

determining a software execution time spent on each component; and

determining resources needed to implement each component.

4. The method of claim 3 , further comprising:

mapping software-based components into resources of a hardware accelerator.

5. The method of claim 3 , further comprising:

mapping multiple software-based components into resources of a hardware accelerator to form a data-flow pipeline.

6. A computer program product for partitioning an Unstructured information Management Architecture (UIMA)-Processing Engine ARchive (PEAR) file into software-based and hardware-accelerated components, the computer program product comprising a non-transitory computer readable storage having program instructions embodied therewith, the program instructions executable by a computer, to cause the computer to perform a method comprising:

creating a data-flow graph representation of the UIMA-PEAR-file;

evaluating the software and hardware complexity of each component of the data-flow graph representation;

flattening hierarchies of the data-flow graph representation; and

selecting the components to be hardware accelerated from the flattened hierarchies of the data-flow graph representation based on data dependencies of data types produced and consumed by each component of the flattened data-flow graph, wherein selecting the components to be hardware accelerated comprises:

removing off-loadable software components from analysis engines, wherein the highest level of hierarchy of the data-flow graph representation is composed of the analysis engines and wherein each analysis engine is a composition of a set of library-based or user-defined software components;

removing analysis engines that have no more off-loadable software components;

merging remaining analysis engines to form new hardware-based components while preserving the data dependencies of the data types, wherein at least some of the merged analysis engines are from different levels of hierarchy of the data-flow graph representation of the UIMA-PEAR file; and

adding the new hardware-based components to be implemented in hardware.

7. The computer program product of claim 6 , wherein the data-flow graph representation comprises:

a plurality of nodes, each node representing a component that either produces data or consumes data; and

a plurality of edges, each edge representing a data type and connecting a component that produces the data type with a component that consumes the data type.

8. The computer program product of claim 7 , wherein evaluating software and hardware complexity of each component comprises:

determining a software execution time spent on each component; and

determining resources needed to implement each component.

9. The computer program product of claim 8 , further comprising program instructions for:

mapping software-based components into resources of a hardware accelerator.

10. The computer program product of claim 8 , further comprising program instructions for:

mapping multiple software-based components into resources of a hardware accelerator to form a data-flow pipeline.

11. A system for partitioning an Unstructured Information Management Architecture (UIMA)-Processing Engine ARchive (PEAR) file into software-based and hardware-accelerated components, the system comprising a processor, memory accessible by the processor, and computer program instructions stored in the memory and executable by the processor to perform:

creating a data-flow graph representation of the UIMA-PEAR-file;

flattening hierarchies of the data-flow graph representation; and

selecting the components to be hardware accelerated from the flattened hierarchies of the data-flow graph representation based on data dependencies of data types produced and consumed by each component of the flattened data-flow graph, wherein selecting the components to be hardware accelerated comprises:

removing off-loadable software components from analysis engines, wherein the highest level of hierarchy of the data-flow graph representation is composed of the analysis engines and wherein each analysis engine is a composition of a set of library-based or user-defined software components;

removing analysis engines that have no more off-loadable software components;

merging remaining analysis engines to form new hardware-based components while preserving the data dependencies of the data types, wherein at least some of the merged analysis engines are from different levels of hierarchy of the data-flow graph representation of the UIMA-PEAR file; and

adding the new hardware-based components to be implemented in hardware.

12. The system of claim 11 , wherein the data-flow graph representation comprises:

a plurality of nodes, each node representing a component that either produces data or consumes data; and

a plurality of edges, each edge representing a data type and connecting a component that produces the data type with a component that consumes the data type.

13. The system of claim 12 , wherein evaluating software and hardware complexity of each component comprises:

determining a software execution time spent on each component; and

determining resources needed to implement each component.

14. The system of claim 13 , further comprising computer program instructions for:

mapping software-based components into resources of a hardware accelerator.

15. The system of claim 13 , further comprising computer program instructions for:

mapping multiple software-based components Into resources of a hardware accelerator to form a data-flow pipeline.