Add empty nodes to collect disonnected graphs?

interactive-cookbook / tagger-parser

Tagger and parser models used on our recipes corpus (data), handled with pre- and postprocessing scripts for data conversion (data-conversions)

0 stars 3 forks source link

Add empty nodes to collect disonnected graphs? #17

Open TheresaSchmidt opened 2 years ago

TheresaSchmidt commented 2 years ago

Problem described in alignment-models/issues/7: When converting recipe graphs, into action graphs, nodes or sub-graphs can get cut off creating disconnected action graphs.

Proposed solution: Add one empty head node to each recipe graph, s.t. from all other nodes, there is a path to this head node and this node will be preserved in all conversions such as conversion to action graphs or FAT graphs.

TheresaSchmidt commented 2 years ago

Open questions:

new code as option in json_to_conllu.py or as separate post-processing script?
what label should the new node have? Action? New label?

Note:

also need to update crowdsourcing scripts with the new data format
don't forget to mark as new version

TheresaSchmidt commented 2 years ago

Other solution: Assuming that, if the sub-graphs are disconnected, the respective parts of the recipe could be executed independently, i.e. one after the other. Therefore we could connect the sub-graphs sequentially. Contra: doesn't preserve the information where the graphs were disonnected originally.