Need clarity on Dependency Relations output by Dependency Parser.

[damzC]

Hi,

While using your dependency parser for English, I found that the dependency relations output by the parser are not in CoNLL Universal Dependency Relations format or, Stanford format. Can you please point me to a document/ page where I can find the meaning of these relations.

For eg:

For the sentence: "I want to take a leave today"
The output is:
1 I _ PRP PRP _ 2 SBJ
2 want _ VBP VBP _ 0 ROOT
3 to _ TO TO _ 2 OPRD
4 take _ VB VB _ 3 IM
5 a _ DT DT _ 6 NMOD
6 leave _ NN NN _ 4 OBJ
7 today _ NN NN _ 4 TMP

I cannot find what the relations 'IM', 'OPRD' mean, anywhere. The same sentence output by the Stanford dependency parser is:

1 I I PRP O 2 nsubj
2 want want VBP O 0 ROOT
3 to to TO O 4 mark
4 take take VB O 2 xcomp
5 a a DT O 6 det
6 leave leave NN O 4 dobj
7 today today NN DATE 4 nmod:tmod

Can you please help me in finding the meaning of these dependency relations.

Thanks in advance.

Arindam

[erickrf]

Hi @damzC
The English dependency model was trained on a version of the Penn Treebank converted with the LTH tool. I couldn't find any organized description of all the labels, since the Penn Treebank was originally annotated with constituent trees, but in this page you can find a good description.