This dataset is part of the IA Gwened Project, an independent research and development initiative focused on Breton language technologies, with special attention to the Vannetais (Gwenedeg) dialect. This dataset provides a parallel corpus derived from DIHUNAMB, a Breton periodical directed by Loeiz Herrieu, issue no. 01 published in 1905. The original texts are written in Breton (Vannetais dialect) and have been aligned with their French equivalents. The corpus has been transcribed, normalized, segmented, and sentence-aligned to ensure compatibility with linguistic research workflows, natural language processing tools, and machine translation frameworks such as Apertium. The dataset includes three complementary versions of the corpus:– an original version, preserving the errors and specific features of the historical source;– an annotated version, documenting these errors and particularities;– a normalized version, in which the errors have been corrected to produce a clean and consistent text. The intellectual responsibility for the original historical content lies with the original editor of the periodical. The present dataset focuses exclusively on the digital processing, corpus structuring, and alignment of the source material.