Narrative based hypothesis generation using event coherence and probabilistic inference

This is a project in the DARPA AIDA program. Quoting from their webpage: "The goal of Active Interpretation of Disparate Alternatives (AIDA) is to develop a multihypothesis semantic engine that generates explicit alternative interpretations of events, situations, and trends from a variety of unstructured sources, for use in noisy, conflicting, and potentially deceptive information environments"

In the University of Texas AIDA project, we view the task of AIDA hypothesis generation as narrative generation: A model selects from a collection of events and participants to generate a coherent narrative. To generate hypotheses that are useful for human analyses, we combine a local generator model with a global constraint model that uses rules that enforce large-scale consistency among selected elements, and conformity with the AIDA ontology.



Graduate students:


Pengxiang Cheng and Katrin Erk. Implicit Argument Prediction as Reading Comprehension. Proceedings of AAAI 2019.

Su Wang, Eric Holgate, Greg Durrett, and Katrin Erk. Picking Apart Story Salads. Proceedings of EMNLP 2018. Code and data.

Su Wang, Greg Durrett, and Katrin Erk. Modeling semantic plausibility by injecting world knowledge. Proceedings of NAACL 2018.


Story Salads