-
Ok thank you for your comment. I think I did introduce the feature through a macro, but seemingly the <GenRel> features inherits other features besides pred. I'll check better, anyhow...
2009-09-01 13:27:20 UTC in OpenCCG: The OpenNLP CCG Library
-
The question of how to cope with Arabic writing is indeed interesting. Some good result has been obtained by statistical tools, but several alternatives might be envisaged.
In openccg, the best solution could be to hack the parser in order to let it accept not only the exact form present in the lexicon, but also any subset of it lacking of one or ore more optional diacritics. To simplify...
2009-09-01 13:24:45 UTC in OpenCCG: The OpenNLP CCG Library
-
in a toy grammar I am trying to implement a family of "adjective nouns" like 'gold' in 'a ring in gold', which have such a family description:
def adj_noun() {
n<~2> [X] \* n<2> [X] \* p[lex=in] : X(<GenRel> *)
}
the problem is that the GenRel feature inherits (by virtue of the default '*' mark) all semantic features from the head noun 'gold', including the...
2009-08-25 12:41:42 UTC in OpenCCG: The OpenNLP CCG Library
-
Thank you for your answer. In fact, I tried the option you suggest, but the problem is that in written Arabic you can have any situation between no diacritics and no diacritics, so you'd need to generate 2n entries, where n is the number of diacritics in the "full" written form.
As to dotccg, I'll try to get modality variables work by hacking ccg2xml (in fact, working on the .ply...
2009-08-24 07:45:39 UTC in OpenCCG: The OpenNLP CCG Library
-
Dear Mike,
thank you for your answer. In fact my aim was not much to criticize the documentation, but rather to stress how wider and more comprehensive the package is than what the documentation suggests (if this sentence is acceptable in English, which I strongly doubt)...
Some of the features you mentioned in your answer could in fact be interesting even for people not interested in...
2009-08-15 13:15:25 UTC in OpenCCG: The OpenNLP CCG Library
-
I think openccg is the best public domain computational linguistic package available, especially because _it works_. The main problem I'm finding in using it is the lack of thorough documentation; many useful pieces of information can be gathered from the rough guide, tutorials on the wiki and comments in sample grammars, but a lot more can only be guessed or found by trial and error. So, I'm...
2009-06-08 08:09:03 UTC in OpenCCG: The OpenNLP CCG Library
