Automating Direct Speech Variations in Stories and Games

Dialogue authoring in large games requires not only content creation but the subtlety of its delivery, which can vary from character to character. Manually authoring this dialogue can be tedious, time-consuming, or even altogether infeasible. This paper utilizes a rich narrative representation for modeling dialogue and an expressive natural language generation engine for realizing it, and expands upon a translation tool that bridges the two. We add functionality to the translator to allow direct speech to be modeled by the narrative representation, whereas the original translator supports only narratives told by a third person narrator. We show that we can perform character substitution in dialogues. We implement and evaluate a potential application to dialogue implementation: generating dialogue for games with big, dynamic, or procedurally-generated open worlds. We present a pilot study on human perceptions of the personalities of characters using direct speech, assuming unknown personality types at the time of authoring. Dialogue authoring in large games requires not only the creation of content, but the subtlety of its delivery which can vary from character to character. Manually authoring this dialogue can be tedious, time-consuming, or even altogether infeasible. The task becomes particularly intractable for games and stories with dynamic open worlds in which character parameters that should produce linguistic variation may change during gameplay or are decided procedurally at runtime. Short of writing all possible variants pertaining to all possible character parameters for all of a game’s dialogue segments, authors working with highly dynamic systems currently have no recourse for producing the extent of content that would be required to account for all linguistically meaningful character states. As such, we find openworld games today filled with stock dialogue segments that are used repetitively by many characters without any linguistic variation, even in game architectures with rich character models that could give an actionable account of how their speech may vary (Klabunde 2013). Indeed, in general, we are building computational systems that, underlyingly, are far more expressive than can be manifested by current authoring practice. These concerns can also be seen in linear games, in which the number of Copyright c © 2014, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved. story paths may be limited to reduce authoring time or which may require a large number of authors to create a variety of story paths. Recent work explores the introduction of automatically authored dialogues using expressive natural language generation (NLG) engines, thus allowing for more content creation and the potential of larger story paths (Monfort, Stayton, and Campana 2014; Lin and Walker 2011; Cavazza and Charles 2005; Rowe, Ha, and Lester 2008). Figure 1: NLG pipeline method of the ES Translator. (Walker et al. 2013) explore using a dynamic and customizable NLG engine called PERSONAGE to generate a variety of character styles and realizations, as one way to help authors to reduce the authorial burden of writing dialogue instead of relying on scriptwriters. PERSONAGE is a parameterizable NLG engine grounded in the Big Five personality traits that provides a larger range of pragmatic and stylistic variations of a single utterance than other NLG engines (Mairesse and Walker 2011). In PERSONAGE, narrator’s voice (or style to be conveyed) is controlled by a model that specifies values for different stylistic parameters (such as verbosity, syntactic complexity, and lexical choice). PERSONAGE requires hand crafted text plans, limiting not only the expressibility of the generations, but also the domain. (Reed et al. 2011) introduce SpyFeet: a mobile game to encourage physical activity which makes use of dynamic storytelling and interaction. A descendant of PERSONAGE, called SpyGen, is its NLG engine. The input to SpyGen is a text plan from Inform7, which acts as the content planner and manager. (Reed et al. 2011) show that this architecture Games and Natural Language Processing: Papers from the AIIDE Workshop