Abstract
We present NarratEX, a dataset designed for the task of explaining the choice of the Dominant Narrative in a news article, and intended to support the research community in addressing challenges such as discourse polarization and propaganda detection. Our dataset comprises 1,056 news articles in four languages, Bulgarian, English, Portuguese, and Russian, covering two globally significant topics: the Ukraine-Russia War (URW) and Climate Change (CC). Each article is manually annotated with a dominant narrative and sub-narrative labels, and an explanation justifying the chosen labels. We describe the dataset, the process of its creation, and its characteristics. We present experiments with two new proposed tasks: Explaining Dominant Narrative based on Text, which involves writing a concise paragraph to justify the choice of the dominant narrative and sub-narrative of a given text, and Inferring Dominant Narrative from Explanation, which involves predicting the appropriate dominant narrative category based on an explanatory text. The proposed dataset is a valuable resource for advancing research on detecting and mitigating manipulative content, while promoting a deeper understanding of how narratives influence public discourse.

Abstract
As recommender systems play an increasingly central role in shaping information exposure on platforms like YouTube, understanding the nature of the content they promote, especially in sensitive contexts, requires scalable and reliable labelling methods. This paper investigates the use of Large Language Models (LLM) to label YouTube videos based solely on their metadata. We propose a committee-based approach that aggregates predictions from an ensemble of seven state-of-the-art LLMs through majority voting. Using a novel dataset collected via simulated user interactions on YouTube, we analyse model agreement, labelling behavior, and the influence of model size. To assess label reliability, we also investigate the semantic coherence of label assignments. Our results show that LLM committees produce highly consistent labels in low-disagreement settings. These findings highlight both the promise and limitations of LLM-based annotation for auditing social networks.

Abstract
For eight years, the Text2Story Workshop series has fostered a vibrant research community dedicated to narrative understanding, advancing shared insights into the challenges of modelling narrative structure in text. While earlier approaches laid important foundations, recent progress in Transformers and Large Language Models (LLMs) has fundamentally reshaped the field. Building on the increasing prominence of LLM-based contributions in recent editions, the ninth edition of Text2Story expands the focus toward agentic AI, where systems plan, reason, and interact over time using narratives as internal representations. Recent advances, including long-context architectures, instruction and preference-tuned models, retrieval-augmented generation, and discourse-aware prompting, have broadened the applicability of LLMs to complex narrative tasks. Nevertheless, reliably capturing fine-grained narrative structures remains challenging, particularly for event chains, temporal and causal relations, character development, and perspective consistency. These challenges are amplified in interactive and agentic settings, where narrative coherence, controllability, and reliability are critical. This edition of Text2Story explores both the opportunities and limitations of LLMs and agentic systems for narrative understanding, including the analysis of narratives generated by LLMs themselves with respect to consistency, hallucination, bias, and control. Through a diverse program of research papers, works in progress, demos, resources, and keynote talks, the workshop continues to advance narrative understanding in the era of foundation and agentic models.

Abstract

Abstract
Heterogeneous graphs can represent real-world problems in a way close to reality, supporting diverse types of vertices and edges. However, their inherent heterogeneity poses challenges in interpreting problem semantics. To address this, heterogeneous graph embedding, aiming to map graph elements to low-dimensional vectors, simplifies subsequent machine learning analysis. This approach has gained prominence in machine learning, fueling classification, recommendation, and similarity search applications. Embedding diverse data is essential for efficient data processing. Incorporating language models, like BERT, into heterogeneous graphs enhances semantic context capture, which is particularly useful when one vertex type represents text. Language models stand out in contextual representation, enriching graph vertex embeddings for various tasks. This paper proposes a novel approach to enhancing heterogeneous graph embeddings by combining language models and task class data. Our approach increases vector quality, accounting for graph structure, semantic textual information, and task labels. We compared our proposal with a language model in the aspect-based sentiment analysis task, demonstrating competitive results and, in some cases, a slight superiority. Furthermore, we explore applications of embeddings from auxiliary vertices in another task, highlighting another advantage of the approach over the language model.

Abstract
The development of a robust annotation scheme and corresponding guidelines is crucial for pro- ducing annotated datasets that advance both lin- guistic and computational research. This paper presents a case study that outlines a method- ology for designing an annotation scheme and its guidelines, specifically aimed at represent- ing morphosyntactic and semantic information regarding temporal features, as well as medi- cal information in medical reports written in Portuguese. We detail a multi-step process that includes reviewing existing frameworks, con- ducting an annotation experiment to determine the optimal approach, and designing a model based on these findings. We validated the ap- proach through a pilot experiment where we assessed the reliability and applicability of the annotation scheme and guidelines. In this ex- periment, two annotators independently anno- tated a patient's medical report consisting of six documents using the proposed model, while a curator established the ground truth. The analy- sis of inter-annotator agreement and the annota- tion results enabled the identification of sources of human variation and provided insights for further refinement of the annotation scheme and guidelines.