My research aims to contribute to research in the narrative authoring domain by using cognitive models in narrative plan generation. These cognitive models determine how actions and events in narrative affect the audience. My research intends to leverage these models in narrative planning and use them to provide intelligent narrative plans that are structured to invoke specific responses from audiences when they experience the narrative. This sort of approach would greatly benefit the enrich growing set of variables of narrative planning.

When we were invited to write a retrospective article about our AAAI-99 paper on mutual bootstrapping (Riloff and Jones 1999), our first reaction was hesitation because, well, that algorithm seems old and clunky now. But upon reflection, it shaped a great deal of subsequent work on bootstrapped learning for natural language processing, both by ourselves and others. So our second reaction was enthusiasm, for the opportunity to think about the path from 1999 to 2017 and to share the lessons that we learned about bootstrapped learning along the way. This article begins with a brief history of related research that preceded and inspired the mutual bootstrapping work, to position it with respect to that period of time. We then describe the general ideas and approach behind the mutual bootstrapping algorithm. Next, we overview several types of research that have followed and shared similar themes: multi-view learning, bootstrapped lexicon induction, and bootstrapped pattern learning. Finally, we discuss some of the general lessons that we have learned about bootstrapping techniques for NLP to offer guidance to researchers and practitioners who may be interested in exploring these types of techniques in their own work.

We have constructed an information extraction system called the Mars Target Encyclopedia that takes in planetary science publications and extracts scientific knowledge about target compositions. The extracted knowledge is stored in a searchable database that can greatly accelerate the ability of scientists to compare new discoveries with what is already known. To date, we have applied this system to ~6000 documents and achieved 41-56% precision in the extracted information.

Modern solutions for implicit discourse relation recognition largely build universal models to classify all of the different types of discourse relations. In contrast to such learning models, we build our model from first principles, analyzing the linguistic properties of the individual top-level Penn Discourse Treebank (PDTB) styled implicit discourse relations: Comparison, Contingency and Expansion. We find semantic characteristics of each relation type and two cohesion devices---topic continuity and attribution---work together to contribute such linguistic properties. We encode those properties as complex features and feed them into a NaiveBayes classifier, bettering baselines(including deep neural network ones) to achieve a new state-of-the-art performance level. Over a strong, feature-based baseline, our system outperforms one-versus-other binary classification by 4.83% for Comparison relation, 3.94% for Contingency and 2.22% for four-way classification.

To understand narrative text, we must comprehend how people are affected by the events that they experience. For example, readers understand that graduating from college is a positive event (achievement) but being fired from one's job is a negative event (problem). NLP researchers have developed effective tools for recognizing explicit sentiments, but affective events are more difficult to recognize because the polarity is often implicit and can depend on both a predicate and its arguments. Our research investigates the prevalence of affective events in a personal story corpus, and introduces a weakly supervised method for large scale induction of affective events. We present an iterative learning framework that constructs a graph with nodes representing events and initializes their affective polarities with sentiment analysis tools as weak supervision. The events are then linked based on three types of semantic relations: (1) semantic similarity, (2) semantic opposition, and (3) shared components. The learning algorithm iteratively refines the polarity values by optimizing semantic consistency across all events in the graph. Our model learns over 100,000 affective events and identifies their polarities more accurately than other methods.

Plan-based story generation has operationalized concepts from the Belief-Desire-Intention (BDI) theory of mind to create goal-driven character agents with explainable behavior. However, these character agents are limited in that they do not capture the dynamic nature of intentions. To address this limitation, we define a plan-based intention revision model and propose an evaluation using the QUEST cognitive model to assess the explainability of an intention revision.

Previous work on story planning has lacked a knowledge representation for characters that make mistakes in the execution of their actions. In particular, characters' execution mistakes that arise from errors in belief have not been modeled. In this paper, we describe a state-space planning system and its belief model, together called HeadSpace, that generates stories that track and manipulates characters' belief about the story world around them. This model is used to produce actions in stories that are attempted but that fail. We show an example story plan that contains failed-action content that cannot be generated by typical planning-based approaches to story creation.

A goal of Experience Managers (EM) is to guide users through a space of narrative trajectories, or story branches, in an Interactive Narrative (IN). When a user performs an action that deviates from the intended trajectory, the EM uses a mediation strategy called accommodation to transition the user to a new desirable trajectory. However, generating the trajectory options then selecting the appropriate one is computationally expensive and at odds with the low-latency needs of an IN. We define three desirable properties (exemplar trajectories, narrative-theoretic comparison, and efficiency) that general solutions would possess and demonstrate how our plan-based Intention Dependency Graph addresses them.

Strong story interactive narratives (IN) are stories that branch based on participant actions where all branches conform to a set of predefined constraints. However, participants in these systems may create branches where the constraints no longer hold. Strong story experience management, the process of generating IN trees, can be viewed as a game where the experience management agent wins if the story constraints hold during gameplay and loses if they are broken. In domains where the player has incomplete information of the story world, the experience manager can take action by shifting the player between alternate states that are consistent with the player's observations in order to maximize the probability that constraints will hold. This process is called superposition manipulation. In this paper we present a method of estimating the number of goal states reachable from different states in order to make informed decisions during superposition manipulation.

Previous work on story planning has shown success in the generation of plans that are both intention-coherent and demonstrate aspects of inter-character conflict. However, the initial models of intention and conflict have been limited, in that they lack methods to generate story plots wherecharacters drop sub-plans to achieve their goals in believably consistent and expressive ways and adopt new sub-plans in the face of plan failure. In current work, we have developed models of failed actions in stories that go hand in hand with erroneous belief models for character. Motivated by characterizations of rational agents' intentions as choice combined with commitment, we provide a framing of the plan generation process that is intended to show how characters form their own plans to achieve their own goals, act upon those plans until they feel that conditions no longer support their plans, and then re-plan in the face of adversity to achieve their goals. We show an example story plan that contains several types of character-based intention dynamics targeted by our approach.