Query regarding Object/Subject identification logic:

rock3125 / enhanced-subject-verb-object-extraction

Enhanced Subject Word Object Extraction

Apache License 2.0

148 stars 51 forks source link

Query regarding Object/Subject identification logic: #15

Open vinodhian opened 1 year ago

vinodhian commented 1 year ago

Hi, In the algorithm used,

To find the Subject of a Verb( _get_all_subs() ), it considers only the left children of the verb and to find the Object of a verb ( _get_all_obss() ), it considers only the right children of the verb. Why is that so? Shouldn't we consider both left and right children?
Also, In line 71, inside _findsubs() function, subs = [tok for tok in head.lefts if tok.dep_ == "SUB"] However, there is no dependency relationship called "SUB" in spacy. Am I missing anything in this?

Thanks, Vinodhkumar

rock3125 commented 1 year ago

active English sentences are SVO (Subject on the left of the Verb, then the Object after the verb). So LHS = S and RHS = O. Passive sentences reverse this. This was originally written with Spacy 1.x - so it's well possible that the spacy people changed all that.

vinodhian commented 1 year ago

@rock3125 Thank you. If you could let me know the intention behind checking the "SUB" dependency, I will try to find the equivalent in the current spacy version.

rock3125 commented 1 year ago

I found this list https://github.com/clir/clearnlp-guidelines/blob/master/md/specifications/dependency_labels.md Here "DEP" is an unclassified dependency. Not sure why I picked that ...