US 12,282,976 B2
Automatically enhancing content items through identifying relationships
Anna Tatunashvili, Bratislava (SK); George Thayyil Jacob Sushil, Bangalore (IN); Rupam Bhattacharjee, Karimganj (IN); and Siba Prasad Satapathy, Horamavu (IN)
Assigned to INTERNATIONAL BUSINESS MACHINES CORPORATION, Armonk, NY (US)
Filed by International Business Machines Corporation, Armonk, NY (US)
Filed on Dec. 8, 2021, as Appl. No. 17/545,858.
Prior Publication US 2023/0177631 A1, Jun. 8, 2023
Int. Cl. G06Q 50/20 (2012.01); G06F 9/445 (2018.01)
CPC G06Q 50/205 (2013.01) [G06F 9/44505 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A computer-implemented method, comprising operations for:
receiving one or more content items and an associated transcript for each of the one or more content items;
identifying one or more terms in the one or more content items;
training a content analyzer machine learning model by calibrating first weights;
using the trained content analyzer machine learning model that receives inputs of the one or more content items and each associated transcript and outputs a content tree for each of the one or more content items, wherein the content tree comprises first vertices and first edges, and wherein an edge to a vertex of the first vertices for a term of the one or more terms comprises an indication of whether the term is one of defined and mentioned;
training a graph builder machine learning model by calibrating second weights;
using the trained graph builder machine learning model that receives inputs of the one or more content items and each content tree output by the content analyzer machine learning model and outputs to build a content graph, wherein the content graph comprises second vertices for unique terms and second edges comprising one of defined type connections and mentioned type connections;
identifying one or more subgraphs of the content graph;
training a curriculum creator machine learning model by calibrating third weights;
using the trained curriculum creator machine learning model that receives an input of the content graph output by the graph builder machine learning model and outputs one or more paths through each of the one or more subgraphs, wherein the one or more paths in a subgraph intersect at an intersection content item in the subgraph and identify one or more prerequisite content items with associated second edges comprising the defined type connections for the intersection content item, and one or more subsequent content items with associated second edges comprising the mentioned type connections for the intersection content item; and
in response to selection of a content item of the one or more content items, sending an enhanced content item based on the one or more paths for the content item to a content renderer, wherein the content renderer plays the enhanced content item by playing the one or more prerequisite content items, the content item, and the one or more subsequent content items.