Methodology

Return to the project page

The analyzed corpus includes 533 citations of scientific studies, collected over the period from March 12, 2025 to June 13, 2026. These citations come from content published or distributed by the company “Thérapie Autisme”, represented by Florence Demourant. It includes posts from the Instagram account “therapietsa” up to June 9, 2026, articles from the “Thérapie Autisme” blog up to June 13, 2026, bibliographic documents associated with tools marketed by “Thérapie Autisme”, and the manuscript titled “Variabilité neuro-architecturale dans l’autisme adulte : vers une cartographie morpho-fonctionnelle”, published on the Skool.com platform “Thérapie TSA - Autisme Adultes”.

Process overview

  • 1. Citation. Cited text with author(s), year and title required; journal, volume, pages and DOI optional.
  • 2. Extraction. Separation of visible bibliographic components, without reformatting the citation into an academic style.
  • 3. Method 1. Crossref, OpenAlex and PubMed search from author + year + title.
  • 4. Method 2. Google Scholar search through SerpApi when this source could provide a better candidate.
  • 5. Final score. Selection of the final candidate, element-by-element comparison, general score and title Jaccard index.
  • 6. Records. Presentation of the citation, the reference found, verification links and calculation details.

Corpus construction

The analysis includes citations that contain at least one author, one year and a sufficiently explicit title. Citations without a title, and citations that appear to be approximate French paraphrases of an English scientific title rather than an identifiable bibliographic title, are excluded.

  • Author(s). Must be present in a usable form: single author, multiple authors or “et al.”.
  • Year. Must be present to place the reference and limit homonymy.
  • Title. Must be present as an identifiable bibliographic title, not only as a reformulation or summary.
  • Journal, DOI, volume, pages. Useful for assessing fidelity, but not required to include the citation.

Candidate reference search

  • M1: Crossref, OpenAlex, PubMed. Search from author, year and title. The cited DOI does not drive the search, because an incorrect DOI may lead directly to the wrong reference. This method may miss some titles or return incomplete metadata.
  • M2: Google Scholar via SerpApi. Reinforcement source when the citation is incomplete, imprecise or phrased differently. Its metadata are less structured; journal or volume may be missing.
  • Final. M1 or M2. This is the reference used in the charts and records. Ambiguous cases still require human reading.

Final candidate selection

When several candidates were available, the selected reference was the one whose title had the best correspondence with the cited title according to the Jaccard index. In case of a tie, the general score separated the candidates. This correspondence takes account of title, authors, year, journal and DOI. If Google Scholar found a better candidate, that candidate was taken into account; when the DOI matched a more structured source, normalized metadata from bibliographic databases were preferred for journal, volume and pages.

  • M1 has the best title correspondence. The M1 reference is retained.
  • M2 has the best title correspondence. The Google Scholar reference is retained.
  • Tie between M1 and M2 on title correspondence. The general score separates the candidates.
  • M1 and M2 DOI are identical. Journal, volume and pages may be taken from M1 because these metadata are more structured.

General score

The general score is a weighted average of the elements actually available in the citation. A missing element does not lower the score: its weight is removed from the denominator.

  • Title: 20%. Comparison by significant words and controlled tolerances.
  • Authors: 20%. Positional comparison of cited authors, in the available order.
  • Year: 20%. Exact or close year depending on the case.
  • Journal: 20%. Journal name, with volume/issue if cited.
  • DOI: 20%. Exact correspondence if the DOI is cited.
  • Missing DOI example. (Title 80×0.20 + Authors 100×0.20 + Year 100×0.20 + Journal 100×0.20) / 0.80 = 95.0/100.

Title fidelity: Jaccard index

The Jaccard index compares significant words in the cited title and the title found: shared words divided by all significant words present in both titles.

  • Normalization. Case, accents, punctuation, compound words, singular/plural.
  • Acronyms. ASD or ASC may correspond to Autism Spectrum Disorder(s) / Condition(s).
  • Language variants. behavior and behaviour are treated as equivalent.
  • Overly generic words. autism, spectrum, disorder(s), autistic, ASD, ASC and condition(s) are excluded from the calculation.
  • Incomplete title. A part before or after a colon may be tolerated only if all cited words correspond to the remaining segment.

Special rules and edge cases

  • Cited but incorrect DOI. It does not drive the search. It is compared afterward and may penalize the DOI.
  • Abbreviated journal. Standard acronyms or abbreviations may be accepted: JADD = Journal of Autism and Developmental Disorders.
  • Volume only: 41 vs 41(9). Considered correct if the volume is identical and the issue was not cited.
  • Different volume: 41(2) vs 40(3). Considered divergent.
  • Severe false positive. If no author is shared and the title is below 50% correspondence, the reference is invalidated and the final score is set to 0.
  • Missing element. Not penalized; weight removed from the calculation.
  • Approximate citation. The procedure accepts some variations to verify whether a real and coherent reference exists, while still reporting observed divergences.

Verification and corrections

  • Verification links. Records present the citation source, the selected bibliographic source and a Google Scholar search built from the citation.
  • Reading records. Detailed comparisons show authors, year, title, journal, DOI and the Jaccard index to support manual checking.
  • Correction. Any error or useful correction can be reported through the AutiHub contact page.