OntoKG-EQ: A provenance-grounded, competency-question-governed knowledge graph for auditable analyst querying
Furqan Nasir, Muhammad Atif Saeed, Muhammad Ehsan, Sher Jeel Ahmad, Abdul Moiz Altaf
cs.ET · cs.AI · cs.DB
Quick verdict
Claims a CQ-governed, SHACL-gated KG making analyst queries auditable by construction; honest scoping but 4-firm slices and demonstrator data limit evidence.
Summary
OntoKG-EQ is a knowledge-engineering method that makes recurring equity-analyst queries over emerging-market data reproducible, evidence-linked, and auditable. Five frozen competency questions (CQ1–CQ5) govern the ontology: every class, property, SHACL shape, derived metric, and inference rule must be justified by one of them, a governance invariant the authors verify mechanically (24 classes, 45 properties, 9 metrics, 15 shapes). A seven-layer pipeline materialises curated market data into provenance-aware RDF, computes nine derived metrics (window returns, correlations, market-adjusted cumulative abnormal return, abnormal volume), gates on SHACL conformance, answers the questions with parameterised SPARQL, materialises typed findings via rules, and auto-generates per-result evidence bundles tracing to observations, sources, and provenance. On curated four-firm datasets from Pakistan, Malaysia, and Indonesia the ontology, shapes, queries, and rules are reused byte-identical (100%); a 64-stock Indonesia run (37,046 triples) answers in milliseconds on Oxigraph (sub-second at ~1.1M triples). A seriously engineered SQLite baseline returns identical analytics — a deliberate negative result locating the contribution in governance and open standards, not analytics. Eight open LLMs transcribing the same evidence bundles show provenance coverage from 0.00 to 1.00; a within-subject user study (n=17) shows large trust and completeness gains, with an acknowledged order-effect confound. Main limitations: hand-mapped, structurally similar markets; demonstrator comparators; CQ4 unexercised at scale.
Review complete
Assessment at a glance
4
risk flags identified
● high 0
● medium 2
● low 2
5 strengths · 5 weaknesses · 5 author questions
AI text signal = Binoculars zero-shot detector score over 20 sampled passages (Fast-DetectGPT co-reported) — absence does not prove human authorship, and a flag does not imply misconduct.
Levels reflect verification complexity — not a quality score. Evolve never rates your paper.
Key claims under review
- Five frozen competency questions act as a mechanically verifiable governance invariant: no ontology class, property, SHACL shape, derived metric, or inference rule exists without a justifying CQ (Section 7.1).
- Once each market's data is hand-mapped into a common nine-sheet schema, the ontology, SHACL shapes, CQ SPARQL templates, and rules are reused byte-identical across three emerging markets (100% reuse, Section 7.2).
- Faithfulness is a structural guarantee, not an estimate: answers rendered deterministically from a SHACL-validated graph with no generative step cannot drift from evidence, making OntoKG-EQ a provenance-grounded reference for measuring GraphRAG/LLM faithfulness (Sections 3.3, 7.3.1, 8.3).
- The knowledge-graph layer changes no analytics — a deliberate negative result — with the value residing in governance, provenance, and self-explaining structure on open standards (Section 7.3, Table 4).
- Attaching the evidence bundle significantly raises perceived trust (+2.87) and justification completeness (+3.35) in a within-subject study (n=17, both p < 0.001, Table 6), though order effects may contribute.
Strengths
Mechanically verified governance invariant
Section 7.1 introspects the ontology and shows all 24 classes, 22 object + 23 datatype properties, 9 derived metrics, and 15 SHACL shapes map to at least one of the five frozen CQs, with no uncovered terms; the CQs are released as versioned first-class artifacts (Section 2.2), going beyond the publish-your-CQs gap the paper documents in the literature.
A deliberate, serious negative result
Section 7.3 builds a genuine SQLite baseline (foreign-key/check constraints, provenance and lineage tables, explanation joins) that returns identical CQ1/CQ3 entities (e.g., PSX CQ3 = {ENGRO, OGDC}), and the authors state plainly that the graph layer adds no analytics — the contribution is governance, provenance, and open standards. Rare honesty that pre-empts the obvious 'why not SQL' objection.
Genuine fault-detection and scaling evidence
Statistical hygiene above the systems-paper bar
End-to-end reproducibility package
Weaknesses
Very small curated evaluation slices
Demonstrator economics throughout
Headline trust properties are definitional, not empirical
User-study confounds acknowledged but unresolved
Fragile CQ2/CQ4 screening semantics
Full Review Report
$1.99 · one-timeUnlock the complete preliminary analysis Evolve produced for this paper.
Report preview
Scope & Claims
OntoKG-EQ (arXiv 2609.08869) is a data-and-knowledge-engineering contribution: a knowledge-graph method making recurring equity-analyst queries over emerging-market data reproducible, evidence-linked, and inspectable. Scope is deliberately bounded by five frozen competency questions (CQ1 fundamentals-vs-market divergence, CQ2 exchange-rate association, CQ3 relative outperformance, CQ4 announcement/event-window abnormality, CQ5 explainability). The central governance claim is that every class, property, SHACL shape, derived metric, and inference rule is justified by at least one CQ, verified mechanically (Section 7.1: 24 classes, 22 object + 23 datatype properties, 9 metrics, 15 shapes, no uncovered terms). The authors explicitly disclaim prediction, causal-economic claims, and automatic portability. Faithfulness is claimed as a structural guarantee: answers rendered deterministically from a SHACL-validated graph with no generative step cannot drift from evidence.
The remaining sections cover Methodology Assessment, Statistical Evidence, Results & Reproducibility, Limitations & Risks, and Verdict & Recommendations.
- All 5 strengths & 5 weaknesses, analyzed in detail
- 4 risk flags with severity, section & reviewer questions
- Methods & statistical evidence assessment
- 5 questions a peer reviewer would ask
- Full six-dimension evaluation report & recommendations
One-time purchase · this paper only
One-time purchase, this paper only. AI-generated preliminary analysis — not peer review.
Frequently asked questions
What is OntoKG-EQ?
A knowledge-graph method that makes recurring equity-analyst questions over emerging-market data reproducible and auditable. A compact ontology is scoped by five frozen competency questions; SHACL validation gates all querying; parameterised SPARQL computes each question's condition; and every answer automatically carries an evidence bundle tracing it to observations, official sources, and provenance records. It is a data-engineering contribution, not a price-prediction system.
Does OntoKG-EQ beat GraphRAG or LLM-based question answering?
No head-to-head accuracy win is claimed. Because nothing is generated — results come from SPARQL over a validated graph — its answers are faithful by construction. The paper instead uses the system as a ground-truth reference to measure how faithfully eight open LLMs (0.5B–7B parameters) transcribe the same evidence bundles, finding provenance coverage ranging from 0.00 to 1.00 across models, unrelated to model size.
Is the code and data available?
Yes. The paper states the complete package — ontology, SHACL shapes, the five SPARQL templates, inference rules, builder and explanation generator, per-market data and materialised graphs, evaluation scripts, and pinned environments — is on GitHub (furqan-nr/OntoKG) and archived on Zenodo (DOI 10.5281/zenodo.21569316), including Indonesia data-acquisition scripts. The pipeline is deterministic, so rebuilds reproduce the same results.
What are the main limitations?
Evaluation markets are small curated slices (four firms each; the 64-stock Indonesia run lacks announcement data, so CQ4 is untested at scale), benchmarks and sector baskets are demonstrator constructions rather than licensed index feeds, cross-market reuse required manual data mapping across structurally similar markets, and the user study always presented the plain result before the evidence-backed version, which may inflate its trust gains.
About this review
This is an AI-generated preliminary review by Evolve.Pub, produced from the openly available text of the paper. It is not peer review and does not represent the authors' views. The canonical record of the paper is at the publisher link above.
- Online Resource 1 and Supplementary Tables S1–S6 are referenced but not included in the provided text; judgments resting solely on them (ablation details, per-query latencies, fault list) are based on main-text descriptions.
- Figures 1–2 are discussed but their visual content cannot be inspected from the extracted text.
Go deeper on this paper
Wrote a paper? Stress-test it before reviewers do.
Upload your manuscript for the same structured scrutiny — reasoning, statistics, evidence, figures, and reproducibility.