Stories are typically represented as a set of events and temporal or causal relations among events. In the metro map model of storylines, participants are represented as histories and events as interactions between participant histories. The metro map model calls for a decomposition of events into what each participant does (or what happens to each participant), as well as the interactions among participants. Such a decompositional model of events has been developed in linguistic semantics. Here, we describe this decompositional model of events and how it can be combined with a metro map model of storylines.

This chapter reviews the research conducted on the representation of events, from theperspectives ofnatural language processing, artificial intelligence (AI), and linguistics. AI approaches to modeling change have traditionally focused on situations and state descriptions. Linguistic approaches start with the description of the propositional content of sentences (or natural language expressions generally). As a result, the focus in the two fields has been on different problems. I argue that these approaches have common elements that can be drawn on to view event semantics from a unifying perspective, where we can distinguish between the surface events denoted by verbal predicates and what I refer to as the latent event structure of a sentence. By clearly distinguishing between surface and latent event structures of sentences and texts, we move closer to a general computational theory of event structure, one permitting a common vocabularyfor events and the relations between them, while enabling reasoning at multiple levels of interpretation.

Understanding the timeline of a story is a necessary first step for extracting storylines. This is difficult because timelines are rarely explicitly given in documents, and fragments of a story may be found across multiple documents. We outline prior work and the state of the art in both timeline extraction and alignment of events across documents. Previous work focused mainly on temporal graph extraction rather than actual timelines. Recently, there has been a growing interest in extracting timelines from these graphs. We review this work and describe our own approach that solves timeline extraction exactly. With regard to event alignment, most efforts have focused on the specific task of cross-document event coreference (CDEC). Current approaches to CDEC perform either event-only clustering or joint event–entity clustering, with neural methods achieving the best results. We outline next steps to advance the field toward full timeline alignment across documents that can serve as a foundation for extraction of higher-level, more abstract storylines.

Event extraction aims to find who did what to whom, when, and where from unstructured data. Over the past decade, event extraction has made advances in three waves. The first wave relied on supervised machine learning models trained from a large amount of manually annotated data and crafted features. The second wave introduced deep neural networks with distributional semantic embedding features but still required large annotated data sets. This chapter provides an overview of a third wave with a share-and-transfer framework, which enhances the portability of event extraction by transferring knowledge from a high-resource setting to another low-resource setting, reducing the need for annotated data. The first share step is to construct a common structured semantic representation space into which these complex structures can be encoded. Then, in the transfer step, we train event extractors over these representations in high-resource settings and apply the learned extractors to target data in the low-resource setting. We conclude with a summary of the current status of this new framework and point to remaining challenges and future research directions to address them.

Traditional event detection systems typically extract structured information on events by matching predefined event templates through slot filling. Automatically linking of related event templates extracted from different documents over a longer period of time is of paramount importance for analysts to facilitate situational monitoring and manage the information overload and other long-term data aggregation tasks. This chapter reports on exploring the usability of various machine learning techniques, textual, and metadata features to train classifiers for automatically linking related event templates from online news. In particular, we focus on linking security-related events, including natural and man-made disasters, social and political unrest, military actions and crimes. With the best models trained on moderate-size corpus (ca. 22,000 event pairs) that use solely textual features, one could achieve an F1 score of93.6%. This figure is further improved to 96.7% by inclusion of event metadata features, mainly thanks to the strong discriminatory power of automatically extracted geographical information related to events.

Witness testimony provides the first draft of history and requires a kind of reading that connects descriptions of events from many perspectives and sources. This chapter examines one critical step in that connective process, namely, how to assess a speaker's certainty about the events they describe. By surveying a group of approximately 300 readers and their approximately 28,000 decisions about speaker certainty, this chapter explores how readers may think about factual and counterfactual statements, and how they interpret the certainty with which a witness makes their statements. Ultimately, this chapter argues that readers of collections of witness testimony were more likely to agree about event descriptions when those providing the description were certain, and that readers' abilities to accept gradations of certainty were better when a witness described factual, rather than counterfactual or negated events. These findings lead to a suggestion for how researchers in natural language processing could better model the question of speaker certainty, at least when dealing with the kind of narrative nonfiction one finds in witness testimony.

For an equivariant commutative ring spectrum R, ?0R has algebraic structure reflecting the presence of both additive transfers and multiplicative norms. The additive structure gives rise to a Mackey functor and the multiplicative structure yields the additional structure of a Tambara functor. If R is an N? ring spectrum in the category of genuine G-spectra, then all possible additive transfers are present and ?0R has the structure of an incomplete Tambara functor. However, if R is an N? ring spectrum in a category of incomplete G-spectra, the situation is more subtle. In this chapter, we study the algebraic theory of Tambara structures on incomplete Mackey functors, which we call bi-incomplete Tambara functors. Just as incomplete Tambara functors have compatibility conditions that control which systems of norms are possible, bi-incomplete Tambara functors have algebraic constraints arising from the possible interactions of transfers and norms. We give a complete description of the possible interactions between the additive and multiplicative structures.

Stories are a pervasive phenomenon of human life. They also represent a cognitive tool to understand and make sense of the world and of its happenings. In this contribution we describe a narratology-based framework for modeling stories as a combination of different data structures and to automatically extract them from news articles. We introduce a distinction among three data structures (timelines, causelines, and storylines) that capture different narratological dimensions, respectively chronological ordering, causal connections, and plot structure. We developed the Circumstantial Event Ontology (CEO) for modeling (implicit) circumstantial relations as well as explicit causal relations and create two benchmark corpora: ECB+/CEO, for causelines, and the Event Storyline Corpus (ESC), for storylines. To test our framework and the difficulty in automatically extract causelines and storylines, we develop a series of reasonable baseline systems

A variety of approaches exist for annotating temporal and event information in text, but it has been difficult to compare and contrast these different corpora. The Richer Event Description (RED) corpus, as an ambitious annotation of temporal, causal, and coreference annotation, provides one point of comparison for discussing how different annotation decisions contribute to the timeline and causal chains which define a document. We present an overview of how different event corpora differ and present new methods for studying the impact of these temporal annotation decisions upon the resulting document timeline. This focus on illuminating the contribution of three particular sources of information – event coreference, causal relations with temporal information, and long-distance temporal containment – to the actual timeline of a document. By studying the impact of specific annotation strategies and framing the RED annotation in the larger context of other event–event relation annotation corpora, we hope to provide a clearer sense of the current state of event annotation and of promising future directions for annotation.

In the past we experimented with variations of an approach we call semantic storytelling, in which we use multiple text analytics components including named entity recognition and event detection. This chapter summarizes some of our previous work with an emphasis on the detection of movement action events, and describes the long-term semantic storytelling vision as well as the setup and approach of our future work towards a robust technical solution, which is primarily driven by three industry use cases. Ultimately, we plan to contribute an implemented approach for semantic storytelling that makes use of various analytics services and that can be deployed in a flexible way in various industrial production environments.

A crucial aspect of understanding and reconstructing narratives is identifying the underlying causal chains, which explain why certain things happened and make a coherent story. To build such causal chains, we need to identify causal links between events in the story, which may be expressed explicitly as well as understood implicitly using commonsense knowledge.

This chapter reviews research efforts on the automated extraction of such event causality from natural language text. It starts with a brief review of existing causal models in psychology and psycholinguistics as a building block for understanding causation. These models are useful tools for guiding the annotation process to build corpora annotated with causal pairs. I then outline existing annotated resources, which are used to build and evaluate automated causality extraction systems. Furthermore, circumstantial events surrounding the causal complex are rarely expressed with language as they are part of common sense knowledge. Therefore, discovering causal common sense is also important to fill the gaps in the causal chains, and I discuss existing work in this line of research.

This chapter reviews the current landscape of ontological and lexical resources that motivated the development of the Rich Event Ontology (REO). Aimed at a whole that is greater than the sum of its parts, REO functions as a conceptual organization of event types that facilitates mapping between FrameNet, VerbNet, and the Entities, Relations, and Events corpus annotation from the Linguistic Data Consortium.

The definition of the homotopy limit of a diagram of left Quillen functors of model categories has been useful in a number of applications. In this chapter we review its definition and summarise some of these applications. We conclude with a discussion of why we could work with right Quillen functors instead, but cannot work with a combination of the two.

This chapter presents techniques for examining the distributional properties of narrative schemas in a subset of the New York Times (NYT) Corpus. In one technique, the narrative argument salience through entities annotated (NASTEA) task, we use the event participants indicated by narrative schemas to replicate salient entity annotations from the NYT Corpus. In another technique, we measure narrative schema stability by generating schemas with various permutations of input documents. Both of these techniques show differences between homogeneous and heterogeneous document categories. Homogeneous categories tend to perform better on the NASTEA task using fewer schemas and exhibit more stability, whereas heterogeneous categories require more schemas applied on average to peak in performance at the NASTEA task and exhibit less stability. This suggests that narrative schemas succeed at detecting and modeling the repetitive nature of template-written text, whereas more sophisticated models are required to understand and interpret the complex novelty found in heterogeneous categories.