Greek Model number

[Ctaaffe]

"σ΄. ψιμμυθίου δραχ. ρν΄. πηγάνου δραχ. ρ΄. σταφίδος ἀγρίας δραχ. κε΄. ὄξους ξεστ. γ΄. μυρσίνου λίτρας γ΄. τρῖβε κατὰ μέρος ἐπιβάλλων τὸ ὑγρὸν καὶ συντίθει"

Using this recipe from Galen that includes Greek letters as numerals; the numbers are marked with the "΄" and are lemmatized as words and the "΄" is "." any fix that would allow pie-extended to recognize the numbers without having to retrain the entire model. Thank you for any answer.

[PonteIneptique]

This is probably possible, but soneone would need to come up with an Ancien Greek numeral regular expression that I would then have to inject in the tokenizer, thus fixing the lemmatization.

[Ctaaffe]

I could do that, do you want the expression directly in pyhton or in an other format ?

[PonteIneptique]

I'd need it in python to capture any numeral, such as

nlp-pie-taggers/pie_extended/pipeline/tokenizers/utils/regexps.py

Line 4 in ae081a4

RomanNumbers = r"(?:M{1,4}(?:CM|CD|D?C{0,3})(?:XC|XL|L?X{0,3})" \

:)

[PonteIneptique]

I can't promise to include it before early February though.

[Ctaaffe]

Thank you, for you answer no problem I'll see when I'll have the time to write it and use it on my own install of pie-extended in the meantime.

[etymologika]

Bonjour à tous deux, voici les deux règles pour qu'une lettre ait valeur de chiffre en grec:

1. Une seule lettre ou un groupe de plusieurs lettres de la liste suivante, suivies du signe ʹ (apostrophe légèrement oblique, dénommée dexia keraia, code Unicode 0374), forment un chiffre (ex.: αʹ =1, ιϛʹ =16, κδʹ = 24).
   α
   β
   γ
   δ
   ε
   ϛ
   ζ
   η
   θ
   ι
   κ
   λ
   μ
   ν
   ξ
   ο
   π
   ϟ
   ρ
   σ
   τ
   υ
   φ
   χ
   ψ
   ω
   ϡ
2. Une seule lettre ou un groupe de lettres peut être précédé ou interrompu par le signe ͵ (aristeri keraia, code Unicode 0375), avec ou sans signe ʹ (dexia keraia) à la fin, exemples: ͵α =1000͵ α͵ροστ΄ = 1176.
   Grand merci d'avance!