This paper describes qualitative spatial representations relevant to cartoon motion incorporated into NarrativeML, an annotation scheme intended to capture some of the core aspects of narrative. These representations are motivated by linguistic distinctions drawn from cross-linguistic studies. Motion is modeled in terms of transitions in spatial configurations, using an expressive dynamic logic with the manner and path of motion being derived from a few basic primitives. The manner is elaborated to represent properties of motion that bear on character affect. Such representations can potentially be used to support cartoon narrative summarization and question-answering. The paper discusses annotation challenges, and the use of computer vision to help in annotation. Work is underway on annotating a cartoon corpus in terms of this scheme.
Narrative is emerging as a notion that may enable overcoming the limitations of the discovery functionality (only ranked lists of objects) offered by information systems to their users. We present preliminary results on modelling narratives by means of formal ontology, by introducing a conceptualization of narratives and a mathematical expression of it. Our conceptualization tries to capture fundamental notions of narratives as defined in narratology, such as fabula, narration and plot. A validation of the conceptualization and of its mathematical specification is ongoing, based on the Semantic Web standards and on the CIDOC CRM ISO standard ontology.
We give a preliminary description of ProppML, an annotation scheme designed to capture all the components of a Proppian-style morphological analysis of narratives. This work represents the first fully complete annotation scheme for Proppian morphologies, going beyond previous annotation schemes such as PftML, ProppOnto, Bod et al., and our own prior work. Using ProppML we have annotated Propp's morphology on fifteen tales (18,862 words) drawn from his original corpus of Russian folktales. This is a significantly larger set of data than annotated in previous studies. This pilot corpus was constructed via double annotation by two highly trained annotators, whose annotations were then combined after discussion with a third highly trained adjudicator, resulting in gold standard data which is appropriate for training machine learning algorithms. Agreement measures calculated between both annotators show very good agreement (F_1>0.75, kappa>0.9 for functions; F_1>0.6 for moves; and F_1>0.8, kappa>0.6 for dramatis personae). This is the first robust demonstration of reliable annotation of Propp's system.
Branching story games have gained popularity for creating unique playing experiences by adapting story content in response to user actions. Research in interactive narrative (IN) uses automated planning to generate story plans for a given story problem. However, a story planner can generate multiple story plan solutions, all of which equally-satisfy the story problem definition but contain different story content. These differences in story content are key to understanding the story branches in a story problem's solution space, however we lack narrative-theoretic metrics to compare story plans. We address this gap by first defining a story plan summarization model to capture the important story semantics from a story plan. Secondly, we define a story plan comparison metric that compares story plans based on the summarization model. Using the Glaive narrative planner and a simple story problem, we demonstrate the usefulness of using the summarization model and distance metric to characterize the different story branches in a story problem's solution space.
The object of the present paper is a study of two traditional litanies, Litany of the Saints and Litany of Loreto, in their most ancient attested form. We will design a narrative grammar to show how their litanic structures can be generated. We propose the notion of n-selection, a narrative rule which can't be reduced to syntax or semantics: it depends on culture. We will test the grammar on the Litany to the Divine Mercy, written by Saint Faustina Kowalska in the XXth century. Our purpose is to identify the rules of the genre which allow its reproduction in more recent versions.
Although theoretical models of the structure of narrative arising from systematic analysis of corpora are available for domains such as Russian folk tales, there are no such sources for the plot lines of musical theatre. The present paper reports an effort of knowledge elicitation for features that characterise the narrative structure of plot in the particular domain of musical theatre. The following aspects are covered: identification of a valid vocabulary of abstract units to use in annotating musical theatre plots, development of a procedure for annotation - including a spread-sheet format for annotators to use, and a corresponding set of instructions to guide them through the process - selection of a corpus of musical theatre pieces that would constitute the corpus to be annotated, the annotation process itself and the results of post-processing the annotated corpus in search for insights on the narrative structure of musical theatre plots.
The effects of social issue documentaries are diverse. In particular, monetary donations and advocacy on social media are behavioral effects with public consequences. Conversely, information-seeking about an issue is potentially done in private. We designed a combined free-viewing and rapid perceptual decision-making experiment to simulate a real scenario confronted by otherwise uninformed movie-viewers, i.e., to determine what degree of support they will lend to a film based on its trailer. For a cohort of subjects with active video-streaming (e.g., Netflix) and social media accounts (e.g., Facebook), we recorded electroencephalography (EEG) and behavioral responses to trailers of social issue documentaries. We examined EEG using reliable component analysis (RCA), finding reliability within subjects across multiple viewings and across subjects within a given viewing of the same trailer. We found this reliability both over EEG captured from whole-movie viewing, as well as over 5-second movie segments. Behavioral responses following trailer viewing were not consistent from first to second viewings. Rather, support choices both tended towards extremes of support/non-support and were made faster upon second viewing. We hypothesized a relationship between reliability behavioral metrics, finding credible evidence for it in this dataset. Finally, we found that we could suitably train a naive classifier to categorize production value and narrative voice ratings given to the viewed movies from RCA-based metrics alone. In sum, our results show that EEG components during free-viewing of social issue documentary trailers can provide a useful tool to investigate viewers' neural responses during viewing, when coupled with a post hoc behavioral decision-making paradigm. The possibility of this tool being used by producers and filmmakers is also discussed.
Our effort at developing computational models for African narratives, particularly those of Yoruba folktales, is challenged by the diversity in concepts and methodologies in the discipline. This motivated us to pause and consider the various computational models of narratives in the literature. This is with a view to finding the most appropriate or otherwise adapt a closely related one for the purpose. Thorndyke's story grammar was among the models of narrative in the literature which were appraised, found close in structure and was adapted for the modelling of Yoruba folktales narrative. In conclusion we found that the modified version of Thorndyke's model was appropriate for modelling Yoruba folktales narrative.
Motifs are distinctive recurring elements found in folklore, and are used by folklorists to categorize and find tales across cultures and track the genetic relationships of tales over time. Motifs have significance beyond folklore as communicative devices found in news, literature, press releases, and propaganda that concisely imply a large constellation of culturally-relevant information. Until now, folklorists have only extracted motifs from narratives manually, and the conceptual structure of motifs has not been formally laid out. In this short paper we propose that it is possible to automate the extraction of both existing and new motifs from narratives using supervised learning techniques and thereby possible to learn a computational model of how folklorists determine motifs. Automatic extraction would enable the construction of a truly comprehensive motif index, which does not yet exist, as well as the automatic detection of motifs in cultural materials, opening up a new world of narrative information for analysis by anyone interested in narrative and culture. We outline an experimental design, and report on our efforts to produce a structured form of Thompson's motif index, as well as a development annotation of motifs in a small collection of Russian folklore. We propose several initial computational, supervised approaches, and describe several possible metrics of success. We describe lessons learned and difficulties encountered so far, and outline our plan going forward.
The recently born expression "analytic narratives" refers to studies that have appeared at the boundaries of history, political science, and economics. These studies purport to explain specific historical events by combining the usual narrative way of historians with the analytic tools that economists and political scientists find in rational choice theory. Game theory is prominent among these tools. The paper explains what analytic narratives are by sampling from the eponymous book Analytic Narratives by Bates, Greif, Levi, Rosenthal, and Weingast (1998) and covering one outside study by Mongin (2008). It first evaluates the explanatory performance of the new genre, using some philosophy of historical explanation and then checks its discursive consistency, using some narratology. The paper concludes that analytic narratives can usefully complement standard narratives in historical explanation, provided they specialize in the gaps that these narratives reveal and that they are discursively consistent, despite the tension that combining a formal model with a narration creates. Two expository modes, called alternation and local supplementation, emerge from the discussion as the most appropriate ones to resolve this tension
Having access to a large set of stories is a necessary first step for robust and wide-ranging computational narrative modeling; happily, language data - including stories - are increasingly available in electronic form. Unhappily, the process of automatically separating stories from other forms of written discourse is not straightforward, and has resulted in a data collection bottleneck. Therefore researchers have sought to develop reliable, robust automatic algorithms for identifying story text mixed with other non-story text. In this paper we report on the reimplementation and experimental comparison of the two approaches to this task: Gordon's unigram classifier, and Corman's semantic triplet classifier. We cross-analyze their performance on both Gordon's and Corman's corpora, and discuss similarities, differences, and gaps in the performance of these classifiers, and point the way forward to improving their approaches.
The growing supply of online mental health tools, platforms and treatments results in an enormous quantity of digital narrative data to be structured, analysed and interpreted. Natural Language Processing is very suitable to automatically extract textual and structural features from narratives. Visualizing these features can help to explore patterns and shifts in text content and structure. In this study, streamgraphs are developed for different types of "Letters from the Future", an online mental health promotion instrument. The visualizations show differences between as well as within the different letter types, providing directions for future research in both the visualization of narrative structure and in the field of narrative psychology. The method presented here is not limited to "Letters from the Future", the current object of study, but can in fact be used to explore any digital or digitalized textual source, like books, speech transcripts or email conversations.
Models of narrative have been proposed from many perspectives and most of these nowadays promote further the notion that narrative is a transmedial phenomenon: i.e., stories can be told making use of distinct and multiple forms of expressions. This raises a range of theoretical and practical questions, as well as rendering the task of providing computational models of narrative both more interesting and more challenging. Central to this endeavour are issues concerned with the potential mutual conditioning of narrative forms and the media employed. Methods are required for isolating narrative properties and mechanisms that may be generalised across media, while at the same time appropriately respecting differences in medial affordances. In this discussion paper I set out a corresponding approach to characterising narrative that draws on a fine-grained formal characterisation of multimodal discourse developed on the basis of both functional and formal linguistic models of discourse, generalised to the multimodal case. After briefly setting out the theoretical principles on which the account builds, I position narrative with respect to the framework and give an example of how audiovisual narratives such as film are accounted for. It will be suggested that a common anchoring in a well specified notion of discourse as an intrinsically multimodal phenomenon offers beneficial new angles on how narratives can be modelled, as well as establishing bridges between humanistic understandings of narrative and complementary computational accounts of narratives involving communicative goal-based planning.
In this work we propose a model for the representation of the narrative of a literary text. The model is structured in an ontology and a lexicon constituting a knowledge base that can be queried by a system. This narrative ontology, as well as describing the actors, locations, situations found in the text, provides an explicit formal representation of the timeline of the story. We will focus on a specific case study, that of the representation of a selected portion of Homer's Odyssey, in particular of the knowledge required to answer a selection of salient queries, formulated by a literary scholar. This work is being carried out within the framework of the Semantic Web by adopting models and standards such as RDF, OWL, SPARQL, and lemon among others.
A story summarizer benefits greatly from a reader model because a reader model enables the story summarizer to focus on delivering useful knowledge in minimal time with minimal eort. Such a
This paper presents a linguistically uninformed computational model for animacy classification. The model makes use of word n-grams in combination with lower dimensional word embedding representations that are learned from a web-scale corpus. We compare the model to a number of linguistically informed models that use features such as dependency tags and show competitive results. We apply our animacy classifier to a large collection of Dutch folktales to obtain a list of all characters in the stories. We then draw a semantic map of all automatically extracted characters which provides a unique entrance point to the collection.
We report on building a computational model of romantic relationships in a corpus of historical literary texts. We frame this task as a ranking problem in which, for a given character, we try to assign the highest rank to the character with whom (s)he is most likely to be romantically involved. As data we use a publicly available corpus of French 17th and 18th century plays (http://www.theatre-classique.fr/) which is well suited for this type of analysis because of the rich markup it provides (e.g. indications of characters speaking). We focus on distributional, so-called second-order features, which capture how speakers are contextually embedded in the texts. At a mean reciprocal rate (MRR) of 0.9 and MRR@1 of 0.81, our results are encouraging, suggesting that this approach might be successfully extended to other forms of social interactions in literature, such as antagonism or social power relations.
We present a novel representation of narratives at the story level called Impulse. It combines a temporal representation of a story’s actions and events with a representation of the mental models of the story’s characters into a cohesive, logic-based language. We show the expressiveness of this approach by encoding a story fragment, and compare it to other formal story representations in terms of representational dimensions. We also acknowledge the computational complexity of our approach and argue that a restricted subset still provides a high degree of expressive power
A narrative world can be viewed as a form of society in which characters follow a set of social norms whose collective function is to guide the characters through (the creation of) a story arc and reach some conclusion. By modelling the rules of a narrative using norms, we can govern the actions of agents that act out the characters in a story. Agents are given sets of permitted actions and obligations to fulfil based on their and the story's current situation. However, a way to describe stories in terms of social norms is needed. Existing formalisms for narrative do not work at multiple layers of abstraction, and do not provide a rich enough vocabulary for describing stories. We use story tropes as a means of building reusable story components with which we describe the social norms that govern our storyworld agents.
Human experiences are stored in episodic memory and are the basis for developing semantic narrative structures and many of the narratives we continually compose. Episodic memory has only recently been recognized as a necessary module in general cognitive architectures and little work has been done to examine how the data stored by these modules may be formulated as narrative structures. This paper regards episodic memory as fundamental to narrative intelligence and considers the gap between simple episodic memory representations and narrative structures, and proposes an approach to generating basic narratives from episodic sequences. An approach is outlined considering the Soar general cognitive architecture and Zacks’ Event Segmentation Theory.
