[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-81523-en":3,"doc-seo-81523-105":30,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":21,"is_downloadable":21,"audit_status":21,"page_count":22,"language":23,"language_code":24,"site_id":25,"html_lang":24,"table_of_contents":26,"faqs":27,"seo_title":13,"seo_description":14,"update_tm":28,"read_time":29},81523,1099513958762,"Logic","https://ap-avatar.wpscdn.com/avatar/1000023916a998db790?x-image-process=image/resize,m_fixed,w_180,h_180&k=1784791008015729253",8,"Research & Report","Beyond LLMs A Linguistic Approach to Causal Graph Generation from Narrative Texts","Proposes a linguistic, hybrid framework for generating causal graphs from narrative texts, targeting fine-grained event-specific causality beyond high-level statements. The method first extracts agent-centered “vertices” via LLM-based summarization, then introduces an Expert Index of seven linguistically grounded features and integrates them into a STAC (Situation, Task, Action, Consequence) classification model. Using RoBERTa embeddings plus the Expert Index, the system improves causal-link precision over LLM-only baselines and iteratively prompts to build connected graphs. Experiments on 100 chapters and short stories show consistent gains over GPT-4o and Claude 3.5 while keeping readability, yielding an open-source, interpretable tool for causal chain mining.","Beyond LLMs: A Linguistic Approach to Causal Graph Generation from  \nNarrative Texts  \nZehan Li Ruhua Pan Xinyu Pi  \nUniversity of California, San Diego  \n{zel025, r3pan, [xpi}@ucsd.edu](xpi}@ucsd.edu)  \narXiv :2504 .07459v2 [ cs .CL] 10 Jul 2026  \nAbstract  \nWe propose a novel framework to generate causal graphs from narrative texts, bridging the gap between high-level causality and finergrained event-specific relationships. Our approach first extracts concise, agent-centered“vertices” using an LLM-based summarization strategy. We then introduce an Expert Index—seven linguistically grounded features—and incorporate them into a STAC (Situation, Task, Action, Consequence) classification model. This hybrid system (RoBERTa embeddings + Expert Index) achieves superior precision in identifying causal links compared to LLM-only baselines. Finally, we apply a structured, five-iteration prompting process torefine and construct a connected causal graph. Experiments on 100 chapters and short stories show that our method consistently outperforms GPT-4o and Claude 3.5 across key dimensions of causal graph quality, while maintaining comparable readability. The resulting open-source tool offers an interpretable and efficient solution for capturing nuanced causal chains within narrative texts.  \n1 Introduction  \nCausal research has historically leveraged knowledge graphs to explore relationships between events (JM;, 1999) . Modern approaches, such as AI-driven causal graph generation, have gained prominence for their ability to summarize causal events at scale (Jaimini and Sheth, 2022 ; Pieperet al., 2023) . However, current AI models largely focus on high-level causality (e.g., \"HIV leads to AIDS\"), and they fall short in capturing nuanced causal relationships in specific narratives, such as political events or historical occurrences(Donnelly, 2025) . Addressing this gap, we propose a method for generating causal graphs from texts that describe discrete, event-specific narratives.  \nUnderstanding these finer-grained causal relationships is crucial for researchers and practition-  \ners who analyze how certain events lead to tangible outcomes in areas like social movements, policy-making, and historical trends. By capturing causal links from narrative texts, stakeholders can more accurately trace the chain of events that precipitate significant changes, enabling better decision-making, deeper historical insight, and more targeted interventions. Furthermore, automated causal graph generation facilitates scalable analysis of large document collections, providing structured representations that can be easily interpreted, queried, and expanded upon.  \nMost existing methods for generating causal graphs follow a two-stage pipeline: (1) a Causality Finder to detect causal relations, and (2) a graph Generator to construct knowledge graphs from these relations. While effective, these methods face limitations in interpretability and accuracy, particularly when dealing with complex sentence structures or implicit causal links (Kyono et al., 2024) .  \nCausality finders have evolved through three phases: (1) early pattern-based models that learned causal relationships from fixed sentence structures (Hidey and McKeown, 2016) (Heindorf et al., 2020) ,(2) BERT-based approaches that addressed issues in text training but failed to account for semantic context (Tan et al., 2023) (Dasgupta et al., 2018) (Li et al., 2020), and (3) LLMs, which improved contextual reasoning but struggled to distinguish intricate causal relationships (Kıcıman et al., 2024) (Shen et al., 2022) (Luo et al., 2024) .  \nIn this paper, we present a novel framework that leverages linguistic feature extraction to enhance causal graph generation from narrative texts. Our approach introduces a Quaternary Classification system to categorize sentences into four components: (1) Situation,(2) Task,(3) Action, and (4) Consequences. This structured decomposition allows for more precise identificati","cbCaimAtvZJWvZHo","https://ap.wps.com/l/cbCaimAtvZJWvZHo","pdf",682010,3,1,16,"English","en",105,"# Introduction\n## Problem Setting","[{\"question\":\"What is the main goal of the proposed framework?\",\"answer\":\"To generate causal graphs from narrative texts by bridging high-level causality and more precise, event-specific causal relationships within narratives.\"},{\"question\":\"How does the framework represent causal structure?\",\"answer\":\"It extracts agent-centered vertices from the text, then classifies sentence components using a STAC model (Situation, Task, Action, Consequence) and uses this to identify causal links and construct a connected graph.\"},{\"question\":\"How is causality defined when building edges in the graph?\",\"answer\":\"Event A causes Event B if, in combination with other factors, Event A is a necessary or sufficient condition for Event B, or if the occurrence of Event A raises the probability of Event B occurring.\"}]",1784173985,40,{"code":4,"msg":31,"data":32},"ok",{"site_id":25,"language":24,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":28},"beyond-llms-a-linguistic-approach-to-causal-graph-generation-from-narrative-texts","",{"@graph":36,"@context":85},[37,53,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,50],{"item":41,"name":42,"@type":43,"position":21},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":20},"https://docshare.wps.com/document/research-report/",{"item":51,"name":13,"@type":43,"position":52},"https://docshare.wps.com/document/beyond-llms-a-linguistic-approach-to-causal-graph-generation-from-narrative-texts/81523/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":24,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":41,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-25","2026-07-16",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"What is the main goal of the proposed framework?","Question",{"text":75,"@type":76},"To generate causal graphs from narrative texts by bridging high-level causality and more precise, event-specific causal relationships within narratives.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How does the framework represent causal structure?",{"text":80,"@type":76},"It extracts agent-centered vertices from the text, then classifies sentence components using a STAC model (Situation, Task, Action, Consequence) and uses this to identify causal links and construct a connected graph.",{"name":82,"@type":73,"acceptedAnswer":83},"How is causality defined when building edges in the graph?",{"text":84,"@type":76},"Event A causes Event B if, in combination with other factors, Event A is a necessary or sufficient condition for Event B, or if the occurrence of Event A raises the probability of Event B occurring.","https://schema.org",{"og:url":51,"og:type":87,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":89,"canonical":51},"index,follow",{"doc_id":7,"site_id":25},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,119,122,127,130,134],{"id":21,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":46,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":46,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":29,"slug":118},7,"Healthcare","healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":120,"slug":121},30,"research-report",{"id":123,"doc_module":4,"doc_module_name":46,"category_name":124,"show_sort_weight":125,"slug":126},9,"Religion & Spirituality",20,"religion-spirituality",{"id":125,"doc_module":4,"doc_module_name":46,"category_name":128,"show_sort_weight":125,"slug":129},"World Cup","world-cup",{"id":131,"doc_module":4,"doc_module_name":46,"category_name":132,"show_sort_weight":131,"slug":133},10,"Lifestyle","lifestyle",{"id":135,"doc_module":4,"doc_module_name":46,"category_name":136,"show_sort_weight":106,"slug":137},19,"General","general"]