Using classified text, deep learning algorithms and blockchain to identify risk in low-frequency, high value situations, and provide early warning
Deep learning is used to identify specific, potential risks to an enterprise while such risks are still internal electronic communications. The combination of Deep Learning and blockchain technologies is a system for overcoming the problem of “small training sets” for highly adverse situations. Each enterprise's data is secure; is not revealed to any other enterprise and yet is being aggregated using blockchain technology into a training set that is provably viable for building a Deep Learning model which is specific to a given adverse situation. When deployed, the Deep Learning model may provide an early warning alert to an enterprise's corporate counsel (or leaders) of a potential adverse situation the enterprise would like to know about in time to conduct an internal investigation in order to prevent or avoid the risk.
1. A method of using classified text augmented with a blockchain and deep learning to identify risk comprising:
obtaining one or more training datasets comprising of textual data corresponding to one or more adverse risk situations;
aggregating a plurality training data into a blockchain for one of said one or more training datasets, wherein each one of said plurality of training data consists of a specific adverse risk situation of said one or more adverse risk situations provided by one of a plurality of enterprises, wherein said training data comprises a number string;
training one or more deep learning algorithms using said one or more training datasets;
extracting an internal electronic communication of an enterprise;
applying said one or more deep learning algorithms to said internal electronic communication to identify and report any one of said one or more adverse risk situations;
determining if said identified one of said one or more adverse risk situations is a false positive or a true positive;
re-training said one or more deep learning algorithms if said identified one of said one or more adverse risk situations is a false positive; and
saving said internal electronic communication in a true positive database if said identified one of said one or more adverse risk situations is a true positive.
2. The method of claim 1 , wherein said one or more deep learning algorithms is a framework for natural language processing of text.
3. The method of claim 1 , wherein said one or more deep learning algorithms is a recurrent neural network with a multiplicity of layers and various features, including but not limited to long short-term memory or gated recurrent units.
4. The method of claim 1 , wherein the one or more deep learning algorithms have been trained with different classifiers using previously classified data sourced and provided by a subject matter expert to become models for specific adverse risk situations.
5. The method of claim 1 , wherein each one of said one or more deep learning algorithms has also been trained with one or more datasets unrelated to the adverse risk situations.
6. The method of claim 1 , wherein said internal electronic communication is indexed.
7. The method of claim 6 , wherein each one of said one or more deep learning algorithms scores the data for accuracy with the deep learning model classification of the data.
8. The method of claim 1 , wherein each of said number string comprises a small set of Classified Examples.
9. The method of claim 8 , wherein said Classified Examples comprises one or more specific risk situations identified in internal documents by an enterprise.
10. A method of using classified text augmented with a blockchain and deep learning to identify risk comprising:
obtaining one or more training datasets comprising of textual data corresponding to one or more adverse risk situations;
aggregating a plurality training data into a blockchain for one of said one or more training datasets, wherein each one of said plurality of training data consists of a specific adverse risk situation of said one or more adverse risk situations provided by one of a plurality of enterprises, wherein said training data comprises a number string;
training one or more deep learning algorithms using said one or more training datasets;
extracting one or more internal electronic communications of an enterprise; and
applying said one or more deep learning algorithms to said one or more internal electronic communications to identify any one of said one or more adverse risk situations.
11. The method of claim 10 , further comprising retraining said one or more deep learning algorithms if said identified one of said one or more adverse risk situations is a false positive.
12. The method of claim 11 , wherein determination of said false positive is performed by a user through a graphical user interface.
13. The method of claim 10 , wherein said number string comprises a set of Classified Examples.
14. The method of claim 13 , wherein said Classified Examples comprises one or more specific risk situations identified in internal documents by an enterprise.
15. A computer program product for using classified text augmented with a blockchain and deep learning algorithms to identify risk, the computer program product comprising non-transitory computer-readable media encoded with instructions for execution by a processor to:
obtain one or more training datasets comprising of textual data corresponding to one or more adverse risk situations;
aggregate a plurality training data into a blockchain for one of said one or more training datasets, wherein each one of said plurality of training data consists of a specific adverse risk situation of said one or more adverse risk situations provided by one of a plurality of enterprises, wherein said training data comprises a number string;
train one or more deep learning algorithms using said one or more training datasets;
extract one or more internal electronic communications of an enterprise; and
apply said one or more deep learning algorithms to said one or more internal electronic communications to identify any one of said one or more adverse risk situations.
16. The computer program product of claim 15 , wherein said number string comprises a set of Classified Examples.
17. The computer program product of claim 16 , wherein said Classified Examples comprises one or more specific risk situations identified in internal documents by an enterprise.