This is for generating (shallow and heuristic) links across different datasets so that they can be queried in conjunction with each other.
Link and test using
$> make
- For some morphemes, the linking algorithm overgenerates and produces links without matches. This may be an issue with blank nodes.
- For language identification, UniMorph uses ISO639-3 tags and the others use BCP47 tags. For German, we just match the first two characters, for other languages, this heuristic is either not applicable (Spanish:
spa
vs.es
) or produces incorrect results (rom
Romani vs.ro
Romanian).