hltfbk / Excitement-Open-Platform

Excitement Open Platform for Recognizing Textual Entailments
http://hltfbk.github.io/Excitement-Open-Platform/
86 stars 74 forks source link

[distsim] LIN-DEP (seems to) generate only ADJ rules (German, CONLL format) #292

Open gilnoh opened 10 years ago

gilnoh commented 10 years ago

After successful generation and redis-conversion; the lexical resource based on Lin dependency works for the German.

However, the resource has no (or almost no) nouns, or verbs. I tried with various common terms that shoud be existing; but couldn't generate any match.

For the moment, I do not know how I can iterate over all rules (or all entries), so this is just my guess: But it is quite likely that LIN-DEP resource generated with existing configuration only has ADJs. (Not even ADVs or Vs, it seems...)

Again, I can be wrong, since I don't know how to iterate over all the rules / elmements. Is this normal? (I guess not).

Also, ADJs were quite ... strange. For example, (much more common) ADJs like gut /schlete (good/ bad) does not existing, while some ADJs like recht (right) are there, etc ...

gilnoh commented 10 years ago

You can reproduce this with the following intermediate size corpus: (1/30th of SDEWAC)
http://www.cl.uni-heidelberg.de/~noh/sdewac_part01.mstparsed.utf8.conll.gz