Narrative event segmentation in the cortical reservoir

Recent research has revealed that during continuous perception of movies or stories, humans display cortical activity patterns that reveal hierarchical segmentation of event structure. Thus, sensory areas like auditory cortex display high frequency segmentation related to the stimulus, while semantic areas like posterior middle cortex display a lower frequency segmentation related to transitions between events. These hierarchical levels of segmentation are associated with different time constants for processing. Likewise, when two groups of participants heard the same sentence in a narrative, preceded by different contexts, neural responses for the groups were initially different and then gradually aligned. The time constant for alignment followed the segmentation hierarchy: sensory cortices aligned most quickly, followed by mid-level regions, while some higher-order cortical regions took more than 10 seconds to align. These hierarchical segmentation phenomena can be considered in the context of processing related to comprehension. In a recently described model of discourse comprehension word meanings are modeled by a language model pre-trained on a billion word corpus. During discourse comprehension, word meanings are continuously integrated in a recurrent cortical network. The model demonstrates novel discourse and inference processing, in part because of two fundamental characteristics: real-world event semantics are represented in the word embeddings, and these are integrated in a reservoir network which has an inherent gradient of functional time constants due to the recurrent connections. Here we demonstrate how this model displays hierarchical narrative event segmentation properties beyond the embeddings alone, or their linear integration. The reservoir produces activation patterns that are segmented by a hidden Markov model (HMM) in a manner that is comparable to that of humans. Context construction displays a continuum of time constants across reservoir neuron subsets, while context forgetting has a fixed time constant across these subsets. Importantly, virtual areas formed by subgroups of reservoir neurons with faster time constants segmented with shorter events, while those with longer time constants preferred longer events. This neurocomputational recurrent neural network simulates narrative event processing as revealed by the fMRI event segmentation algorithm provides a novel explanation of the asymmetry in narrative forgetting and construction. The model extends the characterization of online integration processes in discourse to more extended narrative, and demonstrates how reservoir computing provides a useful model of cortical processing of narrative structure.

Narrative Event Segmentation in the Cortical Reservoir

During continuous perception of movies or stories, awake humans display cortical activity patterns that reveal hierarchical segmentation of event structure. Sensory areas like auditory cortex display high frequency segmentation related to the stimulus, while semantic areas like posterior middle cortex display a lower frequency segmentation related to transitions between events (Baldassano et al. 2017). These hierarchical levels of segmentation are associated with different time constants for processing. Chien and Honey (2020) observed that when two groups of participants heard the same sentence in a narrative, preceded by different contexts, neural responses for the groups were initially different and then gradually aligned. The time constant for alignment followed the segmentation hierarchy: sensory cortices aligned most quickly, followed by mid-level regions, while some higher-order cortical regions took more than 10 seconds to align. These hierarchical segmentation phenomena can be considered in the context of processing related to comprehension. Uchida et al. (2021) recently described a model of discourse comprehension where word meanings are modeled by a language model pre-trained on a billion word corpus (Yamada et al 2020). During discourse comprehension, word meanings are continuously integrated in a recurrent cortical network. The model demonstrates novel discourse and inference processing, in part because of two fundamental characteristics: real-world event semantics are represented in the word embeddings, and these are integrated in a reservoir network which has an inherent gradient of functional time constants due to the recurrent connections. Here we demonstrate how this model displays hierarchical narrative event segmentation properties. The reservoir produces activation patterns that are segmented by the HMM of Baldassano et al (2017) in a manner that is comparable to that of humans. Context construction displays a continuum of time constants across reservoir neuron subset, while context forgetting has a fixed time constant across these subsets. Virtual areas formed by subgroups of reservoir neurons with faster time constants segmented with shorter events, while those with longer time constants preferred longer events. This neurocomputational recurrent neural network simulates narrative event processing as revealed by the fMRI event segmentation algorithm of Baldassano et al (2017), and provides a novel explanation of the asymmetry in narrative forgetting and construction observed by Chien and Honey (2020). The model extends the characterization of online integration processes in discourse to more extended narrative, and demonstrates how reservoir computing provides a useful model of cortical processing of narrative structure.

Elkövetőből áldozat – a csoportközi konfliktusok narratív konstrukciójának percepciója*

Háttér és célkitűzések:Tanulmányunkban azokat a narratív eseménykonstrukciós eszközöket vizsgáljuk, amelyek összefüggésbe hozhatók a kollektív áldozati tudat (Bar-Tal, Chernyak-Hai, Schori és Gundar, 2009) közvetítésével és fenntartásával. Igazolni kívánjuk, hogy a László (2012) által felvázolt áldozati narratív kompozíciós eszközök (nyelvi ágencia, értékelés, és pszichológiai perspektíva csoportelfogult használata) révén közvetíthető egy csoport áldozati pozíciója. Továbbá megvizsgáljuk, hogy egy csoport konfl iktustörténetének percepcióját képes-e megváltoztatni a narratív kompozíció kísérleti úton történő manipulálása: lehetséges-e elkövetőből áldozatot kreálni pusztán a nyelvi megszerkesztettség útján?Módszer:A társas észlelési paradigmára épülő vizsgálatban nemzeti csoportok áldozati történeteinek szisztematikus nyelvi manipulációján keresztül kialakított elfogult és elfogulatlan változatát megítélve, kérdőíves módszerrel (Egyéni és Csoportvélekedés Skála, Eidelson, 2009) mértük fel a narratív kompozíciós eszközök észlelésre gyakorolt hatását.Eredmények:Az áldozati narratívum kompozíciós eszközei statisztikai értelemben is hatással voltak a bemutatott csoportok áldozati pozíciójának észlelésére. A csoportok megítélése attól függően változott, hogy a résztvevők melyik szövegváltozatot olvasták: az elfogulatlan eseményleírás esetén az „áldozati” csoport, az elfogult változat esetén az „agresszor” észlelt áldozati pozíciója válik hangsúlyosabbá. Egyúttal azt is sikerült bizonyítani, hogy pusztán a nyelvi megszerkesztettség útján megváltoztatható egy agresszor csoport észlelése, és áldozati színezettel is bemutathatók tetteik.Következtetések:A László és munkatársai által leírt narratív kompozíció közvetíti az áldozati tudattal összefüggő hiedelmeket, és a csoport szemantikus szerepe képes felülírni az objektíven meghatározott cselekményszerepeket.Background and goals:In this paper we explore the narrative event-constructional devices that can be linked to the transmission and sustainment of collective victim consciousness (Bar-Tal, Chernyak-Hai, Schori, and Gundar, 2009). Our goal is to verify that with the narrative compositional devices (linguistic agency, evaluation, group-biased use of psychological perspective) described by László (2012), a group’s victim position can be transmitted. It is further explored, if the perception of a group’s confl ict-story can be altered by the experimental manipulation of the narrative composition: is it possible to create a victim from a perpetrator by just the linguistic composition?Method:The study is based on the social perception paradigm, in which biased and unbiased variants of national groups’ victimhood stories were created through systematic linguistic manipulation. The effect of the narrative compositional devices on the perception of the stories was evaluated with a questionnaire (Individual- and Group Beliefs Scale, Eidelson, 2009).Results:The narrative compositional devices of the victimhood narrative had a statistically signifi cant effect on the perception of the introduced groups’ victimhood position. The evaluation of the groups changed according to which variant of the story was introduced: in the case of the unbiased event-description, the „victim” group’s victim position is salient; and in the case of the biased event-description, the „perpetrator” group’s victim position becomes more salient. In addition, it is demonstrated that the perception of a perpetrator group can be changed by only the narrative construction and their actions can acquire a „victim tone”.Conclusion:The narrative compositional devices described by László et al. transmit the beliefs linked to victimhood consciousness, and the group’s semantic role can overwrite the objectively defi ned roles.

Positioning Selves in Narrative Accounts of Military Trauma

Linguistic and narrative strategies employed in two dyadic interviews of male veterans of the war in Afghanistan were analysed and compared. Each interviewee told chains of connected stories that positioned them in relation to catastrophic events and their effects. These incidents were framed as being linked to decisions that the teller had taken in perilous circumstances. Sequences of generic clauses in sections of orientation were used to manage knowledge asymmetries, to establish story world norms, to display professional, soldierly and veteran identities and to present danger, serious injury, and death as normative in the context of military work in a conflict zone. Resolutions to narratives involving death or injury resided not in specific narrative event clauses but in sections of evaluation that framed the outcome of a preceding story chain in terms of its personal, current significance to the speaker. Our exploration of contrasting accounts of similarly catastrophic events by two storytellers, one of whom was diagnosed with posttraumatic stress disorder, offers insights into ways in which trauma is represented that could be useful in psychotherapeutic contexts.

The parallelism of the characters in the literature class in the 10th grade: Andrei Bolkonsky and Nikolai Rostov

The purpose of the article is to present the methodology of work in the school with such a compositional principle as parallelism of characters. The introductory part focuses on how the term “parallelism” is interpreted in Russian Philology. In the article the understanding of the term is presented in linguo-stylistic and architectonic contexts. Further, by the example of L.N. Tolstoy’s novel “War and peace” the author reveals the working procedure of paralleling such characters as Andrei Bolkonsky and Nikolai Rostov, the characters that practically do not interfere with each other in the narrative-event sphere of the novel. at the same time, special attention is paid to the comparative analysis in the following areas: perception of boundless natural space; perception of music; ideal and its failure; sense of justice, honesty; sense of guilt; perception of people in war; attitude to death. The logic of the training session dictates the comparison of episodes in a given perspective. Greater part of them were considered earlier at the stage of the textual analysis of the literary text. It should be noted that in the lesson there are in the synthesis two types of analysis – determining and functional, which inevitably “work” on the logic of interpretation. On the basis of the results of the comparative activity, the students reveal the artistic meaning of the parallelism of the images of Andrei Bolkonsky and Nikolai Rostov in the novel. At the lesson presented in the article, the principles and techniques of the classical methodology are kept current, which largely ensures the effectiveness of cognitive activity of schoolchildren.

Aspect and Narrative Event Segmentation

Time is central to human cognition, both in terms of how we understand the world and the events that unfold around us as well as how we communicate about those events. As such, language has morphological systems, such as temporal adverbs, tense, and aspect to convey the passage of time. The current study explored the role of one such temporal marker, grammatical aspect, and its impact on how we understand the temporal boundaries between events conveyed in narratives. In Experiments 1 and 2, participants read stories that contained a target event that was either conveyed with a perfective (e.g., watched a movie) or imperfective aspect (e.g., was watching a movie) and engaged in an event segmentation task. Events described in the perfective aspect were more often perceived as event boundaries than events in the imperfective aspect, however, event duration (long vs. short) did not impact this relationship in Experiment 2. Experiment 3 demonstrated that readers were sensitive to grammatical aspect and event duration in the context of a story continuation task. Overall this study demonstrates that grammatical aspect interacts with world knowledge to convey event structure information that influences how people interpret the end and beginning of events.

Temporal-order Iconicity Bias in Narrative Event Understanding and Memory

Incongruence between the narrated (encoded) order and the actual chronological order of events is ubiquitous in various kinds of narratives and information modalities. The iconicity assumption in text comprehension proposes that readers will by default assume the chronological order to match the narrated order. However, it is not clear whether this iconicity assumption would directly bias the inferred chronology of events and the memory of the narrated order. In the current study, using non-linearly narrated video narratives as encoding materials, we dissociated the narrated order and the underlying chronological order of events. In Experiment 1, we found that participants’ judgments of the chronological order of events were biased by the narrated order, but not vice versa. In Experiment 2, when the chronological positions of events were provided during encoding, participants’ judgments of the chronological order were not biased by the narrated order, rather, their memory of the narrated order of events was biased by the chronological order. Interpreting the bias under a descriptive Bayesian framework, we offer a new perspective on the role of the iconicity assumption as prior belief, apart from prior knowledge about event sequences, in event understanding as well as memory.

Constructing Narrative Event Evolutionary Graph for Script Event Prediction

Script event prediction requires a model to predict the subsequent event given an existing event context. Previous models based on event pairs or event chains cannot make full use of dense event connections, which may limit their capability of event prediction. To remedy this, we propose constructing an event graph to better utilize the event network information for script event prediction. In particular, we first extract narrative event chains from large quantities of news corpus, and then construct a narrative event evolutionary graph (NEEG) based on the extracted chains. NEEG can be seen as a knowledge base that describes event evolutionary principles and patterns. To solve the inference problem on NEEG, we present a scaled graph neural network (SGNN) to model event interactions and learn better event representations. Instead of computing the representations on the whole graph, SGNN processes only the concerned nodes each time, which makes our model feasible to large-scale graphs. By comparing the similarity between input context event representations and candidate event representations, we can choose the most reasonable subsequent event. Experimental results on widely used New York Times corpus demonstrate that our model significantly outperforms state-of-the-art baseline methods, by using standard multiple choice narrative cloze evaluation.