[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-85189-en":3,"doc-seo-85189-105":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},85189,962075114101,"Seraphina","https://ap-avatar.wpscdn.com/avatar/e000253a75eb197efd?x-image-process=image/resize,m_fixed,w_180,h_180&k=1780044092746381165",8,"Research & Report","Multilingual Semantic Retrieval for Apple Music Search","Apple Music serves listeners across 150+ storefronts and rapidly expanding catalogs, making search recall for misspelled, transliterated, and cross-lingual queries a key driver of session quality—especially for tail queries that dominate unique requests. The work introduces ELISE, a 305M multilingual Siamese bi-encoder fine-tuned from GTE-multilingual-base using curriculum-scheduled multi-objective training. A hybrid dense-plus-token retrieval integration with quantile distribution matching is deployed without retraining downstream rankers, improving Hit@10 by 69% offline and lifting conversion rate by 2.28% online while cutting no-result rate by 86%. Gains concentrate on tail queries (+7.93% CR) without head regressions.","Multilingual Semantic Retrieval for Apple Music Search∗  \nVishalaksh Aggarwal†  \nApple Cupertino, CA, USA [vishalaksh@apple.com](vishalaksh@apple.com)  \nKevin Sebastian†  \nApple Cupertino, CA, USA [k_chittinappilyse@apple.com](k_chittinappilyse@apple.com)  \nVivek Kanojiya  \nApple Cupertino, CA, USA [vkanojiya@apple.com](vkanojiya@apple.com)  \nLeo Le  \nApple Cupertino, CA, USA [leo_le@apple.com](leo_le@apple.com)  \nNick Tucey  \nApple Cupertino, CA, USA[ntucey@apple.com](ntucey@apple.com)  \nSantosh Shankar  \nApple Cupertino, CA, USA [santosh_shankar@apple.com](santosh_shankar@apple.com)  \narXiv :2607 . 10239v 1 [ cs .IR] 11 Jul 2026  \nAbstract  \nApple Music serves listeners across 150+ storefronts in dozens of languages, with a catalog that grows by hundreds of thousands of new tracks daily. At this scale, search recall on misspelled, transliterated, and cross-lingual queries becomes a dominant driver of session quality, particularly for tail queries that account for the majority of unique queries. We present a multilingual semantic retrieval system built on a 305M-parameter Siamese bi-encoder fine-tuned from GTE-multilingual-base [18] with curriculum-scheduled multi-objective training. The model is integrated into the search stack via a hybrid retrieval architecture that blends dense nearest-neighbor results with the existing token-based index using quantile distribution matching, enabling deployment without retraining downstream rankers. Offline, the model achieves a 69% relative improvement in Hit@10 over GTE-multilingual-base. In a worldwide online A/B test, the system delivers a 2.28% relative conversion-rate (CR) lift overall, an 86% reduction in the no-result rate, and gains across every storefront with no observed regressions. The improvement is concentrated where it is needed most: tail queries see a 7.93% relative CR lift, compared with 0.89% for mid-frequency queries and 0.14% for head queries—evidence that semantic retrieval improves recall on hard queries without disturbing well-served popular ones. To our knowledge, this is one of the largest search-quality improvements deployed on the platform.  \nCCS Concepts  \n• Information systems → Music retrieval; Retrieval models and ranking; Information retrieval query processing.  \nKeywords  \nsemantic retrieval, multilingual search, music information retrieval, bi-encoder, approximate nearest neighbor, hybrid retrieval  \n1 Introduction  \nApple Music users search across a catalog that spans content types such as songs, albums, artists, playlists, and stations. The existing retrieval system uses a token-based index with language-specific  \n∗ Accepted to the Industry Track of the 20th ACM Conference on Recommender Systems (RecSys 2026) . This is the authors’ accepted manuscript; the final version will appear in the ACM Digital Library.  \n†Equal contribution.  \nanalyzers: query tokens are matched against normalized document metadata, and a downstream ranker re-scores the lexical candidates. This works well for head queries, where lexical overlap is high and engagement signals are abundant. However, our analysis of search sessions indicates that, among sessions with poor result quality, the majority are recall failures rather than ranking errors, with misspellings and other surface-form mismatches leading to poor or no recall being the dominant failure modes.  \nThese recall failures are concentrated in tail queries—queries with low historical frequency that account for 83% of unique queries and roughly one-third of all search sessions. The conversion rate1 is lower for tail queries than for head queries, where token-matching pipelines struggle because lexical overlap between query and intended item is minimal or absent. For example, a user searching for “gnarly by cat’s eye”(a misspelling of Gnarly by KATSEYE) finds nothing; a Persian-speaking user searching “ﺭ؇ዛኞ ܂ਜಾﺭ؇݁” for the song “Bahar” by Martik sees zero matches when translation coverage for that item is incomplete in th","cbCairXfrvfjuTQW","https://ap.wps.com/l/cbCairXfrvfjuTQW","pdf",172192,2,1,9,"English","en",105,"# Abstract\n# 1 Introduction\n## Challenges in multilingual music search\n## ELISE: multilingual semantic bi-encoder\n## Hybrid retrieval without ranker retraining\n## Tail-targeted deployment results","[{\"question\":\"Why does search quality decline for multilingual and tail queries in Apple Music?\",\"answer\":\"Quality declines because many failed sessions are recall failures caused by misspellings, surface-form mismatches, incomplete translation coverage, or queries not present in the catalog. These issues are concentrated in tail queries, which have low historical frequency and minimal lexical overlap with the intended items.\"},{\"question\":\"What is ELISE and how is it trained for music search?\",\"answer\":\"ELISE is a 305M-parameter Siamese bi-encoder fine-tuned from GTE-multilingual-base. It uses curriculum-scheduled, multi-objective training over Query–Item and Query–Query tasks, along with data strategies such as storefront frequency capping, synthetic misspelling augmentation, and world-knowledge injection.\"},{\"question\":\"How is the model integrated into the search stack without retraining rankers?\",\"answer\":\"The system uses a hybrid retrieval architecture that fuses dense approximate nearest neighbor (ANN) scores with the existing token-based index. Quantile distribution matching preserves the original token-match score distribution, enabling deployment without retraining downstream rankers.\"}]",1784201637,23,{"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},"multilingual-semantic-retrieval-for-apple-music-search","",{"@graph":36,"@context":85},[37,53,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,47,50],{"item":41,"name":42,"@type":43,"position":21},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":20},"https://docshare.wps.com/document/","Document",{"item":48,"name":12,"@type":43,"position":49},"https://docshare.wps.com/document/research-report/",3,{"item":51,"name":13,"@type":43,"position":52},"https://docshare.wps.com/document/multilingual-semantic-retrieval-for-apple-music-search/85189/",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","2026-07-23","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},"Why does search quality decline for multilingual and tail queries in Apple Music?","Question",{"text":75,"@type":76},"Quality declines because many failed sessions are recall failures caused by misspellings, surface-form mismatches, incomplete translation coverage, or queries not present in the catalog. These issues are concentrated in tail queries, which have low historical frequency and minimal lexical overlap with the intended items.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"What is ELISE and how is it trained for music search?",{"text":80,"@type":76},"ELISE is a 305M-parameter Siamese bi-encoder fine-tuned from GTE-multilingual-base. It uses curriculum-scheduled, multi-objective training over Query–Item and Query–Query tasks, along with data strategies such as storefront frequency capping, synthetic misspelling augmentation, and world-knowledge injection.",{"name":82,"@type":73,"acceptedAnswer":83},"How is the model integrated into the search stack without retraining rankers?",{"text":84,"@type":76},"The system uses a hybrid retrieval architecture that fuses dense approximate nearest neighbor (ANN) scores with the existing token-based index. Quantile distribution matching preserves the original token-match score distribution, enabling deployment without retraining downstream rankers.","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},[93,97,101,105,110,115,120,123,127,130,134],{"id":21,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":20,"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":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":22,"doc_module":4,"doc_module_name":46,"category_name":124,"show_sort_weight":125,"slug":126},"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"]