Zeri Auction Catalogues

Scripts and data of 1.9K auction catalogues from the Zeri photo archive

RDF transformation

Zeri_cataloghi_RDF.ipynb --> zac_catalogues_<date>.trig ; reconciled_agents.csv

RDF transformation of bibliographic data from the Zeri foundation (export from Sebina), stored in a google spreadsheet (link in the notebook).

Automatic reconciliation of auction houses. Outputs are uploaded in a separated tab of the above spreadsheet for human-revision.

TODO:

• finalise human-revision of reconciliation
• regenerate the RDF dataset to add Wikidata links
• revise classes assignment to people / groups (incorrect)
• remove duplicate entities (different forms of same name in the original data generate different URIs)

OCR

transcription.py

Initial attempts with Pixtral, script used directly on the server where images are accessed on the file system

1_ocr.py --> all.md

• Select pages to be parsed (those with lots description) from a given list
• Transform images in greyscale
• Use Docling (ocrmac) to perform OCR of pages.
• Concat all transcriptions of each catalogue into the output file all.md.

The script is run locally on a benchmark group of catalogue images, outputs not included here.

2_chunking.py --> all_chunks.csv ; inconsistencies.csv

• Parses all.md files,
• performs regex to separate lot descriptions and
• concat them into chunks.csv for each catalogue (not included here).
• All chunks are concat in all_chunks.csv, available on the aforementioned spreadsheet for human revision.
• Errors detected in the chunking (mainly based on numbering sequence inconsistencies) are collected in all_inconsistencies.csv, also included in the spreadsheet.

TODO:

• improve chunking docling:
  • e.g. "203 bis."
  • wrong numbers at the beginning of lines (from OCR) that split one lot in two. e.g. "18. bla \n 2.bla \n 19. bla"
• finalise pipeline for pixtral with the same benchmark group of images
• compare outputs of both pipelines