> Not to be negative about the research, but I’m pretty sure I could have surmised the pigs’ emotional state in many of these situations (castration?!) without the benefit of a translator!
That's the point. The obvious emotional states are necessary in order to correctly label the data for training. The interesting part is to then use the trained model to identify emotional states in situations where it isn't so obvious.
Garbage in garbage out. How do we know the emotional states are what they are? Better to use as label something objectively measurable but costly to observe like some internal brain or chemical state of the pigs.
How about just named after the situations they were observed in? Then when you give it a new pig sound, the model can classify it and we can decide how to respond to "slaughterhouse=0.8, castration=0.2" or whatever.
I'm especially suspicious of fighting being negative. Humans can get pleasure from fighting. Why not pigs?
That's the point. The obvious emotional states are necessary in order to correctly label the data for training. The interesting part is to then use the trained model to identify emotional states in situations where it isn't so obvious.