Abstract
In order to tell stories in different voices for different audiences, interactive story systems require: (1) a semantic representation of story structure, and (2) the ability to automatically generate story and dialogue from this semantic representation using some form of Natural Language Generation (NLG). However, there has been limited research on methods for linking story structures to narrative descriptions of scenes and story events. In this paper we present an automatic method for converting from Scheherazade's story intention graph, a semantic representation, to the input required by the Personage NLG engine. Using 36 Aesop Fables distributed in DramaBank, a collection of story encodings, we train translation rules on one story and then test these rules by generating text for the remaining 35. The results are measured in terms of the string similarity metrics Levenshtein Distance and BLEU score. The results show that we can generate the 35 stories with correct content: the test set stories on average are close to the output of the Scheherazade realizer, which was customized to this semantic representation. We provide some examples of story variations generated by personage. In future work, we will experiment with measuring the quality of the same stories generated in different voices, and with techniques for making storytelling interactive.
Introduction
Storytelling is a dynamic social act where narrators adjust their tone, level of detail, and vocabulary based on their audience. However, most current interactive storytelling systems are brittle, relying on hand-crafted text or simple "story trees" that make it difficult to change the narrative's "voice" without significant manual effort. The core technical challenge lies in the gap between content planning (what happens in the story) and surface realization (how it is actually said). This paper introduces a pipeline for automatically generating multiple versions of the same story by bridging two specialized tools: Scheherazade, which encodes the semantic "meaning" of a story, and PERSONAGE, an NLG engine capable of generating text in various stylistic "voices."
Methodology
The authors developed an automated "bridge" between two existing frameworks to map deep semantic structures to syntactic ones.
- Scheherazade: An annotation tool that represents stories as a Story Intention Graph (SIG). This graph encodes events as predicate-argument structures anchored in ontologies like WordNet and VerbNet.
- PERSONAGE: An NLG engine that takes Deep Syntactic Structures (DSyntS) as input and applies stylistic parameters (based on Big Five personality traits) to produce varied outputs.
The pipeline involves extracting actions and timelines from the SIG encoding via the Scheherazade API, mapping these semantic "frames" into lexico-syntactic classes, and automatically building a DSyntS dependency tree. This tree is then fed into PERSONAGE, which uses the RealPro surface realizer to turn the syntax into actual sentences. By adjusting PERSONAGE’s 67 different parameters, the system can generate the same story in various voices, such as "Shy" or "Laid-back."
Evaluation and Results
The study utilized DramaBank, a collection of story encodings. One fable, "The Fox and the Grapes," was used to develop the translation rules, which were then tested on the remaining 35 Aesop Fables. The system was evaluated using BLEU scores (n-gram overlap) and Levenshtein Distance (edit distance) to compare PERSONAGE’s output against a baseline Scheherazade output.
On the test set, the system achieved a Levenshtein Mean of 72 and a BLEU score of 0.32. While the BLEU score was lower than the development set (0.59), the results were statistically similar to the baseline in terms of content correctness. Qualitatively, the system successfully produced distinct versions of stories. For example, where a formal version might say "The fox didn't obtain the grapes," a "Laid-back" version could say "He damn was hungry... the fox failed to get the group of grapes, okay?" This demonstrates that a single symbolic representation of a story can serve as the source for infinite stylistic variations.
Conclusion and Future Work
While the system demonstrates the feasibility of automated semantic-to-syntactic mapping, it currently struggles with some linguistic nuances, such as irregular verb forms and complex tenses. Furthermore, metrics like BLEU and Levenshtein measure "correctness" rather than "stylistic quality." Future work will involve human subjective judgments to evaluate the effectiveness of different voices and extending the stylistic models to individual characters within the story to provide unique personalities during dialogue.
Summary updated