Figure 1
Temporal distribution of the 10,898 novels. Production peaks in the 1880s, reflecting the golden age of the roman-feuilleton. The drop after 1900 is partly an artefact of Gallica’s digitization coverage and of copyright constraints on twentieth-century works.
Figure A.1
Precision (dashed) and recall (dotted) of MinHash-based duplicate detection as a function of the Jaccard threshold, with 95% bootstrap confidence intervals (shaded). The hybrid pipeline used for the released corpus combines this content-level signal with the rule-based metadata signal described in Section 2, which is why the operational regime sits at a precision level higher than the curve indicates for MinHash alone.