Document marking techniques using semantically similar phrases for document source detection

One or more computer storage media storing computer readable instructions thereon that, when executed by a processor, cause the processor to perform a method comprising:

• accessing a first set of alternative terms generated for a first term in a document and a second set of alternative terms generated for a second term in the document;
• determining tone for the alternative terms of the first set of alternative terms and the second set of alternative terms using a machine learning model;
• selecting one or more alternative terms from each of the first set of alternative terms and the second set of alternative terms based on a tone of the first term and a tone of the second term relative to a tone of the one or more alternative terms as determined by the machine learning model;
• generating a plurality of unique copies of the document, each unique copy being distinct from other unique copies of the document and generated by replacing the first term and the second term in unique combinations with the selected one or more alternative terms;
• individually distributing the plurality of unique copies to a plurality of recipients;
• and
• indexing an association between the plurality of unique copies and the plurality of recipients, such that the index identifies a recipient for an identified unique copy.

A computer implemented method performed by one or more processors, the method comprising:

• accessing a set of documents having a same author;
• training a machine learning model to identify alternative terms from input terms based on content within the set of documents;
• determining, using the trained machine learning model, a first set of alternative terms for a first term within a document and a second set of alternative terms for a second term within the document;
• selecting one or more alternative terms from each of the first set of alternative terms and the second set of alternative terms based on a tone of the term and a tone of the second term relative to a tone of the one or more alternative terms;
• generating a plurality of unique copies of the document, each unique copy being distinct from other unique copies of the document and generated by replacing the first term and the second term in unique combinations with the selected one or more alternative terms;
• individually distributing the plurality of unique copies to a plurality of recipients;
• and
• indexing an association between the plurality of unique copies and the plurality of recipients, such that the index identifies a recipient for an identified unique copy.