Skip to content

Instantly share code, notes, and snippets.

@chaonan99

chaonan99/vaff_anno.md

Created May 7, 2018 15:01
Show Gist options
  • Save chaonan99/c6c00b4edeae31644c0869a7967211a7 to your computer and use it in GitHub Desktop.
Save chaonan99/c6c00b4edeae31644c0869a7967211a7 to your computer and use it in GitHub Desktop.
Download ZIP
Raw
vaff_anno.md

vaff annotation

The annotation of vaff follows Umich Affordance Matrix and annotate 1~5 score indicating the confidence of an concept being the keyrole of a predicate.

To annotate, give each predicate-concept pair a 1~5 score, where 5 being most confident to say the action can happen on the object. For example, for the first pair (add, air), ask yourself "Can you add air (to something)?". (Notice that this is aiming for a cooking scenario, and eventually we want to let the robot to conduct the action on the object. But don't restrict your imagination on this. Use your common sense in first place!)

Here are some suggestions for success in this task:

  • Use Google Sheet or MS Excel for the annotation task.
  • Predicate-wise annotation, e.g., fist annotate all (add, *) pairs, then move on to arrange action
  • After annotating each predicate, swap the object column with annotated column.
    • In Google Sheet, see this tutorial
    • I believe there is a similar feature in MS Excel

The estimated annotation time is about 3~5 hours.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment