Customers often upload Excel files that are visually formatted for human readability rather than structured as flat datasets.
Example issue:
  • Multiple sub-tables arranged in one sheet.
  • Merged cells used for headers such as duration or calorie tiers.
  • Repeated columns with different meanings in separate sections.
  • Visual relationships between cells are clear to humans but not machine-readable.
Current AI Knowledge Base parsing struggles to accurately interpret these files, causing incorrect retrieval of pricing or other structured information.
It would be great to enable AI/Knowledge Base to better understand “visually connected” Excel layouts by:
  • Detecting repeated table regions within a sheet.
  • Interpreting merged cells as contextual metadata.
  • Mapping headers hierarchically.
  • Inferring relationships between visually grouped data.
This would reduce manual preprocessing and improve usability for customers uploading pricing lists, menus, catalogs, and similar spreadsheets.