Cognitive iterative minimization of personally identifiable information in electronic documents
Mechanisms are provided to minimize personally identifiable information (PII) in an electronic document. An iterative personally identifiable information minimization (IPIIM) engine receives an electronic document comprising natural language content having a mention of a protected entity and obfuscates the mention of the protected entity to thereby generate a minimized natural language content. A question answering system processes the minimized natural language content to generate a listing of candidate answers and corresponding confidence scores and the IPIIM engine determines whether or not the minimized natural language content is sufficiently obfuscated based on the listing of candidate answers and corresponding confidence scores. In response to determining that the minimized natural language content is sufficiently obfuscated, the minimized natural language content is provided for processing by a requestor computing device.
1. A method, in a data processing system comprising at least one processor and at least one memory, the at least one memory comprising instructions executed by the at least one processor to cause the at least one processor to minimize personally identifiable information (PII) in an electronic document, the method comprising:
receiving, by an iterative personally identifiable information minimization (IPIIM) engine executing in the data processing system, the electronic document comprising natural language content having a mention of a protected entity;
applying, by the IPIIM engine, natural language processing and analytic analysis on the natural language content to obfuscate the mention of the protected entity and thereby generate a first minimized natural language content, wherein the obfuscated mention of the protected entity is indicated to be a focus in the first minimized natural language content;
automatically processing, by a question answering computing system, the first minimized natural language content, as input to the question answering computing system, to generate a first listing of one or more candidate answers and corresponding confidence scores, wherein the one or more candidate answers specify predictions of the protected entity, in the mention of the protected entity, prior to the obfuscation of the mention of the protected entity, and wherein the question answering computing system comprises artificial intelligence computer logic configured to predict an answer corresponding to the focus of the input to the question answering computing system;
determining, by the IPIIM engine, whether or not the first minimized natural language content is sufficiently obfuscated based on the first listing of one or more candidate answers and corresponding confidence scores, wherein the first minimized natural language content is determined to not be sufficiently obfuscated in response to a correct identity of the protected entity being present in a candidate answer, of the one or more candidate answers, which has a corresponding confidence score equal to or above a predetermined threshold confidence score; and
in response to determining that the first minimized natural language content is sufficiently obfuscated, providing the first minimized natural language content as a final minimized natural language content of the electronic document for processing by a requestor computing device.
2. The method of claim 1 , further comprising, in response to determining that the first minimized natural language content is not sufficiently obfuscated, iteratively performing additional minimizations of elements present in the natural language content to increase obfuscation/minimization of elements in the natural language content until a sufficient level of obfuscation is achieved to cause the question answering system to not include the correct identity of the protected entity in the first listing of one or more candidate answers with a corresponding confidence score equal to or higher than a predetermined threshold confidence score.
3. The method of claim 1 , further comprising, in response to determining that the first minimized natural language content is not sufficiently obfuscated:
automatically performing obfuscation of one or more additional elements of the first minimized natural language content to generate second minimized natural language content, wherein the obfuscation of the one or more additional elements comprises designating each of the one or more additional elements as foci for answer generation by the question answering computing system;
processing, by the question answering computing system, the second minimized natural language content to generate a second listing of one or more candidate answers and corresponding confidence scores;
determining, by the IPIIM engine, whether or not the second minimized natural language content is sufficiently obfuscated based on the second listing of one or more candidate answers and corresponding confidence scores; and
in response to determining that the second minimized natural language content is sufficiently obfuscated, providing the second minimized natural language content as a final minimized natural language content of the electronic document for processing by the requestor computing device.
4. The method of claim 1 , further comprising receiving a listing of protected entities whose identities are to be protected in content of electronic documents, wherein applying, by the IPIIM engine, natural language processing and analytic analysis on the natural language content to obfuscate the mention of the protected entity and thereby generate the first minimized natural language content comprises identifying mentions in the natural language content that match an entity in the listing of protected entities.
5. The method of claim 1 , wherein processing, by the question answering computing system, the first minimized natural language content to generate the first listing of one or more candidate answers and corresponding confidence scores comprises:
generating a first input question based on the first minimized natural language content; and
processing the first input question by the question answering computing system at least by applying one or more queries to a reference document corpus to generate the first listing of one or more candidate answers.
6. The method of claim 1 , wherein the requestor computing device is a cognitive computing system that performs operations on the final minimized natural language content without exposing an identity of the protected entity.
7. The method of claim 6 , wherein the operation performed by the cognitive computing system includes one of an application development operation for developing an application or a debugging operation for debugging an application that operates on the natural language content of the electronic document.
8. The method of claim 1 , wherein the requestor computing device comprises a cognitive decision support system that performs operations for generating a decision support output, and wherein the requestor computing device performs, in an application development environment of the requestor computing device, an application development operation for developing an application of the cognitive decision support system based on processing of the final minimized natural language content of the electronic document by the cognitive decision support system, without exposing an identity of the protected entity to the application development environment.
9. A computer program product comprising a computer readable storage medium having a computer readable program stored therein, wherein the computer readable program, when executed by a data processing system, causes the data processing system to:
receive, by an iterative personally identifiable information minimization (IPIIM) engine executing in the data processing system, an electronic document comprising natural language content having a mention of a protected entity;
apply, by the IPIIM engine, natural language processing and analytic analysis on the natural language content to obfuscate the mention of the protected entity and thereby generate a first minimized natural language content, wherein the obfuscated mention of the protected entity is indicated to be a focus in the first minimized natural language content;
automatically process, by a question answering computing system, the first minimized natural language content, as input to the question answering computing system, to generate a first listing of one or more candidate answers and corresponding confidence scores, wherein the one or more candidate answers specify predictions of the protected entity, in the mention of the protected entity, prior to the obfuscation of the mention of the protected entity, and wherein the question answering computing system comprises artificial intelligence computer logic configured to predict an answer corresponding to the focus of the input to the question answering computing system;
determine, by the IPIIM engine, whether or not the first minimized natural language content is sufficiently obfuscated based on the first listing of one or more candidate answers and corresponding confidence scores, wherein the first minimized natural language content is determined to not be sufficiently obfuscated in response to a correct identity of the protected entity being present in a candidate answer, of the one or more candidate answers, which has a corresponding confidence score equal to or above a predetermined threshold confidence score; and
in response to determining that the first minimized natural language content is sufficiently obfuscated, provide the first minimized natural language content as a final minimized natural language content of the electronic document for processing by a requestor computing device.
10. The computer program product of claim 9 , wherein the computer readable program further causes the data processing system, in response to determining that the first minimized natural language content is not sufficiently obfuscated, to iteratively perform additional minimizations of elements present in the natural language content to increase obfuscation/minimization of elements in the natural language content until a sufficient level of obfuscation is achieved to cause the question answering computing system to not include the correct identity of the protected entity in the first listing of one or more candidate answers with a corresponding confidence score equal to or higher than a predetermined threshold confidence score.
11. The computer program product of claim 9 , wherein the computer readable program further causes the data processing system, in response to determining that the first minimized natural language content is not sufficiently obfuscated, to:
automatically perform obfuscation of one or more additional elements of the first minimized natural language content to generate second minimized natural language content, wherein the obfuscation of the one or more additional elements comprises designating each of the one or more additional elements as foci for answer generation by the question answering computing system;
process, by the question answering computing system, the second minimized natural language content to generate a second listing of one or more candidate answers and corresponding confidence scores;
determine, by the IPIIM engine, whether or not the second minimized natural language content is sufficiently obfuscated based on the second listing of one or more candidate answers and corresponding confidence scores; and
in response to determining that the second minimized natural language content is sufficiently obfuscated, provide the second minimized natural language content as a final minimized natural language content of the electronic document for processing by the requestor computing device.
12. The computer program product of claim 9 , wherein the computer readable program further causes the data processing system to receive a listing of protected entities whose identities are to be protected in content of electronic documents, wherein applying, by the IPIIM engine, natural language processing and analytic analysis on the natural language content to obfuscate the mention of the protected entity and thereby generate the first minimized natural language content comprises identifying mentions in the natural language content that match an entity in the listing of protected entities.
13. The computer program product of claim 9 , wherein processing, by the question answering computing system, the first minimized natural language content to generate the first listing of one or more candidate answers and corresponding confidence scores comprises:
generating a first input question based on the first minimized natural language content; and
processing the first input question by the question answering computing system at least by applying one or more queries to a reference document corpus to generate the first listing of one or more candidate answers.
14. The computer program product of claim 9 , wherein the requestor computing device is a cognitive computing system that performs operations on the final minimized natural language content without exposing an identity of the protected entity.
15. The computer program product of claim 14 , wherein the operation performed by the cognitive computing system includes one of an application development operation for developing an application or a debugging operation for debugging an application that operates on the natural language content of the electronic document.
16. The computer program product of claim 9 , wherein the requestor computing device comprises a cognitive decision support system that performs operations for generating a decision support output, and wherein the requestor computing device performs, in an application development environment of the requestor computing device, an application development operation for developing an application of the cognitive decision support system based on processing of the final natural language content of the minimized electronic document by the cognitive decision support system, without exposing an identity of the protected entity to the application development environment.
17. An apparatus comprising:
a processor; and
a memory coupled to the processor, wherein the memory comprises instructions which, when executed by the processor, cause the processor to:
receive, by an iterative personally identifiable information minimization (IPIIM) engine executing on the processor, an electronic document comprising natural language content having a mention of a protected entity;
apply, by the IPIIM engine, natural language processing and analytic analysis on the natural language content to obfuscate the mention of the protected entity and thereby generate a first minimized natural language content, wherein the obfuscated mention of the protected entity is indicated to be a focus in the first minimized natural language content;
automatically process, by a question answering computing system, the first minimized natural language content, as input to the question answering computing system, to generate a first listing of one or more candidate answers and corresponding confidence scores, wherein the one or more candidate answers specify predictions of the protected entity, in the mention of the protected entity, prior to the obfuscation of the mention of the protected entity, and wherein the question answering computing system comprises artificial intelligence computer logic configured to predict an answer corresponding to the focus of the input to the question answering computing system;
determine, by the IPIIM engine, whether or not the first minimized natural language content is sufficiently obfuscated based on the first listing of one or more candidate answers and corresponding confidence scores, wherein the first minimized natural language content is determined to not be sufficiently obfuscated in response to a correct identity of the protected entity being present in a candidate answer, of the one or more candidate answers, which has a corresponding confidence score equal to or above a predetermined threshold confidence score; and
in response to determining that the first minimized natural language content is sufficiently obfuscated, provide the first minimized natural language content as a final minimized natural language content of the electronic document for processing by a requestor computing device.
18. The apparatus of claim 17 , wherein the instructions further cause the processor, in response to determining that the first minimized natural language content is not sufficiently obfuscated, to iteratively perform additional minimizations of elements present in the natural language content to increase obfuscation/minimization of elements in the natural language content until a sufficient level of obfuscation is achieved to cause the question answering computing system to not include the correct identity of the protected entity in the first listing of one or more candidate answers with a corresponding confidence score equal to or higher than a predetermined threshold confidence score.
19. The apparatus of claim 17 , wherein the instructions further cause the processor to receive a listing of protected entities whose identities are to be protected in content of electronic documents, wherein applying, by the IPIIM engine, natural language processing and analytic analysis on the natural language content to obfuscate the mention of the protected entity and thereby generate the first minimized natural language content comprises identifying mentions in the natural language content that match an entity in the listing of protected entities.
20. The method of claim 1 , wherein the question answering computing system comprises a multi-focus factoid question answering system comprising artificial intelligence computer logic configured to predict entity identities for a plurality of foci in an input to the question answering computing system.