[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"detail-sidebar-cat-0-en-105":3,"doc-seo-273523-105":59,"doc-detail-273523-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","from-entity-mentions-to-tone-an-llm-based-pipeline-for-media-bias-analysis","From Entity Mentions to Tone - An LLM-Based Pipeline for Media Bias Analysis","","This paper introduces an LLM-based pipeline for analyzing media bias and framing in online news. The workflow organizes articles into topics and events, enriches them with named-entity and sentiment annotations, and compares outlets using people-mention tone, source-level patterns, and event-level coverage signals. Evaluation on 8,358 Albanian news articles from GDELT tests agreement with automated annotations and contrasts two prompt strategies for sentiment validation, then provides example framing and gatekeeping indicators.",{"@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/from-entity-mentions-to-tone-an-llm-based-pipeline-for-media-bias-analysis/273523/",{"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/from-entity-mentions-to-tone-an-llm-based-pipeline-for-media-bias-analysis/273523.png","ImageObject",300,407,{"name":92,"@type":93},"Hazel","Person",{"url":74,"name":95,"@type":96},"DocShare","Organization","application/pdf","2026-10-18","2026-09-15",true,{"@type":102,"interactionType":103,"userInteractionCount":105},"InteractionCounter",{"@type":104},"ViewAction",11,{"@type":107,"mainEntity":108},"FAQPage",[109,115,119],{"name":110,"@type":111,"acceptedAnswer":112},"What does the proposed pipeline analyze in online news?","Question",{"text":113,"@type":114},"It analyzes media bias and framing by tracking named entities, sentiment, and how mentions and coverage patterns vary across sources at both source and event levels.","Answer",{"name":116,"@type":111,"acceptedAnswer":117},"How are articles compared to identify differences between news outlets?",{"text":118,"@type":114},"The pipeline groups content into topics and events, then compares people-mention tone, source-level tendencies, and event-level coverage and gatekeeping indicators across outlets.",{"name":120,"@type":111,"acceptedAnswer":121},"What is the impact of using stricter sentiment-validation rules in annotation prompts?",{"text":122,"@type":114},"Stricter rules reduce label-score inconsistencies, but they increase execution time and reduce annotation coverage; the simpler prompt is used for the remaining analysis.","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},273523,1789471864,{"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":39,"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":46,"preview_page_count":4},137441390410,"https://ap-avatar.wpscdn.com/avatar/2000252f4ab5702993?_k=1776741390130283984","From Entity Mentions to Tone: An LLM-Based Pipeline for Media Bias Analysis  \nKlesti Hoxha and Olti Qirici  \nDepartment of Informatics  \nFaculty of Natural Sciences  \nUniversity of Tirana  \nTirana, Albania  \n{klesti.hoxha, [olti.qirici}@unitir.edu.al](olti.qirici}@unitir.edu.al)  \narXiv : 2608 . 17454v 1 [cs .CL] 18 Aug 2026  \nAbstract—This paper presents a pipeline for analyzing media bias and framing in online news. The pipeline groups articles into topics and events, adds named-entity and sentiment annotations, and compares news sources through people mentions, sourcelevel tone, and event-level coverage patterns. We apply it to 8,358 Albanian news articles collected from GDELT and compare the resulting annotations with GDELT’s automated annotations. The results show moderate agreement for sentiment and entity extraction, as well as additional person-entity pairs that can potentially support the bias analysis. We compare two annotation prompts and find that stricter sentiment-validation rules remove label-score inconsistencies but increase execution time and reduce annotation coverage. Based on these results, the simpler prompt is used for the rest of the analysis. We have provided sample analysis on source-level framing profiles, person-level tone differences across sources, and event-level gatekeeping and coverage indicators. These outputs show how the same news collection can be used to examine what sources cover, how they describe public figures, and where coverage is concentrated. The approach is particularly useful in settings where manually verified datasets or specialized language tools are limited.  \nKeywords—media bias, media framing, large language models, named-entity recognition, sentiment analysis, news analysis  \nI. INTRODUCTION  \nOnline news providers have become the default way for many people to be informed. They can be accessed directly or, quite often, through social media platforms. In both cases, recommender-system algorithms [1] boost specific news items or topics [2], making readers vulnerable to disinformation and selective representations of events or facts [3] . Selection may focus on individuals, organizations, or specific topics covered in the news [3], while framing can be related to the tone used when covering people, organizations, or topics in specific news items [4] .  \nAuthor’s accepted manuscript of K. Hoxha and O. Qirici,“From Entity Mentions to Tone: An LLM-Based Pipeline for Media Bias Analysis,” in 2026 18th International Conference on Electronics, Computers and Artificial Intelligence (ECAI), 2026, pp. 1–8, doi: 10.1109/ECAI69016.2026.11613638 .© 2026 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.  \n©2026 IEEE  \nTraditionally, this problem has been addressed through manual investigations of news providers by communication experts [5] . However, this process is heavyweight, costly, and itself prone to bias. Furthermore, it is difficult to apply continuously at the level of individual media providers or news articles.  \nAs an alternative, machine-learning algorithms, such as classifiers, have been used to identify bias or framing in media content [6] . However, these methods also rely on languagespecific training data, which are usually created by human annotators.  \nTo address gatekeeping and reduce disinformation [7] in light of recent advances in news dissemination technologies, reliable and fact-centered tools are needed to continuously assess bias and framing at the level of individual news providers and articles. Such tools are able to enhance news aggregator applications by enabling the visualization of these assessments.  \nThis need is especially visible ","cbCaibMGJBKbnEay","https://ap.wps.com/l/cbCaibMGJBKbnEay","pdf",187722,"English","# I. INTRODUCTION\n## Motivation and challenges in smaller-language settings\n## Need for multi-signal media-bias assessment\n## Role of multilingual LLMs and contributions","[{\"question\":\"What does the proposed pipeline analyze in online news?\",\"answer\":\"It analyzes media bias and framing by tracking named entities, sentiment, and how mentions and coverage patterns vary across sources at both source and event levels.\"},{\"question\":\"How are articles compared to identify differences between news outlets?\",\"answer\":\"The pipeline groups content into topics and events, then compares people-mention tone, source-level tendencies, and event-level coverage and gatekeeping indicators across outlets.\"},{\"question\":\"What is the impact of using stricter sentiment-validation rules in annotation prompts?\",\"answer\":\"Stricter rules reduce label-score inconsistencies, but they increase execution time and reduce annotation coverage; the simpler prompt is used for the remaining analysis.\"}]","From Entity Mentions to Tone - An LLM-Based Pipeline for Media Bias Analysis | PDF"]