The "a torch" part of this output comes from an ExpressSurroundings concept. This is created by `Container`, in the implementation of `conceptualize` inherited from `Containment`.
`Container` is shared between locations and actors, so fixing this isn't a simple matter of making `ExpressSurroundings` stick in the non-semantic "He is carrying" prose, nor even making `Containment.conceptualize` create a different concept. It seems like the solution will need to involve teaching some part of the system about the difference between containment for an actor and containment for a location (and potentially the other kinds of containment - for example, is a box like or dislike a location?).
The "a torch" part of this output comes from an ExpressSurroundings concept. This is created by `Container`, in the implementation of `conceptualize` inherited from `Containment`.
`Container` is shared between locations and actors, so fixing this isn't a simple matter of making `ExpressSurroun dings` stick in the non-semantic "He is carrying" prose, nor even making `Containment. conceptualize` create a different concept. It seems like the solution will need to involve teaching some part of the system about the difference between containment for an actor and containment for a location (and potentially the other kinds of containment - for example, is a box like or dislike a location?).