Recently in this data set are about 280 of total 1.000 pages affected.
The situation is a great deal related to correction attempts of Transkribus output concerning geometrical operations like merging or splitting lines and is subject to re-correction.
Description
As already discovered by @stweil (Fix two issues with textlines in PAGE XML files and Find and fix systematic transcription and data issues), there are differences between text on line-level an the corresponding words.
Recently in this data set are about 280 of total 1.000 pages affected.
The situation is a great deal related to correction attempts of Transkribus output concerning geometrical operations like merging or splitting lines and is subject to re-correction.