Some universal relations do not appear

[nschneid]

The following supposedly universal relations do not appear in the Hebrew data:

• compound (see Use of 'name' #5)
• discourse (but there is conj:discourse)
• expl
• foreign
• list
• remnant
• reparandum
• vocative

Is this because no relevant examples happen to appear in the data? Are there no good heuristics to detect them from the source treebanks? Or are these relations excluded on purpose?

[yoavg]

• compound may be added when we resolve Relabel 'nmod:smixut' as 'compound'? #6
• It is not clear to me from the documentation what discourse stands for, but I am pretty sure it is different from our use of conj:discourse
• expl I do not think we have these in the data. Perhaps יש and אין can be treated this way, but we are already discussing those.. Treatment of existential/possessive יש and אין #8
• foreign, list we have no good way of identifying these
• remnant initially we left out as hard to identify, now I'd wait for the decision on replacement for v2
• reparandum I am pretty sure the data does not have these (or maybe just one or two examples)
• vocative there is no reliable way to identify, but we admittedly did not try very hard

[nschneid]

• It is not clear to me from the documentation what discourse stands for, but I am pretty sure it is different from our use of conj:discourse

I think you're right. At least in English, discourse is used for "please", "yes"/"no", interjections, and emoticons. I see 3 tokens tagged as INTJ in the Hebrew corpus—should these be linked with discourse? Currently, 2 of them are advmod and one is dep.