Abstract
Field Computer-implemented systems for determining which employees require requalification when authoritative reference documents are updated. This publication describes an approach that uses a knowledge dependency graph rather than section-level source-attribution records with reverse-index traversal. Background When a reference document changes, organizations need to identify which employees must be retrained and on what content specifically. One approach maintains per-assessment source-attribution records binding each employee's assessment to the specific document sections tested, then performs a reverse lookup when sections change. The approach described here uses a separate knowledge graph that abstracts documents into atomic concepts with dependency relationships. Technical Description The system maintains a knowledge dependency graph stored as adjacency lists in a relational database. The graph has three node types and three edge types. Node types: knowledge_item (knowledge_id, description, status), doc_fragment (fragment_id, doc_id, section_ref, content_hash), and learning_object (object_id, title, type). Edge types: knowledge_item to doc_fragment (source link, indicating which document fragment defines this knowledge item), knowledge_item to knowledge_item (dependency link, indicating prerequisite relationships), and knowledge_item to learning_object (coverage link, indicating which learning objects teach this item). Employee training records associate each employee_id with knowledge_item_ids they have demonstrated competency on, linked through completed learning_object_ids. When a controlled document is updated, the process works as follows. A content diff service compares the old and new versions of the document at the fragment level, using text hashing for exact matching and embedding similarity for semantic matching. Fragments that changed are identified by their fragment_ids. The system queries the knowledge graph to find all knowledge_items linked to the changed fragments. From those knowledge_items, the system recursively traverses dependency edges to discover downstream knowledge_items that depend on the changed items. The traversal uses breadth-first search with a configurable maximum depth (default: 3 hops). The combined set of directly affected and dependency-affected knowledge_item_ids is then joined against employee training records. Employees who have completed learning objects covering any affected knowledge_item receive requalification assignments scoped to those specific items. The requalification_task record contains: employee_id, knowledge_item_id, triggering_fragment_id, dependency_path (for items found via traversal), and recommended_learning_object_id. Distinguishing Characteristics This system does not maintain source-attribution records binding each assessment to specific document sections via composite identifiers. There is no secondary index on authority unit identifiers enabling reverse-index traversal. The lookup path runs through a separate knowledge graph abstraction: changed fragment to knowledge item to dependent items to employees, not through assessment-to-section attribution records. The granularity unit is the extracted knowledge concept, not the document section or proposition. There are no proposition-level validity bitmaps.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Davis, Kenneth, "Knowledge Dependency Graph Traversal for Document-Change-Triggered Employee Requalification", Technical Disclosure Commons, (March 25, 2026)
https://www.tdcommons.org/dpubs_series/9627