inconsistency in the superscript

[arademaker]

See the superscript in the civil versus signal. One has a only and the other a2??!

[arademaker]

It looks like the case with only a is caused by the sensekey used be not in the list of possible lemmas for the token. But why that happens in the data?

((kind "wf")
   (form . "ordinary")
   (lemmas "ordinary%1" "ordinary%3")
   (tag . "man")
   (senses "ordinary%5:00:02:common:01")
   (meta
    (pos . "JJ")))

[odanoburu]

at least on my machine, where I'm using WN data from openWN, the problem is a small incompatibility between the former's sense keys and PWN's sense keys: the fact that openWN has split the adj.all lexicographer file into adj.all and adjs.all. because of this the already annotated sense of ordinary does not exist in the database, and therefore has no number. because we have already decided to undo this split of adj.all, the problem should disappear by itself in the next commits.

I'll make sensetion complain when it finds an unknown sense key, however.

[arademaker]

What is openWN? Do you mean the https://github.com/own-pt/own-en? BTW, does it means that we will have to map all previous annotations in the glosstag corpus to the new sense keys? Another possibility would be to keep glosstag annotation related to the PWN 3.0 sensekeys. Not sure if I understood the implications of your description above.

[odanoburu]

Do you mean the https://github.com/own-pt/own-en?

yes

BTW, does it means that we will have to map all previous annotations in the glosstag corpus to the new sense keys? Another possibility would be to keep glosstag annotation related to the PWN 3.0 sensekeys.

one of those, yes. we should probably discuss this better, since we're not sure how stable our new id scheme is.

Not sure if I understood the implications of your description above.

the split into adj.all and adjs.all created a new lexicographer file; since all sense keys include the lexicographer file number, all satellite adjectives that have been pre-annotated are not recognized by sensetion (but the sense keys are there and are not eliminated; duplicate annotations may happen, however).