From Code Archival to Knowledge Graph: Bridging Software Heritage, COAR Notify and Wikidata
Camillo Carlo Pellizzari di San Girolamo, Francesco Tosoni
cs.DL · cs.AI · cs.SE
Quick verdict
An open, carefully documented pipeline linking archived code to Wikidata at 4,397 verified pairs; solid infrastructure, but promised precision auditing is still pending.
Summary
The paper presents an end-to-end reconciliation pipeline that harvests paper-to-code links from four venues where the link is editorially verified — JOSS, SoftwareX, IPOL, and SIGMOD's ARI reports — yielding 4,397 curated ⟨DOI, repository⟩ pairs. Two Wikidata application profiles separate scholarly articles from software instances, joined via P921/P1343 and reconciled at two granularities: inline references or standalone nodes carrying repository URLs (P1324) and SWHIDs (P6138). A read-only lookup (29 July 2026) found only 82 of 4,397 repositories (1.9%) already modelled, and human-reviewed batches created 4,182 software items plus 2,326 article items. The input format is shown to map structurally onto COAR Notify payloads, an external protocol left undeveloped. Main limitations: no precision metrics were computed for the self-described baseline corpus; URL-equality matching has known false negatives (SWHID anchoring deferred); SIGMOD extraction was validated via a single LLM cross-check (90/102 = 88.2% agreement); the SoftwareX subset is heavily selected (486 of 2,326); and several numeric reconciliations are unexplained.
Review complete
Assessment at a glance
5
risk flags identified
● high 0
● medium 3
● 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
- An open corpus of 4,397 curated ⟨DOI, repository⟩ pairs with editorially correct paper-to-code links was consolidated from JOSS, SoftwareX, IPOL and SIGMOD ARI (Table 1).
- Two distinct Wikidata application profiles (scholarly article vs software instance), aligned with schema.org and CodeMeta, connect SWH to the scholarly record at reference and node granularity via P921/P1343.
- A read-only, rule-based reconciliation pipeline (exact DOI/URL keys, no name matching) generates provenance-tagged QuickStatements/OpenRefine batches that humans review before any write.
- Only 82 of 4,397 harvested repositories (1.9%) already existed in Wikidata as of 29 July 2026; human-reviewed batches subsequently created 4,182 software items and 2,326 article items.
- COAR Notify Announce payloads map structurally onto the pipeline's ⟨DOI, repo⟩ input format, so a future live stream could reuse the same backend (protocol itself not developed).
Strengths
Editorially verified sources yield a genuinely curated corpus
All four harvests (§3.1) rely on venue workflows where paper↔repo is mandatory and human-checked (JOSS catalogue API 3,596→3,545 pairs; SoftwareX via Crossref PII join; IPOL BibTeX SWHIDs; SIGMOD ARI reports), rather than noisy full-text mention mining (Softcite/SoMeSci).
Manual audits with documented failure cases
§3.1 footnotes document and repair concrete extraction defects: the ElsevierSoftwareX/SOFTX_2020_261 PII mismatch (wrong DOI 10.1016/j.softx.2020.100640 excluded), JOSS paper-support repositories (4 rows), PDF line-wrap URL truncations, and concatenated IPOL SWHIDs repaired via the SWH API.
Writes only through human-reviewed batches
Two-granularity design is well reasoned
Strong reproducibility posture
Weaknesses
Evaluation stops at the lookup stage; the promised precision audit is missing
Three numeric reconciliations are unexplained
SIGMOD extraction validated by a single LLM pass
The SoftwareX subset is heavily selected
URL-equality matching is the sole reconciliation key in practice
Full Review Report
$1.99 · one-timeUnlock the complete preliminary analysis Evolve produced for this paper.
Report preview
Scope & Claims
The paper (Wikidata'26 workshop track) describes an end-to-end reconciliation pipeline that bridges Software Heritage (SWH) and Wikidata. It claims four contributions (§1): (i) an open corpus of 4,397 curated ⟨DOI, repository⟩ pairs from sources where the paper-to-code link is editorially verified (JOSS, SoftwareX, IPOL, SIGMOD ARI); (ii) — called the central contribution — dual application profiles for scholarly articles and software instances aligned with schema.org and CodeMeta; (iii) a rule-based, provenance-tagged reconciliation pipeline with human-reviewed imports; (iv) a structural mapping showing COAR Notify payloads fit the pipeline's input format. The empirical core is Table 1 (per-venue harvest yields), Table 4 (read-only lookup, 29 July 2026: 82 of 4,397 repositories, 1.9%, already modelled; 1,967 of 4,397 articles, 44.7%, existing via WikiCite), and the import totals (4,182 software items, 2,326 article items created).
The remaining sections cover Methodology Assessment, Statistical Evidence, Results & Reproducibility, Limitations & Risks, and Verdict & Recommendations.
- All 5 strengths & 5 weaknesses, analyzed in detail
- 5 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 does this paper actually do?
It builds and runs an open-source pipeline that harvests paper-to-code links from venues where the link is editorially verified (JOSS, SoftwareX, IPOL, SIGMOD's reproducibility reports) and models them in Wikidata: scholarly-article items linked to standalone software items carrying repository URLs and Software Heritage IDs. 4,397 verified DOI-repository pairs and 4,182 newly created Wikidata software items result, all via human-reviewed batches.
What is in the released dataset?
4,397 curated ⟨DOI, repository URL⟩ pairs: 3,545 from the JOSS catalogue, 486 from SoftwareX, 275 from IPOL, and 91 from SIGMOD ARI reports, plus harvest code, reconciliation code, and per-batch QuickStatements import reports on GitHub (github.com/ftosoni/swh-wd-reconciliation).
What is a SWHID?
A SoftWare Hash IDentifier: Software Heritage's intrinsic, content-addressed identifier for any archived source-code object (e.g., a specific directory tree). Because it is derived from the code itself, it stays stable across repository migrations and link rot, and it is standardised as ISO/IEC 18670:2025. The paper attaches SWHIDs to Wikidata items via property P6138.
Is the pipeline already live on Wikidata?
Yes for the offline part: the authors inspected and ran every import batch, creating 4,182 software items and 2,326 article items, all auditable via EditGroups and QuickStatements run reports. The COAR Notify live-enrichment part is prospective only — the paper shows payload compatibility but does not operate a notification stream.
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.
- The abstract in the source feed is truncated mid-sentence, but the embedded full text is complete; figures (1, 2, 4) and tables were assessed from captions and surrounding text only.
- Live-service numbers (Wikidata lookup, SWH Save Code Now outcomes) are point-in-time snapshots dated 29 July 2026 and cannot be re-verified from the text alone.
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.