Skip to content

Latest commit

 

History

History
49 lines (39 loc) · 2.44 KB

README.md

File metadata and controls

49 lines (39 loc) · 2.44 KB

Winnow

Winnow is a declarative domain-specific query language for story sifting, or identifying narratively compelling sequences of events in a larger corpus of narrative material. For instance, Winnow can be used to search for compelling emergent microstories within the output of a simulation-driven game, such as Dwarf Fortress or The Sims.

Example

A typical Winnow sifting pattern looks something like this:

(pattern violationOfHospitality
  (event ?e1 where
    eventType: enterTown,
    actor: ?guest)
  (event ?e2 where
    eventType: showHospitality,
    actor: ?host,
    target: ?guest,
    ?host.value: communalism)
  (event ?e3 where
    tag: harm,
    actor: ?host,
    target: ?guest)
  ;; make sure the guest hasn't left town
  (unless-event ?eMid between ?e1 ?e3 where
    eventType: leaveTown,
    actor: ?guest))

This sifting pattern can be used to find instances of a violation of hospitality microstory, in which a ?guest character enters a town; is shown hospitality by a ?host character who values communalism; but then is somehow harmed by the ?host character before the ?guest has a chance to leave town again.

More examples of Winnow sifting patterns can be found on the tests page.

Use cases

Winnow sifting patterns can be automatically translated into lower-level Felt sifting patterns, allowing you to use Winnow as a more human-friendly syntax for the specification of Felt patterns. However, Winnow also provides unique affordances for incremental story sifting (sifting while the simulation is still live) and partial story sifting (identifying the beginnings of compelling event sequences that haven't yet been completed.)

Further reading

For more information on Winnow, see the following paper:

This paper can be cited as follows:

@inproceedings{kreminski2021winnow,
  title={Winnow: A Domain-Specific Language for Incremental Story Sifting},
  author={Kreminski, Max and Dickinson, Melanie and Mateas, Michael},
  booktitle={Proceedings of the AAAI Conference on Artificial Intelligence and Interactive Digital Entertainment},
  year={2021}
}