[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-seo-239055-105":3,"detail-sidebar-cat-1-en-105":81,"doc-detail-239055-en":126},{"code":4,"msg":5,"data":6},0,"ok",{"site_id":7,"language":8,"slug":9,"title":10,"keywords":11,"description":12,"schema_data":13,"social_meta":74,"head_meta":76,"extra_data":78,"updated_unix":80},105,"en","tweetter-a-benchmark-for-target-entity-retrieval-on-twitter-without-knowledge-bases","TweetTER: A Benchmark for Target Entity Retrieval on Twitter without Knowledge Bases","","Entity linking connects entity mentions to knowledge-base entries, yet standard evaluation assumes models can represent all entities and that a comprehensive knowledge base is available. This constraint becomes impractical in social media, where noisy contexts and massive candidate inventories make conventional benchmarks less suitable. TweetTER (Tweet Target Entity Retrieval) reframes entity linking as a binary retrieval task, evaluating whether a provided entity definition matches the target entity in an input tweet. The approach enables knowledge-base-free assessment across zero-shot, few-shot, and fine-tuning settings, revealing strong performance for encoder fine-tuning while highlighting challenges for in-context learning.",{"@graph":14,"@context":73},[15,34,56],{"@type":16,"itemListElement":17},"BreadcrumbList",[18,23,27,31],{"item":19,"name":20,"@type":21,"position":22},"https://docshare.wps.com","Home","ListItem",1,{"item":24,"name":25,"@type":21,"position":26},"https://docshare.wps.com/template/","Template",2,{"item":28,"name":29,"@type":21,"position":30},"https://docshare.wps.com/template/presentations/","Presentations",3,{"item":32,"name":10,"@type":21,"position":33},"https://docshare.wps.com/template/tweetter-a-benchmark-for-target-entity-retrieval-on-twitter-without-knowledge-bases/239055/",4,{"url":32,"name":10,"@type":35,"image":36,"author":41,"headline":10,"publisher":44,"fileFormat":47,"inLanguage":8,"description":12,"dateModified":48,"datePublished":49,"encodingFormat":47,"isAccessibleForFree":50,"interactionStatistic":51},"DigitalDocument",{"url":37,"@type":38,"width":39,"height":40},"https://docshare.wps.com/thumbnails/tweetter-a-benchmark-for-target-entity-retrieval-on-twitter-without-knowledge-bases/239055.png","ImageObject",442,249,{"name":42,"@type":43},"Mabel","Person",{"url":19,"name":45,"@type":46},"DocShare","Organization","application/pdf","2026-10-29","2026-09-11",true,{"@type":52,"interactionType":53,"userInteractionCount":55},"InteractionCounter",{"@type":54},"ViewAction",12,{"@type":57,"mainEntity":58},"FAQPage",[59,65,69],{"name":60,"@type":61,"acceptedAnswer":62},"What problem does TweetTER address in entity linking for social media?","Question",{"text":63,"@type":64},"TweetTER addresses the practical mismatch of conventional entity linking benchmarks that require a complete knowledge base and assume models understand all candidate entities, which is often unrealistic for noisy social media like Twitter.","Answer",{"name":66,"@type":61,"acceptedAnswer":67},"How does TweetTER reframe entity linking?",{"text":68,"@type":64},"It reframes entity linking as a binary target entity retrieval task, where the model decides whether an entity definition matches the target entity mentioned in an input tweet context.",{"name":70,"@type":61,"acceptedAnswer":71},"How is TweetTER evaluated and what do the results suggest?",{"text":72,"@type":64},"Language models are evaluated on TweetTER using zero-shot, few-shot, and full fine-tuning. Results show successful application of language models, with encoder-based fine-tuning performing best overall and in-context learning approaches proving more challenging.","https://schema.org",{"og:url":32,"og:type":75,"og:title":10,"og:site_name":45,"og:description":12},"article",{"robots":77,"canonical":32},"index,follow",{"doc_id":79,"site_id":7},239055,1791724932,{"code":4,"msg":82,"data":83},"success",[84,88,92,97,102,107,112,117,122],{"id":85,"doc_module":22,"doc_module_name":25,"category_name":29,"show_sort_weight":86,"slug":87},11,90,"presentations",{"id":55,"doc_module":22,"doc_module_name":25,"category_name":89,"show_sort_weight":90,"slug":91},"Resumes",80,"resumes",{"id":93,"doc_module":22,"doc_module_name":25,"category_name":94,"show_sort_weight":95,"slug":96},14,"Invoices",70,"invoices",{"id":98,"doc_module":22,"doc_module_name":25,"category_name":99,"show_sort_weight":100,"slug":101},15,"Posters",60,"posters",{"id":103,"doc_module":22,"doc_module_name":25,"category_name":104,"show_sort_weight":105,"slug":106},16,"Social Media",50,"social-media",{"id":108,"doc_module":22,"doc_module_name":25,"category_name":109,"show_sort_weight":110,"slug":111},17,"Forms",40,"forms",{"id":113,"doc_module":22,"doc_module_name":25,"category_name":114,"show_sort_weight":115,"slug":116},18,"Letters",30,"letters",{"id":118,"doc_module":22,"doc_module_name":25,"category_name":119,"show_sort_weight":120,"slug":121},21,"Paper Templates",5,"papers-templates",{"id":123,"doc_module":22,"doc_module_name":25,"category_name":124,"show_sort_weight":4,"slug":125},158,"General","general-158",{"code":4,"msg":82,"data":127},{"doc_id":79,"user_id":128,"nickname":42,"user_avatar":129,"doc_module":22,"category_id":85,"category_name":29,"doc_title":10,"doc_description":12,"doc_content":130,"file_id":131,"file_url":132,"file_type":133,"file_size":134,"view_count":55,"is_deleted":4,"is_public":22,"is_downloadable":22,"audit_status":22,"page_count":135,"language":136,"language_code":8,"site_id":7,"html_lang":8,"table_of_contents":137,"faqs":138,"seo_title":139,"seo_description":12,"update_tm":140,"read_time":30},7971474920318,"https://ap-avatar.wpscdn.com/avatar/74000ee537e6b0ca360?x-image-process=image/resize,m_fixed,w_180,h_180&k=1788430171632181674","TweetTER: A Benchmark for Target Entity Retrieval on Twitter  \nwithout Knowledge Bases  \nKiamehr Rezaee, Jose Camacho-Collados, Mohammad Taher Pilehvar  \nCardiff NLP, School of Computer Science and Informatics  \nCardiff University, United Kingdom  \n{rezaeek, camachocolladosj, [pilehvarmt}@cardiff.ac.uk](pilehvarmt}@cardiff.ac.uk)  \nAbstract  \nEntity linking is a well-established task in NLP consisting of associating entity mentions with entries in a knowledge base. Current models have demonstrated competitive performance in standard text settings. However, when it comes to noisy domains such as social media, certain challenges still persist. Typically, to evaluate entity linking on existing benchmarks, a comprehensive knowledge base is necessary and models are expected to possess an understanding of all the entities contained within the knowledge base. However, in practical scenarios where the objective is to retrieve sentences specifically related to a particular entity, strict adherence to a complete understanding of all entities in the knowledge base may not be necessary. To address this gap, we introduce TweetTER (Tweet Target Entity Retrieval), a novel benchmark that aims to bridge the challenges in entity linking. The distinguishing feature of this benchmark is its approach of re-framing entity linking as a binary entity retrieval task. This enables the evaluation of language models’ performance without relying on a conventional knowledge base, providing a more practical and versatile evaluation framework for assessing the effectiveness of language models in entity retrieval tasks.  \nKeywords: Entity Linking, Twitter, Target Entity Retrieval  \n1. Introduction  \nWord Sense Disambiguation (Navigli, 2009, WSD) and Entity Linking (Ling et al. , 2015, EL) are two long-standing tasks in language understanding. In these tasks, the primary goal is to associate a given word with either a specific sense in a sense inventory (e.g., WordNet for WSD) or an entity in a knowledge base (e.g., Wikipedia for EL) . However, relying on specific underlying inventories poses certain limitations on the types of models that can be effectively evaluated. It assumes that the models have the ability to store and process information about the entire knowledge base, which may not always be practical or feasible due to the sheer size and complexity of such resources. Knowledge bases such as Wikidata1 are composed of tens of millions of entities, resulting in a substantial number of plausible choices for a given entity, many of which might be either outdated or refer to unconventional usage of the entity. The high granularity often makes the task of entity linking too challenging or even impractical, particularly in noisy settings such as social media (Liu et al. , 2013) .  \nHowever, in realistic scenarios, understanding the specific entity of interest is often sufficient to retrieve relevant sentences from a text corpus. For example, a company launching a new product with an ambiguous name may seek to analyze its impact on social media. In such cases, there is no requirement for a model to encode knowledge about all potential candidate entities. Instead, the focus should  \n1 [https://www.wikidata.org/](https://www.wikidata.org/)  \n\n| P | Context Definition | Obama published his memoirs last year. Former president of the USA |\n| --- | --- | --- |\n| N | Context Definition | I can’t wait to see the new Tom Hanks movie. Renowned physicist and Nobel laureate |\n\nTable 1: Example positive (P) and negative (N) samples for two target entities (underlined) .  \nbe on retrieving information related to a particular sense or instance of the entity, making it unnecessary to store knowledge about every possible entity in the model.  \nTo address these challenges, we propose an alternative formulation for the standard EL task. Inspired by the WiC-TSV dataset (Breit et al. , 2021), we put forward TweetTER (Tweet Target Entity Retrieval), a benchmark for entity linking ","cbCail4KYzmLtPbI","https://ap.wps.com/l/cbCail4KYzmLtPbI","pdf",178880,7,"English","# Abstract\n# 1. Introduction\n# 2. TweetTER: The Dataset\n## 2.1. TweetNERD\n## 2.2. Converting EL to Retrieval\n## 2.3. Evaluation Splits and Statistics","[{\"question\":\"What problem does TweetTER address in entity linking for social media?\",\"answer\":\"TweetTER addresses the practical mismatch of conventional entity linking benchmarks that require a complete knowledge base and assume models understand all candidate entities, which is often unrealistic for noisy social media like Twitter.\"},{\"question\":\"How does TweetTER reframe entity linking?\",\"answer\":\"It reframes entity linking as a binary target entity retrieval task, where the model decides whether an entity definition matches the target entity mentioned in an input tweet context.\"},{\"question\":\"How is TweetTER evaluated and what do the results suggest?\",\"answer\":\"Language models are evaluated on TweetTER using zero-shot, few-shot, and full fine-tuning. Results show successful application of language models, with encoder-based fine-tuning performing best overall and in-context learning approaches proving more challenging.\"}]","TweetTER: A Benchmark for Target Entity Retrieval on Twitter without Knowledge Bases | PDF",1789142953]