| CPC G06V 30/414 (2022.01) [G06F 16/9024 (2019.01); G06F 16/9038 (2019.01); G06F 16/906 (2019.01); G06F 16/93 (2019.01); G06V 30/19173 (2022.01); G06V 30/413 (2022.01)] | 24 Claims |
|
7. A network computer for managing documents, comprising:
a memory that stores at least instructions; and
one or more processors that execute instructions that are configured to cause actions, including:
determining a plurality of terms in a document based on a term location of each term within the document;
generating a layout graph that includes a plurality of term nodes based on the plurality of terms, wherein each of the plurality of terms corresponds to a different one of the plurality of term nodes, and wherein each term node is linked by one or more edges to one or more neighboring term nodes;
determining one or more relationships between the plurality of term nodes based on a traversal of the layout graph, wherein the layout graph is partitioned into one or more content shapes based on a strength of the one or more relationships;
classifying the one or more content shapes to reduce computational resources for extracting information from the document based on one or more types associated with each content shape by performing further actions, including:
associating the one or more types with one or more extraction models that each support one or more of a plurality of different types of document formats, document sources, and document platforms;
generating one or more prompts that include one or more features associated with the one or more content shapes, wherein the one or more prompts are provided to one or more generative artificial intelligence models, and wherein the one or more features include one or more of a term, a style feature, or a layout feature; and
determining the one or more types for the one or more content shapes based on one or more responses from the one or more generative artificial intelligence models; and
employing the one or more extraction models to extract information from the one or more classified content shapes, wherein the extracted information is stored in one or more data stores or included in one or more reports.
|