US 12,242,432 B2
Guiding a generative model to create and interact with a data structure
Moshe Randall Lutz, Bellevue, WA (US)
Assigned to Microsoft Technology Licensing, LLC, Redmond, WA (US)
Filed by Microsoft Technology Licensing, LLC, Redmond, WA (US)
Filed on Mar. 16, 2023, as Appl. No. 18/122,229.
Prior Publication US 2024/0311348 A1, Sep. 19, 2024
Int. Cl. G06F 17/30 (2006.01); G06F 16/21 (2019.01); G06F 16/245 (2019.01); G06F 16/25 (2019.01); G06F 16/355 (2025.01)
CPC G06F 16/211 (2019.01) [G06F 16/245 (2019.01); G06F 16/258 (2019.01); G06F 16/355 (2019.01)] 20 Claims
OG exemplary drawing
 
1. A computer-implemented method for creating and utilizing a structured database, comprising:
receiving unstructured data;
successively extracting prescribed items-of-interest expressed in the unstructured data, using a machine-trained pattern-completion engine;
successively identifying categories associated with the items-of-interest that have been previously extracted, using the machine-trained pattern-completion engine;
successively identifying, using the machine-trained pattern-completion engine, item-to-category relations between the items-of-interest that have been previously extracted and the categories that have been previously identified and category-to-category relations among the categories that have been previously identified, to provide identified relations;
constructing the structured database based on the items-of-interest, categories, and identified relations that have been identified; and
storing the structured database in a data store, the data store being made available to an application for performing an application task,
the structured database being expressed as a directed graph, the directed graph having leaf nodes associated with the identified items-of-interest, and parent and other nodes associated with the identified categories.