A computer-implemented system for generating and executing a machine-readable checklist template from an unstructured electronic document comprising one or more images, the system comprising:
- an import module to import an unstructured electronic document comprising one or more images and a source document from a digital data storage device;
- a GUI module to present a graphical user interface displaying at least a portion of the unstructured electronic document to a user;
- a markup module to receive, via the graphical user interface, a markup of a portion of the unstructured electronic document corresponding to a salient term;
- a GUI term list creation module to generate a term list entry for the salient term, the term list entry comprising:
- a term name for the salient term;
- a data object type associated with the salient term;
- and
- an association between the salient term and the markup of the portion of the unstructured electronic document;
- a ruleset module to generate an extraction ruleset associated with the term list entry, wherein the extraction ruleset includes a contextual extraction rule that identifies, for the salient term, at least one of:
- a relative location of the markup within the unstructured electronic document,
- a relative location of the salient term within the unstructured electronic document,
- identifiable text expected to be proximate to the salient term,
- an identifiable image expected to be proximate to the salient term,
- a formatting style of the salient term, and
- an expected formatting variance of the salient term;
- wherein the contextual extraction rule is stored in the machine-readable checklist template for use in extracting, from a subsequently imported unstructured electronic document an extracted term corresponding to the salient term;
- a GUI checklist rule creation module to receive, via the graphical user interface, an analysis ruleset associated with the term list entry, the analysis ruleset comprising:
- a source term from the source document;
- a comparison rule for comparing the extracted term with a source value corresponding to the source term;
- and
- a reconciliation rule for determining whether the extracted term and the source value match or differ;
- wherein the GUI term list creation module and the GUI checklist rule creation module generate the machine-readable checklist template comprising the term list entry, the extraction ruleset, and the analysis ruleset, the analysis ruleset including the source term, the comparison rule, and the reconciliation rule;
- a ruleset review module configured to:
- present the term list entry, the extraction ruleset, the source term, the comparison rule, and the reconciliation rule to the user for review;
- receive a user modification to at least one rule of the extraction ruleset or the analysis ruleset;
- and
- in response to user confirmation, publish the machine-readable checklist template;
- and
- a trained AI system configured to apply the published machine-readable checklist template to a subsequently imported unstructured electronic document comprising one or more images and a subsequently imported source document by:
- extracting, from the subsequently imported unstructured electronic document and based on the contextual extraction rule, the extracted term;
- extracting, from the subsequently imported source document and based on the source term, the source value;
- applying the comparison rule and the reconciliation rule to the extracted term and the source value to generate a reconciliation result;
- and
- generating an executed checklist comprising the extracted term, the source value, the reconciliation result, and an action to be taken based on whether the reconciliation result indicates a match or difference between the extracted term and the source value.