[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"detail-sidebar-cat-0-en-105":3,"doc-seo-273702-105":59,"doc-detail-273702-en":131},{"code":4,"msg":5,"data":6},0,"success",[7,13,18,23,28,33,38,43,48,51,55],{"id":8,"doc_module":4,"doc_module_name":9,"category_name":10,"show_sort_weight":11,"slug":12},1,"Document","Story & Novel",90,"story-novel",{"id":14,"doc_module":4,"doc_module_name":9,"category_name":15,"show_sort_weight":16,"slug":17},2,"Literature",80,"literature",{"id":19,"doc_module":4,"doc_module_name":9,"category_name":20,"show_sort_weight":21,"slug":22},4,"Exam",70,"exam",{"id":24,"doc_module":4,"doc_module_name":9,"category_name":25,"show_sort_weight":26,"slug":27},5,"Comic",60,"comic",{"id":29,"doc_module":4,"doc_module_name":9,"category_name":30,"show_sort_weight":31,"slug":32},6,"Technology",50,"technology",{"id":34,"doc_module":4,"doc_module_name":9,"category_name":35,"show_sort_weight":36,"slug":37},7,"Healthcare",40,"healthcare",{"id":39,"doc_module":4,"doc_module_name":9,"category_name":40,"show_sort_weight":41,"slug":42},8,"Research & Report",30,"research-report",{"id":44,"doc_module":4,"doc_module_name":9,"category_name":45,"show_sort_weight":46,"slug":47},9,"Religion & Spirituality",20,"religion-spirituality",{"id":46,"doc_module":4,"doc_module_name":9,"category_name":49,"show_sort_weight":46,"slug":50},"World Cup","world-cup",{"id":52,"doc_module":4,"doc_module_name":9,"category_name":53,"show_sort_weight":52,"slug":54},10,"Lifestyle","lifestyle",{"id":56,"doc_module":4,"doc_module_name":9,"category_name":57,"show_sort_weight":24,"slug":58},19,"General","general",{"code":4,"msg":60,"data":61},"ok",{"site_id":62,"language":63,"slug":64,"title":65,"keywords":66,"description":67,"schema_data":68,"social_meta":124,"head_meta":126,"extra_data":128,"updated_unix":130},105,"en","building-and-evaluating-a-synthetic-bengali-speech-resource-for-telecom-customer-care","Building and Evaluating a Synthetic Bengali Speech Resource for Telecom Customer Care","","Speech systems in customer-facing apps depend on reliable coverage of domain-specific language and intents. This work presents a synthetic Bengali speech dataset for telecom customer-care scenarios, including 10,000 audio-text pairs totaling about 26.82 hours at 24 kHz, with fixed train/validation/test splits. Speech is generated via OmniVoice voice cloning and supplied with both original and normalized transcripts for ASR/STT. Evaluation reports WER 2.54% and CER 0.59%, plus reported limitations and ethics.",{"@graph":69,"@context":123},[70,84,106],{"@type":71,"itemListElement":72},"BreadcrumbList",[73,77,79,82],{"item":74,"name":75,"@type":76,"position":8},"https://docshare.wps.com","Home","ListItem",{"item":78,"name":9,"@type":76,"position":14},"https://docshare.wps.com/document/",{"item":80,"name":40,"@type":76,"position":81},"https://docshare.wps.com/document/research-report/",3,{"item":83,"name":65,"@type":76,"position":19},"https://docshare.wps.com/document/building-and-evaluating-a-synthetic-bengali-speech-resource-for-telecom-customer-care/273702/",{"url":83,"name":65,"@type":85,"image":86,"author":91,"headline":65,"publisher":94,"fileFormat":97,"inLanguage":63,"description":67,"dateModified":98,"datePublished":99,"encodingFormat":97,"isAccessibleForFree":100,"interactionStatistic":101},"DigitalDocument",{"url":87,"@type":88,"width":89,"height":90},"https://docshare.wps.com/thumbnails/building-and-evaluating-a-synthetic-bengali-speech-resource-for-telecom-customer-care/273702.png","ImageObject",300,407,{"name":92,"@type":93},"Skyler","Person",{"url":74,"name":95,"@type":96},"DocShare","Organization","application/pdf","2026-10-29","2026-09-15",true,{"@type":102,"interactionType":103,"userInteractionCount":105},"InteractionCounter",{"@type":104},"ViewAction",15,{"@type":107,"mainEntity":108},"FAQPage",[109,115,119],{"name":110,"@type":111,"acceptedAnswer":112},"What dataset is proposed for telecom customer-care, and what does it contain?","Question",{"text":113,"@type":114},"The work introduces a synthetic Bengali speech dataset for telecom customer-care scenarios. It includes 10,000 audio-text pairs, about 26.82 hours of 24 kHz audio, and predefined train/validation/test splits.","Answer",{"name":116,"@type":111,"acceptedAnswer":117},"How is the synthetic speech generated and what text annotations are provided?",{"text":118,"@type":114},"Speech is generated with OmniVoice in voice-cloning mode using a real female reference recording and transcript. The dataset provides the original Bengali text and a normalized transcript field for ASR/STT training and evaluation.",{"name":120,"@type":111,"acceptedAnswer":121},"How is intelligibility evaluated and what performance metrics are reported?",{"text":122,"@type":114},"An automatic intelligibility check is performed over all samples using a domain-adapted Whisper ASR model fine-tuned for Bengali telecom/customer-care. The reported average WER is 2.54% and average CER is 0.59%, with median WER/CER values reported as 0.00%.","https://schema.org",{"og:url":83,"og:type":125,"og:title":65,"og:site_name":95,"og:description":67},"article",{"robots":127,"canonical":83},"index,follow",{"doc_id":129,"site_id":62},273702,1789472569,{"code":4,"msg":5,"data":132},{"doc_id":129,"user_id":133,"nickname":92,"user_avatar":134,"doc_module":4,"category_id":39,"category_name":40,"doc_title":65,"doc_description":67,"doc_content":135,"file_id":136,"file_url":137,"file_type":138,"file_size":139,"view_count":105,"is_deleted":4,"is_public":8,"is_downloadable":8,"audit_status":8,"page_count":44,"language":140,"language_code":63,"site_id":62,"html_lang":63,"table_of_contents":141,"faqs":142,"seo_title":143,"seo_description":67,"update_tm":130,"read_time":144,"preview_page_count":4},2336464648746,"https://ap-avatar.wpscdn.com/davatar_276721f389ce27ea32af1340a28f341c","arXiv : 2608 .20346v 1 [cs .CL] 14 Jun 2026  \nBuilding and Evaluating a Synthetic Bengali Speech Resource for  \nTelecom Customer Care  \nKawshik Kumar Paul 1 Md. Nafiul Alam Fuji2  \nDepartment of Computer Science and Engineering  \nBangladesh University of Engineering and Technology (BUET)  \n[1](1 kawshikbuet17@gmail.com 2 nafiul.fuji@gmail.com)[ kawshikbuet17@gmail.com](1 kawshikbuet17@gmail.com 2 nafiul.fuji@gmail.com)[ 2](1 kawshikbuet17@gmail.com 2 nafiul.fuji@gmail.com)[ nafiul.fuji@gmail.com](1 kawshikbuet17@gmail.com 2 nafiul.fuji@gmail.com)  \nAbstract  \nSpeech systems used in customer-facing applications often require domain-specific language coverage. We present a synthetic Bengali speech dataset for telecom customer-care scenarios. The dataset contains 10,000 audio-text pairs, approximately 26.82 hours of 24 kHz speech, and predefined train, validation, and test splits of 9,000, 500, and 500 examples. It is publicly released on Hugging Face under the CC-BY-4.0 license. The speech was generated with OmniVoice in voice-cloning mode using a real female reference recording and transcript, with bfloat16 precision, 16 diffusion sampling steps, and a speaking-rate control value of 1.0 . Along with the original Bengali text, the dataset provides a normalized transcript field designed for ASR/STT training and evaluation. We report an automatic intelligibility check over all 10,000 samples using a domain-adapted Whisper ASR model fine-tuned from bengaliAI/tugstugi_bengaliai-regional-asr_whisper-medium, along with a manual listening check on selected samples. The evaluation gives an average WER of 2 .54%, an average CER of 0 .59%, and median WER and CER values of 0 .00% . These results suggest strong text-audio consistency under the selected automatic evaluation pipeline, while the paper also discusses the limitations of synthetic speech and STT-based evaluation.  \n1 Introduction  \nMany practical speech systems operate in specific domains where the vocabulary, phrasing, and user intents are repeated. Telecom and customer-care interactions are good examples. A user may ask about a failed recharge, an OTP delay, a package activation issue, a balance mismatch, SIM or account blocking, refund status, registration problems, or mobile banking verification. These situations require speech systems that can handle domain-specific expressions in a reliable way.  \nThis paper presents Bengali Telecom Customer Care Synthetic Speech Dataset, a synthetic Bengali speech resource that we released on Hugging Face (Paul and Fuji, 2026) . The dataset contains 10,000 synthetic speech samples generated from telecom and customer-care style text prompts. It includes both the original text used for speech generation and a normalized text field intended for ASR/STT training and evaluation.  \nThe dataset was generated using OmniVoice (Zhu et al. , 2026) . We use OmniVoice as a speech generation system; the contribution of this work is not a new TTS architecture. The contribution is the construction, release, documentation, and evaluation of a Bengali domainspecific synthetic speech resource.  \nThe main contributions of this paper are:  \n• We release a 10,000-sample synthetic Bengali speech dataset focused on telecom and customercare style utterances.  \n• We provide both original and normalized Bengali text fields to support TTS and ASR/STT workflows.  \n• We document the generation setup, including OmniVoice voice-cloning mode, bfloat16 precision, 16 diffusion sampling steps, and 24 kHz output audio.  \n• We evaluate all 10,000 samples using a fine-tuned Whisper-based STT intelligibility check and report WER/CER summary results.  \n• We discuss practical limitations and ethical considerations for using synthetic voice-cloned speech.  \n2 Related Work  \n2.1 Open speech resources  \nOpen speech corpora have played an important role in multilingual speech research. Common Voice is a massively multilingual speech corpus created through crowdsourced speech collectio","cbCaiteyeOwyk0tA","https://ap.wps.com/l/cbCaiteyeOwyk0tA","pdf",281629,"English","# Abstract\n# 1 Introduction\n# 2 Related Work\n## 2.1 Open speech resources\n## 2.2 Synthetic speech as a dataset resource\n## 2.3 OmniVoice\n## 2.4 Whisper-based ASR","[{\"question\":\"What dataset is proposed for telecom customer-care, and what does it contain?\",\"answer\":\"The work introduces a synthetic Bengali speech dataset for telecom customer-care scenarios. It includes 10,000 audio-text pairs, about 26.82 hours of 24 kHz audio, and predefined train/validation/test splits.\"},{\"question\":\"How is the synthetic speech generated and what text annotations are provided?\",\"answer\":\"Speech is generated with OmniVoice in voice-cloning mode using a real female reference recording and transcript. The dataset provides the original Bengali text and a normalized transcript field for ASR/STT training and evaluation.\"},{\"question\":\"How is intelligibility evaluated and what performance metrics are reported?\",\"answer\":\"An automatic intelligibility check is performed over all samples using a domain-adapted Whisper ASR model fine-tuned for Bengali telecom/customer-care. The reported average WER is 2.54% and average CER is 0.59%, with median WER/CER values reported as 0.00%.\"}]","Building and Evaluating a Synthetic Bengali Speech Resource for Telecom Customer Care | PDF",23]