[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"detail-sidebar-cat-0-en-105":3,"doc-seo-156751-105":59,"doc-detail-156751-en":131},{"code":4,"msg":5,"data":6},0,"success",[7,13,18,23,28,33,38,43,48,51,55],{"id":8,"doc_module":4,"doc_module_name":9,"category_name":10,"show_sort_weight":11,"slug":12},1,"Document","Story & Novel",90,"story-novel",{"id":14,"doc_module":4,"doc_module_name":9,"category_name":15,"show_sort_weight":16,"slug":17},2,"Literature",80,"literature",{"id":19,"doc_module":4,"doc_module_name":9,"category_name":20,"show_sort_weight":21,"slug":22},4,"Exam",70,"exam",{"id":24,"doc_module":4,"doc_module_name":9,"category_name":25,"show_sort_weight":26,"slug":27},5,"Comic",60,"comic",{"id":29,"doc_module":4,"doc_module_name":9,"category_name":30,"show_sort_weight":31,"slug":32},6,"Technology",50,"technology",{"id":34,"doc_module":4,"doc_module_name":9,"category_name":35,"show_sort_weight":36,"slug":37},7,"Healthcare",40,"healthcare",{"id":39,"doc_module":4,"doc_module_name":9,"category_name":40,"show_sort_weight":41,"slug":42},8,"Research & Report",30,"research-report",{"id":44,"doc_module":4,"doc_module_name":9,"category_name":45,"show_sort_weight":46,"slug":47},9,"Religion & Spirituality",20,"religion-spirituality",{"id":46,"doc_module":4,"doc_module_name":9,"category_name":49,"show_sort_weight":46,"slug":50},"World Cup","world-cup",{"id":52,"doc_module":4,"doc_module_name":9,"category_name":53,"show_sort_weight":52,"slug":54},10,"Lifestyle","lifestyle",{"id":56,"doc_module":4,"doc_module_name":9,"category_name":57,"show_sort_weight":24,"slug":58},19,"General","general",{"code":4,"msg":60,"data":61},"ok",{"site_id":62,"language":63,"slug":64,"title":65,"keywords":66,"description":67,"schema_data":68,"social_meta":124,"head_meta":126,"extra_data":128,"updated_unix":130},105,"en","on-the-pursuit-of-fake-newsgraph-neural-network-meets-nlp","On The Pursuit of Fake News:Graph Neural Network meets NLP","","This paper describes methods from the FakeINA team for participating in FakeNews:Corona Virus and Conspiracies MultimediaAnalysis tasks, focusing on text-based misinformation and conspiracy detection. A multimodal neural network is introduced, combining a graph neural network (GIN) that represents each document as an undirected word co-occurrence graph with a multilayer perceptron using textual statistics as features. Experimental results show that while GNN alone can classify effectively, the multimodal approach achieves better performance.",{"@graph":69,"@context":123},[70,84,106],{"@type":71,"itemListElement":72},"BreadcrumbList",[73,77,79,82],{"item":74,"name":75,"@type":76,"position":8},"https://docshare.wps.com","Home","ListItem",{"item":78,"name":9,"@type":76,"position":14},"https://docshare.wps.com/document/",{"item":80,"name":40,"@type":76,"position":81},"https://docshare.wps.com/document/research-report/",3,{"item":83,"name":65,"@type":76,"position":19},"https://docshare.wps.com/document/on-the-pursuit-of-fake-newsgraph-neural-network-meets-nlp/156751/",{"url":83,"name":65,"@type":85,"image":86,"author":91,"headline":65,"publisher":94,"fileFormat":97,"inLanguage":63,"description":67,"dateModified":98,"datePublished":99,"encodingFormat":97,"isAccessibleForFree":100,"interactionStatistic":101},"DigitalDocument",{"url":87,"@type":88,"width":89,"height":90},"https://docshare.wps.com/thumbnails/on-the-pursuit-of-fake-newsgraph-neural-network-meets-nlp/156751.png","ImageObject",300,407,{"name":92,"@type":93},"Riley","Person",{"url":74,"name":95,"@type":96},"DocShare","Organization","application/pdf","2026-10-22","2026-08-29",true,{"@type":102,"interactionType":103,"userInteractionCount":105},"InteractionCounter",{"@type":104},"ViewAction",13,{"@type":107,"mainEntity":108},"FAQPage",[109,115,119],{"name":110,"@type":111,"acceptedAnswer":112},"What tasks and problem setting does the paper address?","Question",{"text":113,"@type":114},"The paper targets FakeNews:Corona Virus and Conspiracies MultimediaAnalysis, specifically the text-based misinformation detection and text-based conspiracy theories recognition sub-tasks.","Answer",{"name":116,"@type":111,"acceptedAnswer":117},"How is the text converted into a graph for the GNN?",{"text":118,"@type":114},"Each document is represented as an undirected graph where words are nodes and edges reflect word co-occurrences computed via a sliding window.",{"name":120,"@type":111,"acceptedAnswer":121},"What is the multimodal model and why does it improve results?",{"text":122,"@type":114},"The multimodal model combines GIN graph embeddings with textual statistics extracted via a text statistics package, feeding concatenated representations into an MLP. Results indicate the multimodal system outperforms using GNN classification alone.","https://schema.org",{"og:url":83,"og:type":125,"og:title":65,"og:site_name":95,"og:description":67},"article",{"robots":127,"canonical":83},"index,follow",{"doc_id":129,"site_id":62},156751,1787965702,{"code":4,"msg":5,"data":132},{"doc_id":129,"user_id":133,"nickname":92,"user_avatar":134,"doc_module":4,"category_id":39,"category_name":40,"doc_title":65,"doc_description":67,"doc_content":135,"file_id":136,"file_url":137,"file_type":138,"file_size":139,"view_count":105,"is_deleted":4,"is_public":8,"is_downloadable":8,"audit_status":8,"page_count":81,"language":140,"language_code":63,"site_id":62,"html_lang":63,"table_of_contents":141,"faqs":142,"seo_title":143,"seo_description":67,"update_tm":130,"read_time":39,"preview_page_count":4},1374391975076,"https://ap-avatar.wpscdn.com/avatar/14000253ca4ec9f6853?x-image-process=image/resize,m_fixed,w_180,h_180&k=1783305029341752051","# On The Pursuit of Fake News:Graph Neural Network meets NLP\n\nZeynep Pehlivan  \nInstitut national de I'audiovisuel,Francezpehlivan@ina.fr  \nCEUR-WS°org/Vol-3181/paper61.pdf  \n## ABSTRACT\n\nThis paper presents the methods proposed by FakeINA teamto par-ticipate The FakeNews:Corona Virus and Conspiracies MultimediaAnalysis tasks.We concentrate our work on text-based misinforma-tion and conspiracy detection.We proposed a multimodal neuralnetwork that combines a graph neural network (GNN)where adocument is represented as graph and a multi-layer perceptronmodel where textual statistics are used as features.Experimentalresults show that however GNNs are able to classify the data,amultimodal performs better.  \n## 1 INTRODUCTION\n\nMediaeval Fake News task[6,7]focuses on the classification oftweet texts aiming detection of fast spreading misinformation.Thistask contains three sub-tasks:Text-Based Misinformation Detec-tion,Text-Based Conspiracy Theories Recognition and Text-BasedCombined Misinformation and Conspiracies Detection.This workproposed a multimodal neural network approach which is onlyapplied to the first two sub-tasks.  \nGraph neural network(GNN)methods have been profoundlyuseful in several domains including natural language processing[9].While it is probably most apparent to regard text as sequentialdata,there are several methods to represent text as various kindsof graphs.Dependency graph construction generates a graph byextracting the dependency relations from the dependency parsingtree.Constituency graph construction captures phrase-based syn-tactic relations in a sentence.Another way of representing textas a graph is to use word co-occurrence and/or document wordrelations.  \nTextGCN[11]proposes to builda single heterogeneous graph forwhole corpus and captures global word co-occurrence information.This approache converts text classification task to node classifica-tion task.On the other hand,TextING[12]builds a graph for eachdocument based on the co-occurrence of words where each node isrepresented as a word embedding and sliding window is used tocapture the relation between words.It learns the fine-grained wordrepresentation of the local structure by GNNto effectively produceembeddings for obscure words in the new text.By representingeach document as agraph,text classification task becomes graphclassification task for GNNs.  \nWe present our approach in Section 2 and we discuss the resultsin Section 3.  \nFigure 1:Our framework of GIN with 2 layers for graphclassification  \n## 2 APPROACH\n\nIn this section,we present the preprocessing applied to the rawtext before graph construction.Two models proposed and theirimplementation details are also presented in this section.  \n### 2.1 Preprocessing\n\nWe use the spaCy[4]library for Python in order to create a text filecontaining the original tokens and their normalized counterparts.For this normalization we have transformed each token to lowercaseand removed stop words.Also,considering the fact that BERTallows a maximum of 512tokens per sequence and the given datasetcontains sentences above that range,Bert tokenizer is used withtruncation option.However,BERT is recommended to use withthe raw text,lemmatization and stemming remain important togenerate a graph since a node(word)would be disrupted by anirrelevant inflection like a simple plural.  \n### 2.2 Graph Construction\n\nWe use the same graph construction approach as described inTextING[12].Each document is represented as a undirected graphwhere nodes are words and co-occurrences between words repre-sented as edges.The co-occurrences is calculated by using a slidingwindow.Embedding ofthe nodes are initialized by extracting wordembeddings from BERT model[2].  \n### 2.3 Models\n\nAfter graph construction,the task converts to the graph classifi-cation task.The main idea behind GNNs is to compute a state foreach node and update this state according to neighbouring nodesstates at each iteration.Graph Isomorphism Network(GIN)[10]wasproposed ","cbCaieKuo5crCD3f","https://ap.wps.com/l/cbCaieKuo5crCD3f","pdf",747683,"English","# ABSTRACT\n# 1 INTRODUCTION\n# 2 APPROACH\n## 2.1 Preprocessing\n## 2.2 Graph Construction\n## 2.3 Models\n## 2.4 Implementation","[{\"question\":\"What tasks and problem setting does the paper address?\",\"answer\":\"The paper targets FakeNews:Corona Virus and Conspiracies MultimediaAnalysis, specifically the text-based misinformation detection and text-based conspiracy theories recognition sub-tasks.\"},{\"question\":\"How is the text converted into a graph for the GNN?\",\"answer\":\"Each document is represented as an undirected graph where words are nodes and edges reflect word co-occurrences computed via a sliding window.\"},{\"question\":\"What is the multimodal model and why does it improve results?\",\"answer\":\"The multimodal model combines GIN graph embeddings with textual statistics extracted via a text statistics package, feeding concatenated representations into an MLP. Results indicate the multimodal system outperforms using GNN classification alone.\"}]","On The Pursuit of Fake News:Graph Neural Network meets NLP | PDF"]