Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

More info needed for solving ambiguous requests #11

Open
laetitia-teo opened this issue Jun 14, 2022 · 1 comment
Open

More info needed for solving ambiguous requests #11

laetitia-teo opened this issue Jun 14, 2022 · 1 comment

Comments

@laetitia-teo
Copy link

Hi !

I'm using ScienceWorld in interactive mode, and I want to make some violet paint, by mixing blue and red paint. I go to the art studio, and I try to pour red paint in wood cup.

The contents of the art studio are:

This room is called the art studio. In it, you see:

    a large cupboard. The large cupboard door is closed.
    a table. On the table is: a jug (containing nothing).
    a wood cup (containing red paint)
    a substance called air
    the agent
    a wood cup (containing blue paint)
    a wood cup (containing yellow paint)

You also see:

    A door to the hallway (that is open)

And this gives me the following observation:

Ambiguous request: Please enter the number for the action you intended (or blank to cancel): 0: pour red paint (in wood cup, in art studio) in wood cup (in art studio) 1: pour red paint (in wood cup, in art studio) in wood cup (in art studio) 2: pour red paint (in wood cup, in art studio) in wood cup (in art studio)

The last observation is not informative enough to take specify one action to take to pour the red paint in the container containing blue paint.

@PeterAJansen
Copy link
Collaborator

Thanks for this note. I struggled a bit with this -- creating the ambiguity resolution message -- in trying to balance a generic method that would give some distinguishing feature, but not be too verbose. In the end, settling on the direct container that an object is in seems to work 80% of the time, but sometimes (as you note) it's still not ideal. Very welcome to hear suggestions and/or alternatives.

The gold path generation does include a method for generating unique referents -- I suppose it could be possible to take the set of referents for each ambiguous object, find the ones that are different between them, and then use those -- though the list of allowed referents can be quite large (and some can be a little unusual, since they're generated automatically using common patterns), so this might occasionally generate some unusual names.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants