search

ufal / treex

Treex NLP framework
33 stars 6 forks source link

get_iset_conll_feat returns other=HASH(0x30ae3c40) #11

Closed martinpopel closed 8 years ago

martinpopel commented 9 years ago

Write::CoNLLX uses $anode->get_iset_conll_feat to print the feat column (if feat_attribute eq 'iset'). Interset can contain feature other which is a hash reference and the current code prints the hash reference address, e.g. 'HASH(0x30ae3c40)', which is probably not the intended behavior.

Simple fix would be to skip other and tagset feature (as it is done in the get_iset_values method). I think a better fix would be to delete these methods completely and use $anode->iset->some_serialization_method. Unfortunately, I don't see any suitable method in Lingua::Interset::FeatureStructure. If it is missing on purpose, I'll do the 'simple fix' above.

dan-zeman commented 9 years ago

Thanks for reporting, this is a legacy of Interset 1 being once used in Treex. Interset 2.047 will have two new methods that can replace some code in Treex::Core::Node::Interset. And it should fix the bug as well. I am now adapting the Treex code in a separate branch.