Citing Government of Canada Open Data: A Verified Citation Corpus and the Limits of Dataset-Level Impact Measurement

This study examines how Government of Canada open data is cited in academic literature. A screening ledger of 110 federal organizations recorded 32,784 cited references in the Web of Science and Scopus; screening and annotation were completed for 59 organizations, accounting for 28,503 of those references. Manual annotation identified 1,337 verified citation instances across 878 articles, of which 1,119 instances in 767 articles met a strict criterion requiring an explicit data reference. Only 2.8 per cent of academic references to a federal organization were traceable citations to its data rather than to a report, guideline, or the organization itself. Among organizations with at least 100 screened references, this rate varied 65-fold, from 0.4 per cent to 26.2 per cent. Citations were concentrated in health, environmental, and immigration-related organizations. Where dataset-level attribution was possible, 187 dataset identifiers were extracted and 107 resolved against the portal catalogue; 81 (75.7 per cent) of these were cited by exactly one article and none by more than five, indicating that dataset-level citation counting cannot presently support impact ranking for a collection of this kind. A BERT model fine-tuned on the verified annotations achieved 0.975 F1 on in-distribution test data, against 0.890 for keyword matching and 0.909 for SPECTER embeddings. A census of all 24 held-out misclassifications shows that the model does not distinguish Government of Canada data from data of other provenance, and that its false positives are concentrated above 0.90 confidence; therefore, thresholding does not compensate. The corpus, screening ledger, annotations, model, and code are openly available.

Authors

Publication Details

Journal
International Journal of Digital Curation
Published
2026-11-16
DOI
https://doi.org/10.2218/ijdc.v20i1.1118
Primary Topic
scientometrics and bibliometrics research
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Citing Government of Canada Open Data: A Verified Citation Corpus and the Limits of Dataset-Level Impact Measurement

Michael Groenendyk
International Journal of Digital Curation
scientometrics and bibliometrics research
article

Citing Government of Canada Open Data: A Verified Citation Corpus and the Limits of Dataset-Level Impact Measurement

Michael Groenendyk
article en

Abstract

This study examines how Government of Canada open data is cited in academic literature. A screening ledger of 110 federal organizations recorded 32,784 cited references in the Web of Science and Scopus; screening and annotation were completed for 59 organizations, accounting for 28,503 of those references. Manual annotation identified 1,337 verified citation instances across 878 articles, of which 1,119 instances in 767 articles met a strict criterion requiring an explicit data reference. Only 2.8 per cent of academic references to a federal organization were traceable citations to its data rather than to a report, guideline, or the organization itself. Among organizations with at least 100 screened references, this rate varied 65-fold, from 0.4 per cent to 26.2 per cent. Citations were concentrated in health, environmental, and immigration-related organizations. Where dataset-level attribution was possible, 187 dataset identifiers were extracted and 107 resolved against the portal catalogue; 81 (75.7 per cent) of these were cited by exactly one article and none by more than five, indicating that dataset-level citation counting cannot presently support impact ranking for a collection of this kind. A BERT model fine-tuned on the verified annotations achieved 0.975 F1 on in-distribution test data, against 0.890 for keyword matching and 0.909 for SPECTER embeddings. A census of all 24 held-out misclassifications shows that the model does not distinguish Government of Canada data from data of other provenance, and that its false positives are concentrated above 0.90 confidence; therefore, thresholding does not compensate. The corpus, screening ledger, annotations, model, and code are openly available.

International Journal of Digital CurationVol. 20(1)
scientometrics and bibliometrics research
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon