[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-seo-661390-105":3,"detail-sidebar-cat-0-en-105":80,"doc-detail-661390-en":130},{"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":73,"head_meta":75,"extra_data":77,"updated_unix":79},105,"en","word-embeddings-analogies-and-machine-learning-beyond-king-man-woman-queen","Word Embeddings, Analogies, and Machine Learning: Beyond King-Man + Woman = Queen","","Solving word analogies is a popular benchmark for word embeddings under the assumption that linear relations between word pairs (e.g., king:man :: woman:queen) reflect embedding quality. This work challenges that assumption, showing information missed by linear offset can be recovered through more advanced search and is encoded in the embedding. Simple averaging over multiple word pairs improves state-of-the-art, and combining cosine similarity with a candidate’s word-class estimate yields further gains, up to 30% for some embeddings and relations. The study also analyzes how embedding-analogy retrieval algorithms interact and how this affects evaluation and extrinsic usage.",{"@graph":14,"@context":72},[15,34,55],{"@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/document/","Document",2,{"item":28,"name":29,"@type":21,"position":30},"https://docshare.wps.com/document/technology/","Technology",3,{"item":32,"name":10,"@type":21,"position":33},"https://docshare.wps.com/document/word-embeddings-analogies-and-machine-learning-beyond-king-man-woman-queen/661390/",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/word-embeddings-analogies-and-machine-learning-beyond-king-man-woman-queen/661390.png","ImageObject",300,407,{"name":42,"@type":43},"Miles","Person",{"url":19,"name":45,"@type":46},"DocShare","Organization","application/pdf","2026-10-29","2026-10-28",true,{"@type":52,"interactionType":53,"userInteractionCount":22},"InteractionCounter",{"@type":54},"ViewAction",{"@type":56,"mainEntity":57},"FAQPage",[58,64,68],{"name":59,"@type":60,"acceptedAnswer":61},"What assumption about word analogies does the paper question?","Question",{"text":62,"@type":63},"It questions the assumption that linguistic relations are identifiable via linear vector offset between word pairs, such as king-man + woman = queen.","Answer",{"name":65,"@type":60,"acceptedAnswer":66},"How does the paper improve over linear offset methods?",{"text":67,"@type":63},"It proposes learning the target relation from multiple word pairs, using averaging as a strong baseline and a further machine-learning approach that combines cosine similarity with an estimate of candidate word-class membership.",{"name":69,"@type":60,"acceptedAnswer":70},"Why is the interaction between embeddings and analogy retrieval algorithms important?",{"text":71,"@type":63},"The paper argues that retrieval algorithms can interact with the embedding representations in ways that affect evaluation results, influencing both benchmark assessment and performance in extrinsic tasks.","https://schema.org",{"og:url":32,"og:type":74,"og:title":10,"og:site_name":45,"og:description":12},"article",{"robots":76,"canonical":32},"index,follow",{"doc_id":78,"site_id":7},661390,1793166173,{"code":4,"msg":81,"data":82},"success",[83,87,91,95,100,104,109,114,119,122,126],{"id":22,"doc_module":4,"doc_module_name":25,"category_name":84,"show_sort_weight":85,"slug":86},"Story & Novel",90,"story-novel",{"id":26,"doc_module":4,"doc_module_name":25,"category_name":88,"show_sort_weight":89,"slug":90},"Literature",80,"literature",{"id":33,"doc_module":4,"doc_module_name":25,"category_name":92,"show_sort_weight":93,"slug":94},"Exam",70,"exam",{"id":96,"doc_module":4,"doc_module_name":25,"category_name":97,"show_sort_weight":98,"slug":99},5,"Comic",60,"comic",{"id":101,"doc_module":4,"doc_module_name":25,"category_name":29,"show_sort_weight":102,"slug":103},6,50,"technology",{"id":105,"doc_module":4,"doc_module_name":25,"category_name":106,"show_sort_weight":107,"slug":108},7,"Healthcare",40,"healthcare",{"id":110,"doc_module":4,"doc_module_name":25,"category_name":111,"show_sort_weight":112,"slug":113},8,"Research & Report",30,"research-report",{"id":115,"doc_module":4,"doc_module_name":25,"category_name":116,"show_sort_weight":117,"slug":118},9,"Religion & Spirituality",20,"religion-spirituality",{"id":117,"doc_module":4,"doc_module_name":25,"category_name":120,"show_sort_weight":117,"slug":121},"World Cup","world-cup",{"id":123,"doc_module":4,"doc_module_name":25,"category_name":124,"show_sort_weight":123,"slug":125},10,"Lifestyle","lifestyle",{"id":127,"doc_module":4,"doc_module_name":25,"category_name":128,"show_sort_weight":96,"slug":129},19,"General","general",{"code":4,"msg":81,"data":131},{"doc_id":78,"user_id":132,"nickname":42,"user_avatar":133,"doc_module":4,"category_id":101,"category_name":29,"doc_title":10,"doc_description":12,"doc_content":134,"file_id":135,"file_url":136,"file_type":137,"file_size":138,"view_count":22,"is_deleted":4,"is_public":22,"is_downloadable":22,"audit_status":22,"page_count":139,"language":140,"language_code":8,"site_id":7,"html_lang":8,"table_of_contents":141,"faqs":142,"seo_title":143,"seo_description":12,"update_tm":79,"read_time":112},13056703019404,"https://ap-avatar.wpscdn.com/davatar_29158cc5080c5b710cf443261637dec0","Word Embeddings, Analogies, and Machine Learning:  \nBeyond King-Man + Woman = Queen  \nAleksandr Drozdy, Anna Gladkovaz, Satoshi Matsuokay  \ny Tokyo Institute of Technology, Meguro-ku, Tokyo 152-8550, Japan  \n[alex@smg.is.titech.ac.jp](alex@smg.is.titech.ac.jp) , [matsu@is.titech.ac.jp](matsu@is.titech.ac.jp)  \nz The University of Tokyo, Meguro-ku, Tokyo 153-8902 Japan [gladkova@phiz.c.u-tokyo.ac.jp](gladkova@phiz.c.u-tokyo.ac.jp)  \nAbstract  \nSolving word analogies became one of the most popular benchmarks for word embeddings on the assumption that linear relations between word pairs (such as king:man :: woman:queen) are indicative of the quality of the embedding. We question this assumption by showing that the information not detected by linear offset may still be recoverable by a more sophisticated search method, and thus is actually encoded in the embedding.  \nThe general problem with linear offset is its sensitivity to the idiosyncrasies of individual words. We show that simple averaging over multiple word pairs improves over the state-of-the-art. A further improvement in accuracy (up to 30% for some embeddings and relations) is achieved by combining cosine similarity with an estimation of the extent to which a candidate answer belongs to the correct word class. In addition to this practical contribution, this work highlights the problem of the interaction between word embeddings and analogy retrieval algorithms, and its implications for the evaluation of word embeddings and the use of analogies in extrinsic tasks.  \n1 Introduction  \nDiscovering analogical relations is currently one of the most popular benchmarks for word embeddings. This trend started after (Mikolov et al., 2013b) showed that proportional analogies (a is to b as c is to d) can be solved by finding the vector closest to the hypothetical vector calculated as c-a + b (e.g. king-man + woman = queen) . Many subsequent studies used this approach to evaluate the performance of word embeddings with the Google test set (Mikolov et al., 2013a); the top current result is over 80% accuracy (Pennington et al., 2014) . The assumption is that a “good” word embedding encodes linguistic relations in such a way that they are identifiable via linear vector offset (see section 2) .  \nAnalogies are interesting not only as a benchmark, but also potentially as a method for discovering linguistic relations (Turney, 2008) . They are already used for morphological analysis (Lavalle and Langlais, 2010), word sense disambiguation (Federici et al., 1997), semantic search (Cohen et al., 2015), and even for broad-range detection of both morphological and semantic features (Lepage and Goh, 2009) . However, Mikolov's study was a demonstration of how word embeddings capture linguistic relations, rather than a proposal of linear vector offset as a method for their discovery. It was later shown to not work as well for a wider range of relations (Kper et al., 2015; Gladkova et al., 2016) .  \nThis study questions the underlying assumption that linguistic relations should translate to linear relations between vectors rather than a more complex correspondence pattern. We show that relations not detected by vector offset may be recoverable by other methods, and thus are actually encoded in the embedding. The method we propose is based on learning the target relation from multiple word pairs, since reliance on single word pair makes linear vector offset sensitive to word idiosyncrasies. A naive average-based baseline outperforms the state-of-the-art. A more sophisticated machine-learning algorithm achieves further improvement (up to 30% for some embeddings and linguistic relations) by combining similarity to a source word vector (king) with the estimate of whether a candidate answer (queen) belongs to the correct class of words (“woman”) .  \nThis work is licensed under a Creative Commons Attribution 4 .0 International Licence. Licence details: [http://](http://)[ ](http://)[creativecommons.org/lice","cbCairCbRKE3nwmI","https://ap.wps.com/l/cbCairCbRKE3nwmI","pdf",404772,12,"English","# Introduction\n## Analogies as a benchmark\n## Limitations of linear vector offset\n## Proposed approach and contributions\n# State of the Art: Analogical Reasoning Based on the Offset of Word Vectors\n## 3CosAdd\n## PairDistance\n## Alternative scoring functions","[{\"question\":\"What assumption about word analogies does the paper question?\",\"answer\":\"It questions the assumption that linguistic relations are identifiable via linear vector offset between word pairs, such as king-man + woman = queen.\"},{\"question\":\"How does the paper improve over linear offset methods?\",\"answer\":\"It proposes learning the target relation from multiple word pairs, using averaging as a strong baseline and a further machine-learning approach that combines cosine similarity with an estimate of candidate word-class membership.\"},{\"question\":\"Why is the interaction between embeddings and analogy retrieval algorithms important?\",\"answer\":\"The paper argues that retrieval algorithms can interact with the embedding representations in ways that affect evaluation results, influencing both benchmark assessment and performance in extrinsic tasks.\"}]","Word Embeddings, Analogies, and Machine Learning: Beyond King-Man + Woman = Queen | PDF"]