WorldPT: A Structured Dataset of Fictional Worlds in Portuguese

Abstract

Dataset Description WorldPT: A Structured Dataset of Fictional Worlds in Portuguese is a curated and fully structured dataset designed for the study, generation, and analysis of fictional universes (worldbuilding). The dataset organizes narrative elements into multiple interconnected categories, including characters, locations, cultures, events, organizations, technologies, magic systems, religions, races, and more, each stored as standardized JSON files written entirely in Portuguese. Each world is represented as an independent directory containing entities with consistent metadata fields such as unique ID, title, genre, narrative text, content type, data source, and automatically detected cross-entity relationships. This relational structure enables the dataset to be used for knowledge graph construction, narrative network analysis, AI world generation, and computational creativity research. WorldPT serves as a reference resource for researchers and developers interested in natural language processing (NLP), semantic modeling, and artificial intelligence applied to storytelling and worldbuilding, especially in the context of the Portuguese language, where such resources remain scarce. At its core, WorldPT models each fictional universe as a graph, where: Entities (nodes) represent elements of the world (e.g., characters, places, artifacts). Relationships (edges) represent how these entities are connected. However, unlike conventional knowledge graphs, WorldPT introduces two critical innovations: Directed Relationships with Semantic Grounding All relationships are directional and meaningful. The direction of a connection encodes a dependency: The source entity is understood as being contextually grounded in the target entity. The target entity provides the context, origin, or constraint for the source. For example: A character is grounded in an organization (membership). An event is grounded in a timeline (temporal placement). This principle ensures that the graph reflects how the world is structured, not just which elements are associated. Multilayer Narrative Structure (Hexapartite System) Each relationship exists within one of six distinct semantic layers, forming a multilayer (or multiplex) graph. This allows the same pair of entities to be connected in different ways depending on context. Each layer represents a different dimension of narrative meaning: 1. Structural Layer Captures spatial organization and hierarchical containment.Examples: A city located within a kingdom A character belonging to a region 2. Causal Layer Represents cause-and-effect relationships and transformations.Examples: A technology influencing an economy An event triggering another event 3. Temporal Layer Encodes chronological structure and time anchoring.Examples: An event occurring within a specific timeline Historical sequencing of occurrences 4. Social Layer Models interpersonal relationships and social structures.Examples: Alliances, kinship, or membership Authority and social roles 5. Ontological Layer Defines the fundamental rules and systems governing the world.Examples: A character using a magic system Entities constrained by metaphysical laws 6. Symbolic Layer Represents meaning, ideology, and cultural influence.Examples: A religion shaping a culture Symbols influencing collective beliefs This layered structure allows the dataset to disambiguate complex relationships that would otherwise be conflated in traditional representations. For instance, a character may simultaneously belong to a group (social), be located in a place (structural), and be influenced by an ideology (symbolic). Ontological Classes WorldPT defines a fixed set of 13 ontological classes to categorize all entities in a fictional world. These classes ensure full coverage of narrative elements while minimizing ambiguity. They include: Agents and groups: characters, races, organizations Spatiotemporal anchors: places, events, timelines Societal systems: culture, religion, politics, economy World constraints: magic systems, technology, artifacts This taxonomy ensures that every entity has a clearly defined role within the world, enabling consistent modeling and analysis