A Bibliometric Dataset Cleaning Checklist
A bibliometric map is only as interpretable as the dataset and decisions behind it. Before producing a visualisation, document your search boundaries, export settings and cleaning steps so that the results can be reviewed.
Keep an untouched export
Save the original database export and a working copy. Record the database, search date, search expression, filters and export format. Avoid silently replacing source records while cleaning.
Check the records before mapping
- Duplicates: Compare identifiers, titles, publication years and author information. Do not rely on title matches alone.
- Completeness: Check fields needed for your intended analysis, such as authors, keywords, affiliations or cited references.
- Name variants: Review author and institution variants carefully. Similar names can belong to different people or organisations.
- Keyword variants: Consider spelling, abbreviations, singular/plural forms and compound terms.
- Scope: Check that the retained document types, years and languages match your protocol.
- Missing values: Record what is missing and how that affects the analysis.
Create a cleaning log
| Decision | Record |
|---|---|
| Term merged | Original variants, preferred term and rationale. |
| Record removed | Identifier, reason and stage of removal. |
| Name disambiguated | Evidence used to distinguish or combine entries. |
| Threshold changed | Previous value, new value and impact on the displayed network. |
Interpret cautiously
Document the counting approach, thresholds and any normalisation settings relevant to your tool and analysis. A prominent node does not automatically establish research quality. Cluster labels are interpretive summaries that should be checked against the underlying records.
Review robustness
Consider whether a reasonable change in inclusion criteria or thresholds materially changes the pattern. Report important sensitivity rather than presenting one attractive map as definitive evidence. Keep the dataset, cleaning decisions and outputs together for review.
Discuss the required outputs and dataset preparation through our bibliometric analysis service. For reporting structure, explore research samples. Tool-specific procedures should follow the documentation for the version used in your project.
2 Responses
[…] human review. Record each merge, exclusion and unresolved ambiguity in a cleaning log. Use our bibliometric dataset cleaning checklist to organise this […]
[…] for the appropriate workflow. Check current manuals when a database export format changes. Our dataset cleaning checklist covers preparation decisions to […]