Skip to main content

A new collaborative tool to improve bibliographic matches in the “Thérapie Autisme” corpus

17 Nov 2026 Project

Scientific reference citation checks
Contribute anonymously to better bibliographic matches in the “Thérapie Autisme” corpus. No account or personal details are required. Proposals are reviewed, and decisions and changes are publicly documented.

Anyone can now easily contribute to improving bibliographic matches in the “Thérapie Autisme” corpus on AutiHub. Participation is anonymous: no account, name or email address is required. Proposals, decisions and changes are publicly documented.

This new collaborative tool allows anyone to suggest a publication that better matches an existing citation, or to report a citation missing from the corpus. The aim is to improve the accuracy of the results through verifiable contributions, with a publicly accessible record of how each proposal is handled.

Prioritising the weakest matches

If you know of a publication that better matches a citation currently assigned a low or zero score, we encourage you to suggest it. Citations with the most uncertain matches are those for which further research may be particularly useful.

A citation may contain an incorrect year, an incomplete title or the wrong journal name while still referring to a real publication. Automated searches may also fail to identify the intended publication. A low or zero score therefore does not demonstrate that the publication does not exist.

The “Improve the corpus” page displays citations with the lowest scores first by default. Each entry shows the citation as published, its source, the reference currently selected and the results of their comparison. A Google Scholar search is also provided to support further checks.

Two indicators are displayed separately: the overall bibliographic concordance score and the title Jaccard index, which measures the proportion of significant words shared by the two titles.

A straightforward way to propose a better reference

For the relevant citation, select “Propose a better reference” and enter the proposed publication’s DOI. This identifier can usually be found on the publisher’s page, often as a link beginning with https://doi.org/.

You can then test the reference to compare its results with those of the currently selected reference, and submit your proposal for review. A short arithmetic question is required for these actions.

An optional comment allows you to explain your proposal. This comment remains private; proposed references, decisions and the reasons for those decisions are publicly accessible.

Testing and submitting are separate actions: testing a reference does not automatically submit a proposal, and submitting a proposal does not immediately replace the published reference. The contribution guide explains the steps.

Human review before any proposal is incorporated

Each proposal is reviewed before it is incorporated into the corpus. A higher score provides useful information, but is not sufficient to establish that the proposed publication is the one intended by the citation.

To replace one publication with another, the proposed reference’s final score must exceed that of the reference selected at the time of review. Verified corrections concerning the same publication may also be accepted, including when they reduce its score. The objective remains a more accurate comparison of bibliographic information.

Decisions include a public explanation in French and English. An accepted community-proposed reference is labelled M3. References obtained through automated searches, M1 and M2, remain available with their results. These labels indicate where the references came from; they use the same comparison rules.

Reporting a missing citation

The tool also allows you to suggest a citation missing from the corpus.

To be considered, the citation must appear in Thérapie Autisme content published between 12 March 2025 and 13 June 2026 inclusive. The original text must already include at least a first author, a year and a title. These details must be reproduced as published, without adding information from the publication you believe the citation refers to.

A link to the source content is required. However, you do not need to have identified a DOI to report a missing citation.

The date range applies to the publication date of the content, not to the year of the study it cites. Each proposed addition is checked for its source, attribution and date.

An updated comparison method

This update also changes how the system chooses between references obtained through automated searches. Previously, title similarity had priority. The selection now gives priority to agreement across all comparable bibliographic information: title, authors, year, journal and DOI when cited and available.

The title remains a component of the score, and its Jaccard index is used to distinguish candidates with equal overall scores. The calculation rules, tolerances and conditions that can result in a final score of zero are explained in the public methodology.

The corpus and the individual “Test a citation” tool use the same scoring rules. These scores assess agreement between bibliographic details. They measure neither a study’s scientific quality nor the validity of conclusions presented in content that cites it.

Public records of proposals and changes

The documentation makes it possible to understand the results and follow their changes. Three complementary histories are available:

  • The proposal history presents submitted references, their scores, decisions and the reasons for those decisions.
  • The change history documents changes to the calculation, references and results, including information before and after each change.
  • The citation search history provides access to individual tests carried out with “Test a citation”.

Previous states are preserved. A pending proposal does not change the corpus statistics.

We invite everyone who knows of better matches, particularly for citations with the lowest scores, to contribute to their verification. Each documented proposal can help make the corpus more accurate and easier to check.

Access the collaborative “Improve the corpus” tool