| CPC G06F 16/345 (2019.01) [G06F 3/167 (2013.01); G06F 16/367 (2019.01); G06F 16/9024 (2019.01); G06F 40/284 (2020.01); G06T 11/206 (2013.01)] | 15 Claims |
|
1. A computer-implemented method comprising:
obtaining, via a user interface, memory input data from one or more users, wherein the memory input data comprises a digital media artifact and one or more pieces of memory-descriptive data that is descriptive of a target user-recalled memory associated with the digital media artifact;
automatically constructing, by one or more processors, a target memory graph based on the memory input data, wherein the target memory graph comprises a plurality of graphical nodes representing a semantic illustration of the memory input data, wherein constructing the target memory graph includes:
computing a plurality of distinct sets of embedding values based on the memory input data;
extracting a plurality of distinct semantic features of the memory input data based on an input of the plurality of distinct sets of embedding values to a semantic data extraction model;
for each of the plurality of distinct semantic features of the memory input data:
creating a distinct graphical node of the plurality of graphical nodes of the target memory graph;
storing in association with the distinct graphical node a given semantic feature of the plurality of distinct semantic features and a given set of embedding values of the plurality of distinct sets of embedding values;
creating, by the one or more processors, a semantic collage using the target memory graph, wherein:
the semantic collage comprises a plurality of distinct memory graphs interconnected via one or more graphical connections,
creating the semantic collage includes setting at least one semantic nexus between a node of the target memory graph and a node of a historical memory graph of the semantic collage,
the at least one semantic nexus relates to a graphical connection that connects nodes of the target memory graph and the historical memory graph that have an identified semantic relationship, wherein setting the at least one semantic nexus between the node of the target memory graph and the node of the historical memory graph includes:
computing a distance between embedding values of the plurality of graphical nodes of the target memory graph and embedding values of a plurality of graphical nodes of the historical memory graph,
determining that a computed distance between the node of the target memory graph and the node of the historical memory graph satisfies or is less than a matching distance threshold,
setting the at least one semantic nexus between the node of the target memory graph and the node of the historical memory graph based on the determination, and
generating, via a machine learning model, a mnemonic narrative artifact based the semantic collage, wherein the mnemonic narrative comprises a semantic summarization of the semantic collage that is synthesized using data extracted from the semantic collage; and
storing, in a queryable computer database, the mnemonic narrative artifact in digital association with the semantic collage and the digital media artifact, wherein if a semantic search of the queryable computer database identifies the semantic collage or the mnemonic narrative artifact, returning via a user interface one or more of the digital media artifact and the mnemonic narrative artifact.
|