Auditable authorship attribution with automatically applied authorship tokens
AssigneeU.S. Bank National Association
InventorsCHRISTOPHER ZIOLKOWSKI, VALERIE LANCELLE, CHRISTOPHER DAVIS
Life of this patent
Estimating the term…
Abstract
A text editor or a plugin thereto automatically generates authorship tokens to identify content authored by a human author or an artificial author. The authorship tokens are applied to the work while the work is being produced. Thus, subsequent review of the work can identify regions produced by a human author and other regions produced by an artificial intelligence.
Claims20
A method comprising:
- detecting an actuation of a first user interface element of text editing software or a plugin thereto and in response to detecting the actuation, exporting text in a manner that preserves authorship tokens associated with regions of the text, wherein the authorship tokens include:
- at least one human authorship token indicating that a first associated region of the text has human authorship rather than artificial authorship based on a human user typing a manual edit at the first associated region of the text;
- at least one external authorship token indicating that a second associated region is based on externally sourced text;
- at least one authorship token indicating that a third associated region of the text is based on text typed in a fashion dissimilar to human typing;
- at least one mixed authorship token indicating that a fourth associated region of the text is associated with mixed human and artificial authorship;
- and
- at least one artificial authorship token indicating that a fifth associated region of the text is associated with artificial authorship.
A non-transitory computer readable medium having instructions stored thereon that, when executed by one or more processors, cause the one or more processors to perform a method comprising:
- providing a first user interface element actuatable to cause activation of an authorship tracking feature of text editing software;
- responsive to detecting actuation of the first user interface element and automatically during an edit session where text is being edited using with the text editing software:
- monitoring edits to the text, determining authorship of the edits, and associating authorship tokens with associated regions;
- providing a second user interface element;
- and
- responsive to detecting the actuation of the first user interface element of text editing software or a plugin thereto and in response thereto, exporting text in a manner that preserves authorship tokens associated with regions of the text, wherein the authorship tokens include three or more types of authorship tokens selected from the group consisting of:
- human authorship tokens indicating that a first associated region of the text has human authorship rather than artificial authorship based on a human user typing a manual edit at the first associated region of the text;
- external authorship tokens indicating that a second associated region is based on externally sourced text;
- dissimilar authorship tokens indicating that a third associated region of the text is based on text typed in a fashion dissimilar to human typing;
- unknown provenance authorship tokens indicting that a fourth associated region of the text has unknown provenance;
- mixed authorship tokens indicating that a fifth associated region of the text is associated with mixed human and artificial authorship;
- and
- artificial authorship tokens indicating that a sixth associated region of the text is associated with artificial authorship and indicating an identity of an artificial intelligence used.
A method comprising:
- actuating a first user interface element of text editing software or a plugin thereto, thereby causing the text editing software or the plugin thereto to export text in a manner that preserves authorship tokens associated with regions of the text, wherein the authorship tokens include:
- at least one human authorship token indicating that a first associated region of the text has human authorship rather than artificial authorship based on a human user typing a manual edit at the first associated region of the text;
- at least one external authorship token indicating that a second associated region is based on externally sourced text;
- at least one authorship token indicating that a third associated region of the text is based on text typed in a fashion dissimilar to human typing;
- at least one mixed authorship token indicating that a fourth associated region of the text is associated with mixed human and artificial authorship;
- and
- at least one artificial authorship token indicating that a fifth associated region of the text is associated with artificial authorship and indicating an identity of an artificial intelligence used.
Dependencies are inferred from the claim text.