So far, the software seems to have done a good job in working with the Scots language, where spelling is non-uniform. As I mentioned in a separate post, it struggled with abbreviations, which makes me think that the issue with abbreviations might differ by language or region. However, in general it did not auto correct the spelling which other software I have used has done in the background.
Would there be a way to train models specific to languages? So I could select that my document is written in English/Scots from x date and the AI will then pull its modeling from locations like the Dictionaries of the Scots Language - https://dsl.ac.uk/
1 Like
Leo’s philosophy is to have an all-in-one model for languages. Unfortunately, it’s not straightforward to rely e.g. on dictionaries to guide spelling recognition. The problem is just that it hasn’t seen enough material in Scots yet. Hopefully as more historians of Scotland start using the app, that will change!