---
title: "KGCL (peer-reviewed, Database 2025) implements node and edge changes but explicitly defers merge and split — the topology-reorganizing operations"
type: "claim"
status: "budding"
audit_status: "verified-verbatim (quotes cross-checked by the capture against both the OUP article-of-record and the PMC full-text mirror, matching wording)"
source_url: "https://doi.org/10.1093/database/baae133"
source_title: "A change language for ontologies and knowledge graphs"
source_author: "Hegde, Vendetti, Goutte-Gattat, Caufield, Graybeal, Harris, Karam, Kindermann, Matentzoglu, Overton, Musen, Mungall"
source_date: "2025-01-22T00:00:00.000Z"
source_venue: "Database (Oxford University Press), 2025, baae133 — peer-reviewed (Received 2024-09-20; Revised 2024-11-21; Accepted 2024-12-30; PMID 39841813)"
source_tier: 1
source_quote: "Instance-level (ABox) changes, like instance addition or deletion, and heuristic changes, such as concept merge and split, are yet to be considered"
provenance: "Promotion from 10-inbox/raw/2026-07-02-is-the-oxford-academic-ontology-evolution-paper-peer-reviewed.md, 2026-07-07, queen cycle 18"
origin: "batch"
derived_from: "10-inbox/raw/2026-07-02-is-the-oxford-academic-ontology-evolution-paper-peer-reviewed.md"
date_created: "2026-07-07T00:00:00.000Z"
tags: ["knowledge-graphs","ontology-evolution","kgcl","vault-governance","schema-restructuring"]
---


The Knowledge Graph Change Language paper — peer-review status confirmed
(full editorial trail on the article of record; PubMed/PMC indexed) — defines
its implemented taxonomy in Table 1 as two families: **NodeChange**
(rename, obsoletion, deletion, class creation, synonym/definition edits) and
**EdgeChange** (edge creation/deletion, NodeMove, PredicateChange).

What it does *not* implement is exactly the pair that reorganizes topology:
"heuristic changes, such as concept merge and split, are yet to be
considered" (Discussion — future work). "Merge" appears in the
introduction's informal framing of the problem space; "split" isn't even
mentioned there. So a "add/modify/merge/split" reading of KGCL (the form the
harvested lead carried) overstates its symmetry: add and modify map cleanly;
merge is acknowledged but unbuilt; split is absent. NodeMove is the closest
implemented operation to a restructuring move, and it relocates one node at
a time.

**Why this is on the vault's shelf.** The vault's own governance question —
targeted hooks-and-lint versus a global consolidation pass
([[question-consolidation-pass-vs-revisit-protocol]]) — is at bottom a
question about when a knowledge graph needs merge/split-class
reorganization rather than in-place edits. The most current peer-reviewed
change vocabulary for knowledge graphs stops short of formalizing exactly
those operations; a vault consolidation pass would be operating past the
edge of what the field has standardized. The cognitive-science taxonomy this
maps onto is now on the shelf: [[claim-schema-change-four-mode-taxonomy]]
(KGCL's node/edge edits ≈ accretion/tuning; the deferred merge/split ≈
restructuring/schema-formation). See [[moc-schema-change-and-restructuring]]
for that cluster and [[moc-knowledge-gap-detection]] for the sibling question
of *finding* what to reorganize.