Oh, ELAN has an interlinearization mode now… anyone tried it?

Stumbled across this this morning:


I’m curious if anyone here has tried it. If so, what did you think?

Just spent like 45 minutes trying to gloss ten words. Did not succeed.

Glad to see this. I need to try it now.

1 Like

I’ve used it with some minor success. As you might expect, it isn’t as good as FLEx’s interlinearization capabilities. Some limitations I encountered:

  • Changes to the lexicon are not implemented in the EAF data created using that lexicon, so you need to manually implement each change you make (e.g. a new gloss) across all of the files in your corpus.
  • Tonal constructions aren’t recognized and tonal data isn’t considered separately from segmental data (not fully supported in FLEx either, AFAIK). This was a big problem for me actually and resulted in multiple entries for the same morpheme with different tone.
  • The ELAN interlinearization system “learns” which potential interlinearizations are most likely to be correct, but still isn’t very good at it. It can produce a very long list of possible matches/permutations, and the correct one is sometimes hidden somewhere in the middle. It doesn’t appear to be sensitive to the lexical class of the root in determining the probability for a given affix (e.g. a nominal suffix never attaches to a verb).

IMO FLEx is still better for the interlinearization of large datasets and ELAN is still better for time-aligned data. You can use them together, of course…