Replied inline... On 24 January 2018 at 15:07, Edward Garrett <[email protected]> wrote: > > can you elaborate on this? while everything you said from the CG-3 side > makes sense to me, just looking at the CoNLL-U format ( > http://universaldependencies.org/format.html), it seems like one could > make a handy converter which just made some choices about how to map the > CoNLL-U columns into CG-3 tags. for example, features could be mapped to > Name=Value, i.e. Case=Num, and so on. i don't know whether this would be a > "general purpose converter" but if it took any CoNLL-U input and produced a > plausible CG-3 output, then it would be useful in my book. >
Sure, from CoNNL-U to CG-3 is easy, though yielding an ugly result for the general case. I assumed the original question was the other way. > as for going in the reverse direction, i agree it would be problematic. i > suppose since the order of the tags and their internal composition is > irrelevant in CG-3, the tags might easily get swapped around and screw > things up. that said, perhaps there could be some guidelines: if you name > your tags in such and such a way, and choose from this inventory of tag > naming conventions, then you can convert your CG-3 file to CoNLL-U using > this script. > This is where nobody can agree. VISL uses ALLUPPER for part of speech, <> for secondary. Apertium has no patterns. Giellatekno has several prefixes that denote secondary tags, with the remainder being primary. Etc. The only thing everyone agrees on is using prefix @ for primary mapping tags, because that was enforced by older CG implementations. So naming conventions are out the window - has to be list based. Those lists could be in the grammar, or supplied to cg-conv somehow. -- Tino Didriksen -- You received this message because you are subscribed to the Google Groups "Constraint Grammar" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To post to this group, send email to [email protected]. Visit this group at https://groups.google.com/group/constraint-grammar. For more options, visit https://groups.google.com/d/optout.
