Hi,
As for the sentences in question, here are some translations that co-organizer Niko Partanen made off the cuff: you can take them as a starting point if useful:
1) "Ja Arwatte sen swren kwlutuxen ia waijwan, quin Mee kaickeden teiden hyffuexenne pitänehet olemma."
"and you can guess that great bearing and effort, what we for the good of all of you have kept"
2) "Pacanat muinen – – ychden pellcurin ja jänixensydämmällisen Sota miehen, owat heitin pitänyt nin cuin idze häpiän"
"The other (?) pagans – – one coward and rabbithearted soldier, they have reacted to them as I have myself been ashamed (or maybe 'as the shame itself'?)"
3) "[Niniven] Turun Cartana piti, 400. Stadia "
"(Niniven ?) as the map of Turku was kept 400. (what is Stadia?)"
Comment: There is a longer fragment elsewhere that says "[Niniven] Turun Cartana piti, 400. Stadia iotca. 12. Somen penicwlemata tekeuet", so the idea could be that the map of something was 400 units of something, that were 12 units in Finnish miles or something.
4) "[Löydetyllä] Heuoisella, Tammalla, ia häriällä, mahta mies työtä tehdä, ia asioitans aia"
"[with found] horse, mare, and ox, may man work do, and drive his things"
5) "Aja myöskin pois minun Sielustani caicki Ylpeys – – eten minä luotais minun Lahjani päälle"
"drive also from my soul all proudness – – so that I would not rely to the gift I have received.
A good thing to keep in mind is that the resource is in Old Literary Finnish, not in Finnish—much like an English system will struggle with Chaucer, we have very little expectations that tools developed for modern Finnish will work on Old Literary Finnish. One major point of difference concerns orthography: for instance, modern Finnish never uses the letter C W and X (outside of foreign names and loanwords from other languages), whereas these letters were frequently used in Old Literary Finnish.
Internally, we used the following code snippet for orthographic normalization:
def normalise(word):
normalised = []
old = 'wcqx'
for char in word:
if char in old:
if char == 'w':
normalised.append('v')
elif char == 'c' or char == 'q':
normalised.append('k')
elif char == 'x':
normalised.append('ks')
else:
normalised.append(char)
return ''.join(normalised)
As for punctuation signs: square brackets and double dashes are found as such in the base data we used to construct the Finnish segment of the AXOLOTL dataset. You are of course welcome to preprocess the dataset in any way you see fit; we elected to provide participants with the actual contents of the base resource we use (rather than second guessing the choices of the original lexicographers).
Hopefully this answers your questions!
Best regards
Timothee Mickus