Narrative Planning Model Acquisition from Text Summaries and Descriptions

Abstract

AI Planning has been shown to be a useful approach for the generation of narrative in interactive entertainment systems and games. However, the creation of the underlying narrative domain models is challenging: the well documented AI planning modelling bottleneck is further compounded by the need for authors, who tend to be non-technical, to create content. We seek to support authors in this task by allowing natural language (NL) plot synopses to be used as a starting point from which planning domain models can be automatically acquired. We present a solution which analyses input NL text summaries, and builds structured representations from which a pddl model is output (fully automated or author in-the-loop). We introduce a novel sieve-based approach to pronoun resolution that demonstrates consistently high performance across domains. In the paper we focus on authoring of narrative planning models for use in interactive entertainment systems and games. We show that our approach exhibits comprehensive detection of both actions and objects in the system-extracted domain models, in combination with significant improvement in the accuracy of pronoun resolution due to the use of contextual object information. Our results and an expert user assessment show that our approach enables a reduction in authoring effort required to generate baseline narrative domain models from which variants can be built.

NaRuto: Automatically Acquiring Planning Models from Narrative Texts

“…Hayton et al (2017) and Huo et al (2020) explored narrative texts, yet their action model generation largely depends on manual interventions. Hayton et al (2020) proposed an automated process, but their action models aim only to mirror the input narrative.…”

“…Recently, researchers have attempted to extract action models from narrative texts such as short stories and movie synopses (Hayton et al 2017;Huo et al 2020;Hayton et al 2020). However, the methods proposed so far either generate quite simple and highly specific action models, or rely on human effort to complement or correct automatic extraction.…”

A Comprehensive Framework for Learning Declarative Action Models

“…Interestingly, this work learns the interactions between the pieces in the games from example sequences of the moves made in playing those games. Another successful field of application of action model learning is narrative generation for story telling (Lindsay, Read, Ferreira, Hayton, Porteous, & Gregory, 2017b;Hayton, Porteous, Ferreira, & Lindsay, 2020). Dif-ferently from the previous applications, the examples in narrative generation are natural language descriptions of activity sequences.…”

“…Most of the reviewed systems assume that an abstract state representation is available. Again this is a too strong assumption for many real-world applications, where images (Asai & Fukunaga, 2018;Asai & Muise, 2020), or natural language text, are the only available learning examples (Lindsay, Read, Ferreira, Hayton, Porteous, & Gregory, 2017a;Hayton et al, 2020). The computation of abstract state representations for effective problem solving is actually a core problem in AI that appears in many different forms such as at the computation of generalized plans (Lotinac, Segovia-Aguas, Jiménez, & Jonsson, 2016;Bonet, Frances, & Geffner, 2019), predicate invention for relational learning (Kok & Domingos, 2007), feature discovering from scene labeling (Farabet, Couprie, Najman, & LeCun, 2013) or in RL (Konidaris et al, 2018).…”

Learning a Planning Domain Model From Natural Language Process Manuals

“…Hayton et al [13] used LOCM to complete the story reconstruction after extracting an action sequence through NLP tools. They also provided a specific method to domain model acquisition for domains that require non-technical experts to create content to populate their models [41]. Janghorbani et al [23] developed virtual agents that can obtain preconditions and effects from a compound of NL sentences.…”