Fine-tuned Qwen2.5-7B on 100 films for probabilistic story graphs

This capability generates probabilistic story graphs by leveraging a fine-tuned Qwen2.5-7B model that has been specifically trained on a dataset of 100 films. It utilizes a transformer architecture to understand narrative structures and relationships between characters and events, allowing it to output complex story arcs based on learned probabilities. The model's training on diverse cinematic narratives enables it to capture a wide range of storytelling techniques and styles, making it distinct in its ability to produce nuanced and varied story graphs.

This capability maps relationships between characters in a story by analyzing dialogue and interactions within the context of the trained films. It employs natural language processing techniques to identify and categorize interactions, allowing users to visualize how characters influence each other's arcs. This mapping is probabilistic, meaning it can suggest potential relationship dynamics based on learned patterns from the training data, providing a unique perspective on character development.

This capability predicts potential narrative arcs by analyzing the structure and flow of stories within the training dataset. It employs machine learning techniques to identify common patterns and tropes in storytelling, allowing it to suggest plausible future events or twists based on the established narrative. This predictive modeling is grounded in the probabilistic nature of the training data, making it capable of generating varied outcomes that align with typical storytelling conventions.