ALTO 3.x support?

[aploshay]

It looks like this is currently using ALTO 2.0, but my tesseract is generating ALTO 3.x -- does ALTO 3 work with all the newspaper_works requirements, or are any changes required?

[ebenenglish]

Hi @aploshay, can you provide a bit more context about your use case?

Are you trying to use NewspaperWorks to ingest newspapers content that already has been OCR'd via tesseract?

[aploshay]

Hi @ebenenglish, sorry for the belated reply:

In the ESSI project, we're using NewspaperWorks components to generate word boundaries to support text search within a scanned document, and do want to support newspapers that have already been OCR'ed via tesseract. My local tesseract produces ALTO 3 output, which seems to work fine (at least for parsing word boundaries), but since it's a different major version I didn't know what might break or if there were reasons to hold back to 2.x.

[ebenenglish]

Hi @aploshay, sorry for the delay in responding. We haven't done any testing with ALTO 3 output on our end, or any research about the major differences.

But it sounds like ALTO 3 is similar enough to ALTO 2 that our current conversion process is working OK, which is encouraging! Though now I see that ALTO 4.1 is the current version...

https://www.loc.gov/standards/alto/

Our target of ALTO 2 is based on what LofC is mandating in their digitization specs for the NDNP program. Even recently uploaded batches from 2019 seem to still be using ALTO 2 (ex: https://chroniclingamerica.loc.gov/data/batches/arhi_electabuzz_ver01/), and the NDNP Digitization Guidelines are still mandating ALTO 2 (see page 9 here: https://www.loc.gov/ndnp/guidelines/NDNP_201921TechNotes.pdf).