Data Modeling
Mathematical and Computational Data Modeling
0
Citations
4.5k
Views
29
Articles
Volume 6 Number 1 2025
RESEARCH ARTICLE (Open Access)
Virtual Personal Assistant (VPA) for Bengali language with BanglaSTT (Whisper) with RL, BanglaBERT (BERT) and BanglaTTS (SileroTTS)
Sadidul Islam 1, Khandokar A. Mamun 2
Data Modeling 6 (1) 1-8 https://doi.org/10.25163/data.6110803
Submitted: 29 September 2025 Revised: 20 November 2025 Accepted: 25 November 2025 Published: 27 November 2025
Abstract
Voice assistants have, by now, become almost invisible parts of everyday life — quietly setting alarms, answering questions, managing small tasks — yet this convenience has remained, for the most part, unevenly distributed. English-speaking users have benefited considerably; speakers of Bangla, one of the world's most widely spoken languages, largely have not. This review examines the development of a Bengali Virtual Personal Assistant (VPA) that brings together three components — speech recognition, conversational understanding, and speech synthesis — into a single, cohesive pipeline. Specifically, it considers how a Whisper-based Bengali speech-to-text model, a BanglaBERT-driven chatbot, and a SileroTTS-based speech synthesizer can be combined, and, perhaps more interestingly, how Reinforcement Learning might allow such a system to gradually adapt to an individual speaker's voice. That last point matters more than it might first seem: in a country where illiteracy remains common and a meaningful share of the population lives with some form of disability, including speech-related limitations, a one-size-fits-all model risks excluding exactly the people who could benefit most. Drawing on demographic and accessibility data alongside a working prototype, this review traces both what current Bengali NLP tools can already do and where the more difficult, unresolved challenges — personalization, dynamic knowledge, real-world deployment — still lie ahead.
Keywords: —Virtual Personal Assistant, Bengali NLP, Speech to Text, Whisper, BanglaBERT, Text to Speech, Reinforcement Learning, Chatbot
References
Islam, S. M., Houya, M. F. A., Islam, S. M., Islam, S., & Hossain, N. (2019). Adheetee: A comprehensive Bangla virtual assistant.
Badhon, S. M. S. I., Rahaman, M. H., & Rupon, F. R. (2020). State of art research in Bengali speech recognition.
Gupta, D., Hossain, E., Hossain, M. S., Andersson, K., & Hossain, S. (2019). A digital personal assistant using Bangla voice command recognition and face detection.
Bellegarda, J. R. (2013). Large-scale personal assistant technology deployment: The Siri experience.
Ricciardelli, E., & Biswas, D. (2019). Self-improving chatbots based on reinforcement learning.
Hussain, S., Ameri Sianaki, O., & Ababneh, N. (2019). A survey on conversational agents/chatbots classification and design techniques.
Moon, D. (2012). Web-based text-to-speech technologies in foreign language learning: Opportunities and challenges.
Kowsher, M., Tithi, F. S., Alam, M. A., Huda, M. N., Moheuddin, M. M., & Rosul, M. G. (2019). Doly: Bengali chatbot for Bengali education.
Apple. (2011). Siri. http://www.apple.com/ios/siri/
Google. (2008). Google mobile. http://www.google.com/mobile/voice-actions
Google. (2012). Google Now. http://www.google.com/landing/now
Microsoft. (2008). Tellme. http://www.microsoft.com/en-us/Tellme/consumers/default.aspx
Nuance. (2011). Dragon Go!. http://www.nuance.com/products/dragon-go-in-action/index.htm
Nuance. (2012). Nina. http://www.nuance.com/for-business/by-solution/customer-service-solutions/solutions-services/mobile-customer-service/nina/index.htm
Speaktoit. (2012). Speaktoit Assistant. http://www.speaktoit.com/index.htm
Emery, C. (2023). The 33 most spoken languages in the world. Langoly. Retrieved July 14, 2023
Mahtab, M., & Selim, F. (2023). UNICEF concerned that more than half of children with disabilities in Bangladesh do not go to school. UNICEF Bangladesh. Retrieved July 14, 2023
Jasim, M. M. (2023). Bangladesh's slow march towards 100% literacy. The Business Standard. Retrieved July 14, 2023
Government of Bangladesh. (2022, December). Report on National Survey on Persons with Disabilities (NSPD) 2021. ReliefWeb.
Pinon, R., & Haydon, J. (2010, December). The benefits of the English language for individuals and societies: Quantitative indicators from Cameroon, Nigeria, Rwanda, Bangladesh and Pakistan. Euromonitor International Ltd.
Hossain, S. (2024). BanglaTTS [Source code]. GitHub. https://github.com/shhossain/BanglaTTS
Hossain, S. (2025). BanglaSpeech2Text [Source code]. GitHub. https://github.com/shhossain/BanglaSpeech2Text
CSEBUETNLP. (2025). BanglaBERT [Source code]. GitHub. https://github.com/csebuetnlp/banglabert
Recommended articles
An Integrated AI-Driven Framework for Maternal Resource Intelligence Shortages Across U.S. Hospital
From Scarcity to Fairness: A Closed-Loop Reinforcement Learning and Geospatial Analytics Framework for Equitable Healthcare Supply Chain Preparedness
Integration of Artificial Intelligence in Bank Customer Relationship Management in Bangladesh
Article metrics
View details
0
Downloads
0
Citations
33
Views
0
Save
Save
0
Citation
Citation
33
View
View
0
Share
Share