parametrize error correction

[andreahorbach]

so that one can choose how the best replacement candidate for a misspelling is selected

• strictly based on Levenshtein distance
• prefer domain material
• etc.

[zesch]

Do we really do that in Escrito itself? We might first need a spelling tool that can be parametrized in those ways (stand-alone project) which is then used here.

[andreahorbach]

We have normalization code that uses the escrito readers corrects spelling mistakes and writes the output, which is the used as new input in a core escrite process. So normalization is not a direct part of the escrito pipeline, but escrito provides ways for spellchecking the data.

[zesch]

ok, my suggestion would then be to allow Escrito to use a spell checker and keep most of the parametrization with the spell checker. Escrito only needs to decide where the spell checker is applied.