For each target bible, we automatically choose an English bible that aligns well to make the alignment more meaningful.
Motivation: improve alignment -> reduce FPs -> increase DC precision > 0.30
Tasks
[ ] add code to try out all available English translations for a given LRL translation and find min perplexity / cross entropy
does it contain many verses (also apocrypha)?
look if alignments have a low perplexity or low cross entropy
e.g., does it mention eng:Euphrates in the same verses as tpi:Yufretis appear?
known seed of words
[ ] evaluate if these metrics correlate with F1 score
Goal
For each target bible, we automatically choose an English bible that aligns well to make the alignment more meaningful. Motivation: improve alignment -> reduce FPs -> increase DC precision > 0.30
Tasks