The bulk update request #11450 is pending approval.
create implication sitting_on_swing (93) -> sitting (348274)
create implication sitting_on_swing (93) -> on_swing (103)
create implication standing_on_swing (3) -> standing (487664)
create implication standing_on_swing (3) -> on_swing (103)
create implication on_swing (103) -> swing (1129)
create implication sitting_on_toilet (561) -> sitting (348274)
create implication sitting_on_toilet (561) -> on_toilet (344)
create implication standing_on_toilet (3) -> standing (487664)
create implication standing_on_toilet (3) -> on_toilet (344)
create implication on_toilet (344) -> toilet (6995)
create implication sitting_on_vehicle (233) -> sitting (348274)
create implication sitting_on_vehicle (233) -> on_vehicle (143)
create implication standing_on_vehicle (7) -> standing (487664)
create implication standing_on_vehicle (7) -> on_vehicle (143)
create implication on_vehicle (143) -> vehicle (55183)
create implication sitting_on_car (229) -> sitting_on_vehicle (233)
create implication sitting_on_car (229) -> on_car (297)
create implication standing_on_car (4) -> standing_on_vehicle (7)
create implication standing_on_car (4) -> on_car (297)
create implication on_car (297) -> car (16329)
create implication on_car (297) -> on_vehicle (143)
create implication standing_on_chair (73) -> standing (487664)
Reason: Okay, this one is a doozy.
The first batch of these implications is for sitting and standing on inanimate objects. The first BUR is a test case pending feedback on whether this is the right approach or overly specific.
"Why imply on_thing instead of thing directly?"
The term on_thing has value in being a catch-all for actions on top of an object. If a verb is missed (such as crouching, kneeling, etc.), then the tagger can at least state the position of the character as on top of the thing. Without the intermediary tag, a tagger would not be able to say a character is on top of something if the verb is missing.
"Why have on_thing at all?"
There is a valid argument to be made that the on_* tags should just be aliased to the object and that actions should be split. You could have a mostly equivalent search between sitting_on_swing, sitting on_swing, and sitting swing. However, aliasing away on_swing would mean that any action that isn't specified would require a new tag, including obscure poses or characters that have atypical bodies that can't perform searchable actions like sitting, standing, or lying. This includes flowing goo, masses of tentacles, etc. - how would you describe a character like this being on top of the object otherwise?
"Where is lying_on_*?"
The lying_on_* tags are trickier since a) some objects are too small or too awkward to lie on, b) it's rarer to lie on top of some things and may not be worth considering (eg. lying_on_swing is improbable and hasn't been tagged before ), and c) any lying_on tag has an argument to be made that it requires a separate laying_on alias since the words are used interchangeably. There is added complexity here that I do not want to approach until users discuss just the more basic actions like sitting and standing.
"If you imply sitting_on_car to sitting_on_vehicle, then sitting_on_vehicle to on_vehicle, isn't on_car implying on_vehicle redundant?"
This creates a redundant implication only if the user tags any of the *_on_car tags. However, it is necessary to include if the user tags on_car directly or it won't be implied.
Please leave feedback on this. I do not mind if people say "no, this is a bad idea", but I would like to hear specifically why you would be for or against these, as well as alternative options that would be cleaner. Note that there is precedent for these types of implication chains with sitting_on_bench and lying_on_bed, but precedent shouldn't deter you from speaking your mind if you disagree with the structure.
Updated