Granted Patent B2

US 12,724,762 · App. 18/987,241 · Granted Sep 1, 2026

Inventors: Craig Trim (Murfreesboro, TN); Janice Cha (Redwood City, CA); Mary Rudden (Denver, CO); Thanh Chi Lam (Aurora, CO)

Assignee: Bast, Inc.

G06F16/2365G06F18/2321

Abstract

Claims (71)

1 . A non-transitory processor-readable medium storing code representing instructions to be executed by a processor, the code comprising code to cause the processor to:

maintain, via the processor, concurrent domain-specific ontologies for a plurality of industries;

establish, via the processor, cross-domain relationships between concepts across different ontologies using semantic similarity algorithms;

map, via the processor, skills, certifications, and domain concepts across industry boundaries;

identify, via the processor, semantic similarities between concepts across different domain ontologies;

detect, via the processor, emerging domains;

establish, via the processor, initial cross-domain mappings for the emerging domains; and

maintain, via the processor, temporal vectors for tracking relationship evolution across the domains, wherein the processor synchronizes updates to the temporal vectors via conflict-free replicated data types; and

generate and present, via the processor, a cross-domain integration dashboard that visually represents the plurality of industries, cross-domain mappings of the plurality of industries that exceed a predefined confidence threshold, and integration metrics, wherein the integration metrics include the predefined confidence threshold and the processor computes the integration metrics using the semantic similarities.

2 . The non-transitory processor-readable medium of claim 1 , further comprising code to cause the processor to:

implement, via the processor, density-based spatial clustering with epsilon calculation using k-distance graphs;

apply, via the processor, an epsilon value of 0.3 for technical domains; and

apply, via the processor, an epsilon value of 0.5 for soft skills domains.

3 . The non-transitory processor-readable medium of claim 1 , further comprising code to cause the processor to:

implement, via the processor, Markov Decision Process modeling for career transitions by:

maintaining, via the processor, a state space comprising skills, roles, industry position, and career stage;

defining, via the processor, an action space comprising upskill, role change, and industry transition; and

implementing, via the processor, a reward function balancing salary gain, growth potential, and market demand.

4 . The non-transitory processor-readable medium of claim 3 , further comprising code to cause the processor to:

calculate, via the processor, the reward function as w1*salary_gain+w2*growth_potential+w3*market_demand; and

apply, via the processor, weight values of w1=0.4, w2=0.3, and w3=0.3.

5 . The non-transitory processor-readable medium of claim 1 , further comprising code to cause the processor to:

maintain, via the processor, bidirectional mappings using adjacency tensor representation; and

enable, via the processor, constant-time lookups for connected nodes.

6 . The non-transitory processor-readable medium of claim 1 , further comprising code to cause the processor to:

implement, via the processor, hypergraph-based relationship tracking; and

update, via the processor, weight vectors through gradient descent.

7 . The non-transitory processor-readable medium of claim 1 , further comprising code to cause the processor to:

apply, via the processor, dimensionality reduction techniques with minimum distance preservation constraints for relationship mapping.

8 . The non-transitory processor-readable medium of claim 1 , further comprising code to cause the processor to:

implement, via the processor, hierarchical clustering with boundary adjustment for emerging skill clusters.

9 . The non-transitory processor-readable medium of claim 1 , further comprising code to cause the processor to:

calculate, via the processor, composite relationship scores using weighted combinations of semantic similarity and usage patterns.

10 . The non-transitory processor-readable medium of claim 1 , further comprising code to cause the processor to:

implement, via the processor, recursive relationship validation with configurable consistency thresholds.

11 . The non-transitory processor-readable medium of claim 1 , further comprising code to cause the processor to:

implement, via the processor, conflict resolution using Conflict-Free Replicated Data Types for distributed updates.

12 . The non-transitory processor-readable medium of claim 1 , further comprising code to cause the processor to:

calculate, via the processor, transitivity scores for indirect relationships across domain boundaries.

13 . The non-transitory processor-readable medium of claim 1 , further comprising code to cause the processor to:

implement, via the processor, parallel processing for relationship discovery across multiple domains.

14 . The non-transitory processor-readable medium of claim 1 , further comprising code to cause the processor to:

maintain, via the processor, separate confidence metrics for different relationship types.

15 . The non-transitory processor-readable medium of claim 1 , further comprising code to cause the processor to:

generate, via the processor, relationship evidence chains for cross-domain mappings.

16 . The non-transitory processor-readable medium of claim 1 , further comprising code to cause the processor to:

generate, via the processor, cross-domain transition paths with confidence scoring.

17 . The non-transitory processor-readable medium of claim 1 , further comprising code to cause the processor to:

maintain, via the processor, relationship provenance with version tracking and validation history.

18 . The non-transitory processor-readable medium of claim 1 , further comprising code to cause the processor to:

apply, via the processor, spectral clustering overlay with normalized graph Laplacian;

calculate, via the processor, adaptive kernel bandwidth as σ=median_distance*scaling_factor; and

adapt, via the processor, the scaling_factor between 0.1 and 0.3 based on cluster density.

19 . A computer-implemented system for cross-domain ontology integration in explainable AI environments, comprising:

an ontology integration machine learning model configured to:

maintain concurrent domain-specific ontologies for a plurality of industries;

establish cross-domain relationships between concepts across different ontologies using semantic similarity algorithms; and

map skills, certifications, and domain concepts across industry boundaries;

maintain temporal vectors for tracking relationship evolution across the domain-specific ontologies, wherein the computer-implemented system synchronizes updates to the temporal vectors via conflict-free replicated data types;

a relationship mapping engine configured to;

identify semantic similarities between concepts across different domain ontologies;

generate and present a cross-domain integration dashboard that visually represents the plurality of industries, cross-domain mappings of the plurality of industries that exceed a predefined confidence threshold, and integration metrics, wherein the integration metrics include the predefined confidence threshold and the integration metrics are computed using the semantic similarities; and

a domain extension mechanism configured to detect emerging domains and establish initial cross-domain mappings.

20 . A method for providing explainable recommendations, the method comprising:

maintaining, via a processor, concurrent domain-specific ontologies for a plurality of industries;

establishing, via the processor, cross-domain relationships between concepts across different ontologies using semantic similarity algorithms; mapping, via the processor, skills, certifications, and domain concepts across industry boundaries;

identifying, via the processor, semantic similarities between concepts across different domain ontologies;

detecting, via the processor, emerging domains;

establishing, via the processor, initial cross-domain mappings for the emerging domains; and

maintaining, via the processor, temporal vectors for tracking relationship evolution across the domains, wherein the processor synchronizes updates to the temporal vectors via Conflict-Free Replicated Data Types; and

generating and presenting, via the processor, a cross-domain integration dashboard that visually represents the plurality of industries, cross-domain mappings of the plurality of industries that exceed a predefined confidence threshold, and integration metrics, wherein the integration metrics include the predefined confidence threshold and the processor computes the integration metrics using the semantic similarities.

Continuity (1)

References Cited (41)