cfmrp / mtool

Software to Manipulate Different Flavors of Semantic Graphs
http://mrp.nlpl.eu
GNU Lesser General Public License v3.0
51 stars 24 forks source link

Mismatch in UCCA remote edge data format #70

Closed longxudou closed 5 years ago

longxudou commented 5 years ago

In CoNLL2019 dataset, the remote edge was tagged with Edge Properties and set it value as true , while it was called Edge Attributes in mrp-website. However, when I try to convert the UCCA_English-Wiki dataset into mrp format using mtool. I find the remote edge was tagged with Edge Attributes.

That is to say, there is a format mismatch in remote edge data format.

This might due to lines 92~94 in mtool/ucca.py .

attributes, values = None, None;
if edge.attrib.get("remote"):
attributes = ["remote"];
values = [True];
danielhers commented 5 years ago

Hi @DreamerDeo, the latest format specifies edge have attributes while nodes have properties, to make a clearer distinction. While older data specifies edge properties, mtool reads them as edge attributes when decoding the json: https://github.com/cfmrp/mtool/blob/master/graph.py#L336

longxudou commented 5 years ago

@danielhers Got it, thank u!