Maybe we can create a tool to allow interested wikisource editors to match text to bounding boxes, which is required for Transcribus (and other OCR software) to process text and build models.