Semantic analyzer for training a policy engine
This disclosure describes systems, methods, and apparatus that monitor any manifestation of an idea, such as typed, written, or verbal message or document creation (e.g., while a user types an email or instant message, or makes a phone call) and analyze the manifestation in real-time to extract a sentiment and based on this sentiment, determine if the idea(s) manifested in the message, document, or other medium poses a risk of violating compliance, policy, or law.
1. A method for enhancing a computerized speed of precluding messages that pose a compliance, policy, or legal violation risk from being recorded in non-volatile memory, the method comprising:
parsing a first document into words and symbols via a parsing module, sequences of words, or words and symbols, being identified as ideas;
passing the first document to a semantic analyzer configured to:
access a policy model stored in a policy engine database;
access a violation threshold stored in the policy engine database;
assign a score to each of the ideas using the policy model;
determine that a total of scores for the first document surpasses the violation threshold;
disable save functionality of a device creating the first document, the save functionality configured to save the first document to non-volatile memory; and
provide the first document or a screenshot of the first document to an analytics dashboard configured to present the first document to a policy reviewer;
receiving a response from the policy reviewer indicating the policy reviewer's assessment of risk of the total of the scores for the first document;
storing the response in the policy engine database;
training the policy model, at a training module, using machine learning that uses the response as one of its inputs, to form an enhanced policy model; and
accessing the enhanced policy model and assigning scores to ideas parsed from a second document using the enhanced policy model.
2. The method of claim 1 , wherein the parsing module is part of the semantic analyzer.
3. The method of claim 1 , wherein the first document is received in real time as it is created.
4. The method of claim 1 , wherein the first document is received via import after document creation.
5. The method of claim 1 , wherein the semantic analyzer further provides the total of scores for the first document to the analytics dashboard.
6. The method of claim 1 , wherein the semantic analyzer requests the screenshot from a device creating the first document in response to determining that the total of scores for the first document surpasses the violation threshold.
7. The method of claim 1 , wherein the analytics dashboard is further configured to provide the first document to a second policy reviewer, the method further comprising:
receiving a second response from the second policy reviewer indicating the second policy reviewer's assessment of risk of the total of the scores for the first document;
storing the second response in the policy engine database; and
training the policy model, at the training module, using the machine learning that uses the second response as a second of its inputs, to form the enhanced policy model.
8. A system for enhancing a computerized speed of precluding messages that pose a compliance, policy, or legal violation risk from being recorded in non-volatile memory during document creation, the system comprising:
a parsing module configured to receive a document in real time or via import and parsing the document into words and symbols sequences of words, or words and symbols, being identified as ideas;
a policy engine database;
an analytics dashboard configured to present the document, or a screenshot of the document, to a policy reviewer, and to receive a response from the policy reviewer indicating the policy reviewer's assessment of risk of a total of scores assigned to the document;
a semantic analyzer configured to:
receive the sequences of words, or words and symbols;
access a policy model stored in the policy engine database;
access a violation threshold stored in the policy engine database;
assign a score to each of the ideas using the policy model;
determine that the total of the scores assigned to the document surpasses the violation threshold;
disable save functionality of a device creating the document, the save functionality configured to save the document to non-volatile memory; and
provide the document or the screenshot of the document to the analytics dashboard;
a training module configured to train the policy model using the response from the policy reviewer as one of its inputs, to form an enhanced policy model.
9. The system of claim 8 , wherein the parsing module is part of the semantic analyzer.
10. The system of claim 8 , wherein the semantic analyzer is further configured to provide the total of the scores assigned to the document to the analytics dashboard.
11. The system of claim 8 , wherein the semantic analyzer is further configured to request the screenshot from a device creating the document in response to determining that the total of scores assigned to the document surpasses the violation threshold.
12. The system of claim 8 , wherein the analytics dashboard is further configured to:
provide the document to a second policy reviewer;
receive a second response from the second policy reviewer indicating the second policy reviewer's assessment of risk of the total of the scores assigned to the document;
store the second response in the policy engine database; and
wherein the training module is further configured to train the policy model, using the second response as a second of its inputs, to form the enhanced policy model.
13. A non-transitory, tangible computer readable storage medium, encoded with processor readable instructions to perform a method of enhancing a computerized speed of precluding messages that pose a compliance, policy, or legal violation risk from being recorded in non-volatile memory during document creation, the method comprising:
parsing a first document into words and symbols via a parsing module, sequences of words, or words and symbols, being identified as ideas;
passing the first document to a semantic analyzer configured to:
access a policy model stored in a policy engine database;
access a violation threshold stored in the policy engine database;
assign a score to each of the ideas using the policy model;
determine that a total of scores for the first document surpasses the violation threshold;
disable save functionality of a device creating the first document, the save functionality configured to save the first document to non-volatile memory; and
provide the first document or a screenshot of the first document to an analytics dashboard configured to present the first document to a policy reviewer;
receiving a response from the policy reviewer indicating the policy reviewer's assessment of risk of the total of the scores for the first document;
storing the response in the policy engine database;
training the policy model, at a training module, using machine learning that uses the response as one of its inputs, to form an enhanced policy model; and
accessing the enhanced policy model and assigning scores to ideas parsed from a second document using the enhanced policy model.
14. The non-transitory, tangible computer readable storage medium of claim 13 , wherein the parsing module is part of the semantic analyzer.
15. The non-transitory, tangible computer readable storage medium of claim 13 , wherein the first document is received in real time as it is created.
16. The non-transitory, tangible computer readable storage medium of claim 13 , wherein the first document is received via import after first document creation.
17. The non-transitory, tangible computer readable storage medium of claim 13 , wherein the semantic analyzer further provides the total of scores for the first document to the analytics dashboard.
18. The non-transitory, tangible computer readable storage medium of claim 13 , wherein the semantic analyzer requests the screenshot from a device creating the first document in response to determining that the total of scores for the first document surpasses the violation threshold.
19. The non-transitory, tangible computer readable storage medium of claim 13 , wherein the analytics dashboard is further configured to provide the first document to a second policy reviewer, the method further comprising:
receiving a second response from the second policy reviewer indicating the second policy reviewer's assessment of risk of the total of the scores for the first document;
storing the second response in the policy engine database; and
training the policy model, at the training module, using machine learning that uses the second response as a second of its inputs, to form the enhanced policy model.