Goto

Collaborating Authors

Generating Narrative Text in a Switching Dynamical System

arXiv.org Artificial Intelligence

Early work on narrative modeling used explicit plans and goals to generate stories, but the language generation itself was restricted and inflexible. Modern methods use language models for more robust generation, but often lack an explicit representation of the scaffolding and dynamics that guide a coherent narrative. This paper introduces a new model that integrates explicit narrative structure with neural language models, formalizing narrative modeling as a Switching Linear Dynamical System (SLDS). A SLDS is a dynamical system in which the latent dynamics of the system (i.e. how the state vector transforms over time) is controlled by top-level discrete switching variables. The switching variables represent narrative structure (e.g., sentiment or discourse states), while the latent state vector encodes information on the current state of the narrative. This probabilistic formulation allows us to control generation, and can be learned in a semi-supervised fashion using both labeled and unlabeled data. Additionally, we derive a Gibbs sampler for our model that can fill in arbitrary parts of the narrative, guided by the switching variables. Our filled-in (English language) narratives outperform several baselines on both automatic and human evaluations.


Minstrel Remixed: User Interface and Demonstration

AAAI Conferences

This demo features a user interface for authoring stories and story fragments for use by the Minstrel Remixed story generation system. It also demonstrates Minstrel Remixed in use, allowing users to author story fragments and then have Minstrel Remixed expand these fragments and generate stories based on them. The focus is on the interface for story-fragment authoring, which exposes Minstrel's graph- of-frames knowledge representation format to the user in an interactive manner. It also exposes Minstrel Remixed's story generation capabilities as they exist currently, including the Author-Level Planning (ALP) and Transform Adapt Recall Methods (TRAM) systems.


Reader-Model-Based Story Generation

AAAI Conferences

Several existing systems have used reader models for story generation, but they have focused on either interactive contexts or pure discourse-level manipulation. I intend to buid a reader-model story generator that not only applies reader modelling to full plot generation, but which also draws on theories about intentionality and emotions put forward by Lisa Zunshine, Keith Oatley, and Raymond Mar. To evaluate the contributions of the reader model, I'll compare it with human-authored stories using measures of reader engagement. I'll also run the model on human-authored stories and compare the results to a human gold-standard analysis.


Improving Plan-Based Interactive Narrative Generation

AAAI Conferences

I am interested in algorithms that generate well structured interactive narratives that allow human participants to effect change in a story world. Narrative can be structured as a sequence of actions to be carried out by a set of characters in a particular environment. This narrative structure can be generated by AI planning. A process called accommodative mediation may be used on a single narrative plan to create a participatory experience that branches based on a human's interactions as a character within the story. However, accommodation is limited by search time and the number of plans the algorithm can possibly reach. In this paper, I discuss an initial modification of accommodative mediation that expands the space of plans the algorithm may search. I also identify several avenues by which the process may continue to be improved.


Computational Approaches to Storytelling and Creativity

AI Magazine

Features relevant to creativity and to stories are analyzed, and existing systems are reviewed under the light of that analysis. The extent to which they implement the key features proposed in recent models of computational creativity is discussed. Limitations, avenues of future research, and expected trends are outlined. Yet over the last few years there has been a surge of research efforts concerning the combination of both subjects. This article tries to shed light on these efforts.