[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-84090-en":3,"doc-seo-84090-105":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},84090,1099514067438,"River Wang","https://ap-avatar.wpscdn.com/avatar/100002539ee87300030?x-image-process=image/resize,m_fixed,w_180,h_180&k=1780474512215547542",8,"Research & Report","From Sinhala to Dhivehi Cross-Lingual Transfer Learning for Low-Resource Speech Recognition","Dhivehi, the national language of the Maldives, lacks sufficient data for automatic speech recognition (ASR), limiting NLP development for its speakers. The study examines whether cross-lingual transfer learning from Sinhala, a related higher-resource Insular Indo-Aryan language, improves Dhivehi ASR. Seventeen experiments cover Dhivehi-only baselines, sequential fine-tuning, multilingual fine-tuning, continual pretraining, and a Turkish unrelated-language control. Continual pretraining on Sinhala followed by Dhivehi fine-tuning with KenLM achieves 12.89% WER and 2.70% CER, exceeding the Dhivehi-only baseline.","From Sinhala to Dhivehi: Cross-Lingual Transfer Learning for Low-Resource Speech Recognition  \nLukmal Ilyas∗ , Nevidu Jayatilleke†  \n∗ School of Computing, Informatics Institute of Technology, Sri Lanka  \n[ilyas.20221732@iit.ac.lk](ilyas.20221732@iit.ac.lk)  \n†Department of Computer Science & Engineering, University of Moratuwa, Sri Lanka  \n[nevidu.25@cse.mrt.ac.lk](nevidu.25@cse.mrt.ac.lk)  \narXiv :2607 .06289v 1 [ cs .CL] 7 Jul 2026  \nAbstract—Dhivehi, the national language of the Maldives, is currently under-resourced for automatic speech recognition (ASR) and other NLP tasks. This study investigates whether cross-lingual transfer learning from Sinhala, a linguistically related, relatively well-resourced Insular Indo-Aryan language, can improve Dhivehi ASR. We conduct seventeen experiments across five transfer learning paradigms: Dhivehi-only baselines, sequential fine-tuning, multilingual fine-tuning, continual pretraining, and a control using Turkish as an unrelated language. The strongest system, continual pre-training on Sinhala followed by fine-tuning on Dhivehi with KenLM, achieves 12.89% WERand 2.70% CER, outperforming the Dhivehi-only baseline by 13.50% WER and 3.02% CER. However, the adaptation strategy and decoding configuration are equally critical for a successful transfer learning experiment. We conduct seventeen controlled experiments spanning five transfer learning paradigms: Dhivehionly baselines, sequential fine-tuning, multilingual fine-tuning, continual pre-training, and a control experiment using Turkish as an unrelated language. The strongest system, continual pretraining on Sinhala followed by fine-tuning on Dhivehi with KenLM, achieves 12.89% WER and 2.70% CER, outperforming the Dhivehi-only baseline by 13.50% WER and 3.02% CER. The Turkish control experiment confirms that observed improvements stem from linguistic relatedness; adaptation strategy and decoding configuration are also critical.  \nIndex Terms—Automatic Speech Recognition, Cross-lingual Transfer Learning, Low-resource Languages, Dhivehi, Sinhala  \nI. INTRODUCTION  \nModern ASR systems achieve strong performance for highresource languages such as English and Mandarin, which benefit from large-scale speech datasets [1]. However, the majority of the world’s approximately 7,000 languages remain underrepresented in speech technology, restricting access to digital services, educational tools, and communication technologies for hundreds of millions of speakers [2] .  \nDhivehi (also called Maldivian), the national language of the Maldives [3], is spoken by roughly 335,000 to 410,000 people across the Maldives, Minicoy (India), and expatriate communities [4] . Mozilla Common Voice provides only 61 recorded hours of speech, of which 37 hours are validated for Dhivehi [5], far below the data volumes typically required for competitive ASR systems.  \nSelf-supervised models such as Wav2Vec [6] and multilingual extensions such as XLS-R [7] have substantially reduced labelled data requirements through large-scale pre-  \nTABLE I  \nDHIVEHI-SINHALA SENTENCE COMPARISONS SHOWING SHARED INDO-ARYAN COGNATES (ROMANISED) .  \n\n| Meaning | Dhivehi | Sinhala |\n| --- | --- | --- |\n| I look at the sea\u003Cbr>The village is far I eat fish My ear hurts The road is long | aharen muhudhu balaneegamu dhuru veyaharen mas kaneekan dhanee\u003Cbr>magu dhigu | mama muhuda balanava¯ gama durayi\u003Cbr>mama mas kanava¯ mage kana ridenava¯ maga digayi |\n\ntraining on raw audio. Fine-tuning on limited target data often underperforms because pretrained encoders lack exposure to the target language’s acoustic characteristics. Cross-lingual transfer leveraging data from a related, higher-resource language offers a complementary strategy that has shown promise across several language families [8–10] .  \nDhivehi and Sinhala both belong to the Insular Indo-Aryan subgroup, sharing phonological, lexical, and structural features documented in comparative linguistic work [4] . Figure 1 illustrates their g","cbCaidG8uCnQ3CWp","https://ap.wps.com/l/cbCaidG8uCnQ3CWp","pdf",505849,1,7,"English","en",105,"# Introduction\n## Problem of low-resource ASR for Dhivehi\n## Related-language transfer: Sinhala to Dhivehi\n## Research questions\n# Method Overview\n## Transfer learning paradigms","[{\"question\":\"Why is Dhivehi ASR considered a low-resource problem?\",\"answer\":\"Dhivehi has limited publicly available transcribed speech data, with Mozilla Common Voice providing only 37 validated hours for Dhivehi—far below the data volumes typically needed for competitive ASR.\"},{\"question\":\"What transfer learning approaches does the study compare for improving Dhivehi ASR?\",\"answer\":\"The study evaluates five paradigms: Dhivehi-only baselines, sequential fine-tuning, multilingual fine-tuning, continual pretraining, and a control experiment using Turkish as an unrelated language.\"},{\"question\":\"Which system yields the best ASR results and what are its metrics?\",\"answer\":\"Continual pretraining on Sinhala followed by fine-tuning on Dhivehi with KenLM performs best, reaching 12.89% WER and 2.70% CER, improving over the Dhivehi-only baseline by 13.50% WER and 3.02% CER.\"}]",1784192678,18,{"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},"from-sinhala-to-dhivehi-cross-lingual-transfer-learning-for-low-resource-speech-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/from-sinhala-to-dhivehi-cross-lingual-transfer-learning-for-low-resource-speech-recognition/84090/",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},"Why is Dhivehi ASR considered a low-resource problem?","Question",{"text":74,"@type":75},"Dhivehi has limited publicly available transcribed speech data, with Mozilla Common Voice providing only 37 validated hours for Dhivehi—far below the data volumes typically needed for competitive ASR.","Answer",{"name":77,"@type":72,"acceptedAnswer":78},"What transfer learning approaches does the study compare for improving Dhivehi ASR?",{"text":79,"@type":75},"The study evaluates five paradigms: Dhivehi-only baselines, sequential fine-tuning, multilingual fine-tuning, continual pretraining, and a control experiment using Turkish as an unrelated language.",{"name":81,"@type":72,"acceptedAnswer":82},"Which system yields the best ASR results and what are its metrics?",{"text":83,"@type":75},"Continual pretraining on Sinhala followed by fine-tuning on Dhivehi with KenLM performs best, reaching 12.89% WER and 2.70% CER, improving over the Dhivehi-only baseline by 13.50% WER and 3.02% CER.","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},[92,96,100,104,109,114,118,121,126,129,133],{"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":110,"doc_module":4,"doc_module_name":45,"category_name":111,"show_sort_weight":112,"slug":113},6,"Technology",50,"technology",{"id":21,"doc_module":4,"doc_module_name":45,"category_name":115,"show_sort_weight":116,"slug":117},"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"]