[HN Gopher] Automated reading of medieval manuscripts: Alternati...
       ___________________________________________________________________
        
       Automated reading of medieval manuscripts: Alternative for
       palaeography classes?
        
       Author : danso
       Score  : 24 points
       Date   : 2022-08-16 15:02 UTC (1 days ago)
        
 (HTM) web link (www.leidenmedievalistsblog.nl)
 (TXT) w3m dump (www.leidenmedievalistsblog.nl)
        
       | tarl0s wrote:
       | The Roma Tre University has a research project named In Codice
       | Ratio. One of its objectives is to transcribe through AI and OCR
       | the whole Vatican Secret Archives - one of the biggest collection
       | of manuscripts, some of them more than a thousand years old.
       | 
       | The code hasn't been released (yet) but you can find some
       | preliminary results here:
       | http://www.inf.uniroma3.it/db/icr/preliminary-results.html
        
       | jofer wrote:
       | I kept reading that as "paleogeography", which is a common
       | geological term. (i.e. reconstructing ancient landscapes)
       | 
       | Weird how the brain jumps to "term I know" rather than actually
       | reading a word.
        
       | azangru wrote:
       | Curious about the scope of the Digital Editing course mentioned
       | in the article, but the course page is in Dutch :-(
        
       | avyeed_desa wrote:
       | I like Transkribus a lot and it is extremely helpful to get quick
       | transcriptions, especially when the models are trained well. It
       | will never get to 100%, but manual intervention is always needed.
       | And Transkribus is a really, realls well-thought out piece of
       | software, even though its heavy dependencies on Java make it
       | slow, especially on 50+ page documents.
       | 
       | However, i never liked their move from a research project to a
       | commercial model. Their signup has plenty of credits for an
       | individual who just wants to edit their family documents, but i
       | still think it should be a bit more lenient for personal use.
       | 
       | Thankfully there is eScriptorium. Even if it is still in early
       | development it is a more user-friendly alternative to
       | Transkribus. https://gitlab.com/scripta/escriptorium
        
         | wjnc wrote:
         | I came to comment on a few things:
         | 
         | 1. Indeed, a publicly funded European research project turned
         | commercial software (closed source?) that expects institutions
         | to pay for annual fees. Hmm. I understand OSS still needs
         | constributors and has ongoing maintenance costs, but couldn't
         | there be a more efficient way? It had a very German academic
         | feel to me (nofi, and indeed it's an endeavor started at 4
         | German universities.)
         | 
         | 2. The blogpost almost reads like a nineties description of the
         | value of IT. (Fun read and perspective though! This is the
         | positivist approach to history that underpins many interpretive
         | histories of the future. Great and underestimated work.) The
         | whole point of computer and user augmenting each other
         | continuously somewhat falls short with the author saying how
         | impressive, but fallible students and computers are. Along the
         | lines of "okay, the output is x 1000 and of pretty good
         | quality, but it's not professional academic quality". When I
         | think of chess, or poker: computers have given people /new ways
         | of studying/ even before applying. I think that point is still
         | missed here. The software should point out mistakes by the
         | students in training, while it learns by the additional input.
         | That is the virtuous cycle of continuous improvement.
         | 
         | And 3. Things like the scientific R and Python ecosystems, or
         | like Stan have shown me the power of creating open source tools
         | for other use. Like Andrew Gelman, who has remarked multiple
         | times that he never could have expected the use cases Stan has
         | now. (There are Bayesian sport scientist now..!) Teach people
         | and give them tools, but don't dictate the entire workflow.
         | Please let outsiders have a chance of swapping models, doing
         | proof of concepts etc.
        
       ___________________________________________________________________
       (page generated 2022-08-17 23:01 UTC)