[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-85553-en":3,"doc-seo-85553-105":29,"detail-sidebar-cat-0-en-105":90},{"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":4,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":13,"seo_description":14,"update_tm":27,"read_time":28},85553,16904993612988,"Olivia Brown","https://ap-avatar.wpscdn.com/davatar_a8503ba1806abce46bf441b54a3ca4cd",8,"Research & Report","An Empirical Recipe for Universal Phone Recognition","Phone recognition (PR) is a key enabler for multilingual and low-resource speech processing, yet robustness across languages remains limited. English-focused PR models struggle to generalize, while multilingual approaches underutilize pretrained self-supervised learning representations and explore few training objectives. PhoneticXEUS is trained on large-scale multilingual data and delivers state-of-the-art results on multilingual speech (17.7% PFER) and accented English (10.6% PFER). Controlled ablations quantify effects of SSL representations, data scale, and loss objectives, with open data and code.","An Empirical Recipe for Universal Phone Recognition  \nShikhar Bharadwaj  1, Chin-Jou Li  1, Kwanghee Choi  2, Eunjung Yeo  2, William Chen  1,  \nShinji Watanabe  1, David R. Mortensen  1  \n1 Carnegie Mellon University, USA 2 The University of Texas at Austin, USA  \n{sbharad2,[dmortens](dmortens}@andrew.cmu.edu)[}](dmortens}@andrew.cmu.edu)[@andrew.cmu.edu](dmortens}@andrew.cmu.edu)  \narXiv :2603 .29042v2 [ cs .CL] 13 Jul 2026  \nAbstract  \nPhone recognition (PR) is a key enabler of multilingual and low-resource speech processing tasks, yet robust performance remains elusive. Highly performant English-focused models do not generalize across languages, while multilingual models underutilize pretrained representations. It also remains unclear how data scale, architecture, and training objective contribute to multilingual PR. We present PhoneticXEUS—trained on large-scale multilingual data and achieving state-of-the-art performance on both multilingual (17.7% PFER) and accented English speech (10.6% PFER) . Through controlled ablations with evaluations across 100+ languages under a unified scheme, we empirically establish our training recipe and quantify the impact of SSL representations, data scale, and loss objectives. In addition, we analyze error patterns across language families, accented speech, and articulatory features. All data and code are released openly.1  \nIndex Terms: speech recognition, phone recognition  \n1. Introduction  \nPhone Recognition (PR) enables important multilingual speech processing technologies, especially for zero text-resource languages [1–3] via cross-lingual transfer [4–6] . PR is also heavily employed in atypical speech assessment [7, 8], computerassisted language learning [9–11] and linguistic fieldwork [12– 14] . For English PR, several systems have been developed since the 1950s [15, 16], with recent systems [17, 18] pushing performance by leveraging larger and better-quality datasets. However, their generalization to multilingual settings is limited (Figure 1) . On the other hand, recent progress has led to a proliferation of multilingual PR models and benchmarks [19–21], yet these models still leave performance potential on the table by not fully leveraging self-supervised learning (SSL) based representations and by only exploring a narrow selection of training objectives. Stated succinctly, while English PR models do not generalize well to diverse languages, recent multilingual PR research has scarcely explored the space of possible training methods. To address this gap, we conduct extensive empirical experiments and establish a training recipe for a PR system that attains state-of-the-art performance in both unseen multilingual settings and on accented English.  \nWhile prior approaches to multilingual PR train on a limited number of languages [25–28] or treated PR as a tool to explore unsupervised ASR (Wav2Vec2Phoneme) [24], recent methods [19,20] have shown that scaling to diverse datasets [29] automatically generated via Grapheme-to-Phoneme (G2P) can significantly improve multilingual performance, allowing appli-  \n1 [https://github.com/changelinglab/PhoneticXeus](https://github.com/changelinglab/PhoneticXeus).  \nP FER  \n16  \n15  \n14  \n13  \n12  \n11  \n10  \nAccented English  \nMultiIPA W2V2P-LV60  \nPOWSM-CTCW2V2P-XLSR53 ZIPA-CTC  \nZIPA-CTC-NS  \nPhoneticXEUS  \n23  \n22  \n21  \n20  \n19  \n18  \n17  \nMultilingual  \nMultiIPA W2V2P-LV60  \nPOWSM-CTCW2V2P-XLSR53 ZIPA-CTC  \nZIPA-CTC-NS  \nPhoneticXEUS  \nFigure 1: PhoneticXEUS achieves SOTA performance on both accented English and multilingual speech. Details in Table 1.  \ncations to real-world use cases [30] . However, more recent approaches do not utilize the capabilities of pretrained representations [31, 32] .  \nFurthermore, the architecture and training objectives of prior systems have been chosen based their specific goals: ZIPA [19] uses Zipformer [33] with CR-CTC [34] for training efficiency, and POWSM [20] uses CTC-Attention joint training in an au","cbCaiqIBGArpR57j","https://ap.wps.com/l/cbCaiqIBGArpR57j","pdf",309985,1,6,"English","en",105,"# Introduction\n## Background and Motivation\n## Gap in Existing Multilingual Phone Recognition\n## Proposed Approach and Contributions","[{\"question\":\"What problem does PhoneticXEUS address in universal phone recognition?\",\"answer\":\"PhoneticXEUS targets the lack of robust cross-language generalization in phone recognition, especially when English-focused models fail to generalize and multilingual models do not fully leverage self-supervised learning representations or diverse training objectives.\"},{\"question\":\"How does PhoneticXEUS evaluate performance across multilingual and accented English settings?\",\"answer\":\"It reports state-of-the-art results on both multilingual speech (17.7% PFER) and accented English (10.6% PFER). Evaluations use a unified scheme and controlled ablations across 100+ languages.\"},{\"question\":\"What does the document claim about the contribution of SSL representations, data scale, and training objectives?\",\"answer\":\"The training recipe is established through controlled ablations that empirically quantify how SSL representations, data scale, and loss objectives affect multilingual phone recognition performance.\"}]",1784204490,15,{"code":4,"msg":30,"data":31},"ok",{"site_id":24,"language":23,"slug":32,"title":13,"keywords":33,"description":14,"schema_data":34,"social_meta":85,"head_meta":87,"extra_data":89,"updated_unix":27},"an-empirical-recipe-for-universal-phone-recognition","",{"@graph":35,"@context":84},[36,53,67],{"@type":37,"itemListElement":38},"BreadcrumbList",[39,43,47,50],{"item":40,"name":41,"@type":42,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":44,"name":45,"@type":42,"position":46},"https://docshare.wps.com/document/","Document",2,{"item":48,"name":12,"@type":42,"position":49},"https://docshare.wps.com/document/research-report/",3,{"item":51,"name":13,"@type":42,"position":52},"https://docshare.wps.com/document/an-empirical-recipe-for-universal-phone-recognition/85553/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":23,"description":14,"dateModified":61,"datePublished":61,"encodingFormat":60,"isAccessibleForFree":62,"interactionStatistic":63},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":40,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-16",true,{"@type":64,"interactionType":65,"userInteractionCount":4},"InteractionCounter",{"@type":66},"ViewAction",{"@type":68,"mainEntity":69},"FAQPage",[70,76,80],{"name":71,"@type":72,"acceptedAnswer":73},"What problem does PhoneticXEUS address in universal phone recognition?","Question",{"text":74,"@type":75},"PhoneticXEUS targets the lack of robust cross-language generalization in phone recognition, especially when English-focused models fail to generalize and multilingual models do not fully leverage self-supervised learning representations or diverse training objectives.","Answer",{"name":77,"@type":72,"acceptedAnswer":78},"How does PhoneticXEUS evaluate performance across multilingual and accented English settings?",{"text":79,"@type":75},"It reports state-of-the-art results on both multilingual speech (17.7% PFER) and accented English (10.6% PFER). Evaluations use a unified scheme and controlled ablations across 100+ languages.",{"name":81,"@type":72,"acceptedAnswer":82},"What does the document claim about the contribution of SSL representations, data scale, and training objectives?",{"text":83,"@type":75},"The training recipe is established through controlled ablations that empirically quantify how SSL representations, data scale, and loss objectives affect multilingual phone recognition performance.","https://schema.org",{"og:url":51,"og:type":86,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":88,"canonical":51},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":91},[92,96,100,104,109,113,118,121,126,129,133],{"id":20,"doc_module":4,"doc_module_name":45,"category_name":93,"show_sort_weight":94,"slug":95},"Story & Novel",90,"story-novel",{"id":46,"doc_module":4,"doc_module_name":45,"category_name":97,"show_sort_weight":98,"slug":99},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":45,"category_name":101,"show_sort_weight":102,"slug":103},"Exam",70,"exam",{"id":105,"doc_module":4,"doc_module_name":45,"category_name":106,"show_sort_weight":107,"slug":108},5,"Comic",60,"comic",{"id":21,"doc_module":4,"doc_module_name":45,"category_name":110,"show_sort_weight":111,"slug":112},"Technology",50,"technology",{"id":114,"doc_module":4,"doc_module_name":45,"category_name":115,"show_sort_weight":116,"slug":117},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":119,"slug":120},30,"research-report",{"id":122,"doc_module":4,"doc_module_name":45,"category_name":123,"show_sort_weight":124,"slug":125},9,"Religion & Spirituality",20,"religion-spirituality",{"id":124,"doc_module":4,"doc_module_name":45,"category_name":127,"show_sort_weight":124,"slug":128},"World Cup","world-cup",{"id":130,"doc_module":4,"doc_module_name":45,"category_name":131,"show_sort_weight":130,"slug":132},10,"Lifestyle","lifestyle",{"id":134,"doc_module":4,"doc_module_name":45,"category_name":135,"show_sort_weight":105,"slug":136},19,"General","general"]