Generating actionable information from documents
AssigneeCodexo
InventorsSébastien HUPPE, Marc DANGEARD
Life of this patent
Estimating the term…
Abstract
Systems and methods for creating actionable data from documents are described. In some implementations, the systems and methods deconstruct, or break up, documents into blocks (e.g., a paragraph or other segment or portion of text of the document) and store the blocks in a database. The systems and methods tag the blocks with intelligence about the content of the blocks. For example, the systems and methods can tag blocks with commitment markers or other information that identify the blocks as containing or reciting a commitment (e.g., a contractual provision, term, or obligation) within a contract document.
Claims7
A method for generating a graph database of document blocks from multiple documents provided by an enterprise, the method comprising:
- extracting one or more commitments from the multiple documents, wherein a commitment is represented by a block of a document;
- tagging each of the blocks associated with the document with commitment markers, wherein a commitment marker relates a block to a commitment;
- storing the tagged blocks in a graph database for the enterprise.
A non-transitory computer-readable medium whose contents, when executed by a computing system, cause the computing system to perform a method for generating a database of document blocks from multiple documents provided by an enterprise, the method comprising:
- extracting multiple blocks from the multiple documents,
- wherein a block includes a portion of text within a document;
- tagging each of the blocks associated with the document with commitment markers, wherein a commitment marker relates a block to a commitment, and
- wherein a commitment includes a provision identified in the portion of text;
- storing the tagged blocks as commitments in a graph database for the enterprise,
- wherein the graph database maps tagged blocks to documents from which the blocks are extracted.
Dependencies are inferred from the claim text.