`wh` sentences not dealt with at all

[saraswat]

Arguably, handling wh sentences (who / what / when / where) is out of scope for a project that aims to discover the propositional content of sentences. Hence this "issue" really just documents the current behavior of the system, which is a bit unusual.

(Generated by -t -c --dep --oie, -c is my extension that produces the json logical form, it is just a deterministic rewrite of the dependency structure produced by props.)

Who was Jim married to ?
1	Who	_	WP	WP	_	2	nsubj	_	_
2	was	_	VBD	VBD	_	0	root	_	_
3	Jim	_	NNP	NNP	_	2	dep	_	_
4	married	_	JJ	JJ	_	2	dep	_	_
5	to	_	TO	TO	_	4	prep	_	_
6	?	_	.	.	_	2	punct	_	_

1	1,Who	WP	0	0	sameAs_arg,7
3	3,Jim	NNP	0	0	sameAs_arg,7
4	4,married	JJ	0	0	dep,7
5	5,to	TO	0	0	prep,4
7	2,SameAs	VBD	1	1	

{'lf': {'dep': {'prep': 'to', 'word': 'married'},
        'pred': 'SameAs',
        'sameAs_arg': ['Who', 'Jim']},
 'sentence': 'Who was Jim married to ?'}

SameAs:(sameAs_arg:Who , sameAs_arg:Jim , dep:married to )

Why was he killed ?
1	Why	_	WRB	WRB	_	4	advmod	_	_
2	was	_	VBD	VBD	_	4	auxpass	_	_
3	he	_	PRP	PRP	_	4	nsubjpass	_	_
4	killed	_	VBN	VBN	_	0	root	_	_
5	?	_	.	.	_	4	punct	_	_

8	4,killed	VBN	1	1	
9	1,Why	WRB	0	0	mod,8
11	3,he	PRP	0	0	obj,8

{'lf': {'mod': 'Why', 'obj': 'he', 'pred': 'killed'},
 'sentence': 'Why was he killed ?'}

killed:(obj:he , mod:Why )

When did he die ?
1	When	_	WRB	WRB	_	4	dep	_	_
2	did	_	VBD	VBD	_	4	aux	_	_
3	he	_	PRP	PRP	_	4	nsubj	_	_
4	die	_	VB	VB	_	0	root	_	_
5	?	_	.	.	_	4	punct	_	_

14	4,die	VB	1	1	
15	1,When	WRB	0	0	dep,14
17	3,he	PRP	0	0	subj,14

{'lf': {'dep': 'When', 'pred': 'die', 'subj': 'he'},
 'sentence': 'When did he die ?'}
die:(subj:he , dep:When )

Who killed him ?
1	Who	_	WP	WP	_	2	dep	_	_
2	killed	_	VBD	VBD	_	0	root	_	_
3	him	_	PRP	PRP	_	2	dobj	_	_
4	?	_	.	.	_	2	punct	_	_

20	2,killed	VBD	1	1	
21	1,Who	WP	0	0	dep,20
23	3,him	PRP	0	0	dobj,20

{'lf': {'dep': 'Who', 'dobj': 'him', 'pred': 'killed'},
 'sentence': 'Who killed him ?'}

killed:(dobj:him , dep:Who )

Where did he drink the poison ?
1	Where	_	WRB	WRB	_	4	advmod	_	_
2	did	_	VBD	VBD	_	4	aux	_	_
3	he	_	PRP	PRP	_	4	nsubj	_	_
4	drink	_	VB	VB	_	0	root	_	_
5	the	_	DT	DT	_	6	det	_	_
6	poison	_	NN	NN	_	4	dobj	_	_
7	?	_	.	.	_	4	punct	_	_

25	4,drink	VB	1	1	
26	1,Where	WRB	0	0	mod,25
28	3,he	PRP	0	0	subj,25
30	6,poison	NN	0	0	dobj,25

{'lf': {'dobj': 'poison', 'mod': 'Where', 'pred': 'drink', 'subj': 'he'},
 'sentence': 'Where did he drink the poison ?'}
drink:(subj:he , dobj:the poison , mod:Where )

Who is he ?
1	Who	_	WP	WP	_	2	dep	_	_
2	is	_	VBZ	VBZ	_	0	root	_	_
3	he	_	PRP	PRP	_	2	nsubj	_	_
4	?	_	.	.	_	2	punct	_	_

34	1,Who	VBZ	1	1	
36	3,he	PRP	0	0	prop_of,34

{'lf': {'pred': 'Who', 'prop_of': 'he'}, 'sentence': 'Who is he ?'}

Who:(prop_of:he )

[gabrielStanovsky]

We indeed left these constructions out of scope, and it's interesting to see what the output for these end up looking like. I suspect that this output is due to the dependency mark relation appearing in both conditionals and wh sentences.