Machine learning methods and systems for protection and redaction of privacy information

A system comprising:

• one or more processors;
• and
• a non-transitory computer readable medium storing a plurality of instructions, which when executed, cause the one or more processors to:
• receive one or more initial electronic documents, the electronic documents including respective portions of content comprising textual or audio data;
• store the received one or more initial electronic documents in a data store device;
• evaluate the one or more initial electronic documents to build a content profile based on detecting information at one or more of the respective portions of content that indicates a pre-defined category, from a plurality of different types of pre-defined categories, that corresponds to each of the one or more respective portions of content;
• for each respective portion of content, apply at least one machine learning model that corresponds with the indicated pre-defined category of the respective portion to determine a probability value of whether the respective portion of content includes data considered likely to be privacy information, wherein a first type of machine learning model that corresponds with a first pre-defined type of category indicated for a first portion of content differs from a second type of machine learning model that corresponds with a second pre-defined type of category indicated for a second portion of content;
• and
• recreate the one or more initial documents according to one or more privacy information redactions at respective locations of the respective portions of content;
• wherein apply at least one machine learning model comprises:
• apply a pre-determined machine learning model for a financial business domain category to a header portion of a particular electronic document that includes content indicative of the financial business domain;
• wherein determine a probability value comprises:
• determine, via application of the pre-determined machine learning model for the financial business domain category, a first probability value of whether the content in the header portion is considered to be financial privacy information;
• wherein recreate the one or more initial documents according to one or more privacy information redactions comprises:
• generate a version of the particular electronic document to include a visual modification at a document position of the header portion, the visual modification obscuring the content indicative of the financial business domain.

A computer program product comprising computer-readable program code to be executed by one or more processors when retrieved from a non-transitory computer-readable medium, the program code including instructions to:

• receive one or more initial electronic documents, the electronic documents including respective portions of content comprising textual or audio data;
• store the received one or more initial electronic documents in a data store device;
• evaluate the one or more initial electronic documents to build a content profile based on detecting information at one or more of the respective portions of content that indicates a pre-defined category, from a plurality of different types of pre-defined categories, that corresponds to each of the one or more respective portions of content;
• for each respective portion of content, apply at least one machine learning model that corresponds with the indicated pre-defined category of the respective portion to determine a probability value of whether the respective portion of content includes data considered likely to be privacy information, wherein a first type of machine learning model that corresponds with a first pre-defined type of category indicated for a first portion of content differs from a second type of machine learning model that corresponds with a second pre-defined type of category indicated for a second portion of content;
• and
• recreate the one or more initial documents according to one or more privacy information redactions at respective locations of the respective portions of content;
• wherein apply at least one machine learning model comprises:
• apply a pre-determined machine learning model for a financial business domain category to a header portion of a particular electronic document that includes content indicative of the financial business domain;
• wherein determine a probability value comprises:
• determine, via application of the pre-determined machine learning model for the financial business domain category, a first probability value of whether the content in the header portion is considered to be financial privacy information;
• wherein recreate the one or more initial documents according to one or more privacy information redactions comprises:
• generate a version of the particular electronic document to include a visual modification at a document position of the header portion, the visual modification obscuring the content indicative of the financial business domain.