Structure-aware translation
How can grammatical and verb-centered representations improve English-to-Bangla translation while handling idioms, modifiers, and varied sentence forms?
01 / Research focus
Computing across Bangla text, speech, and images.
Explore Masud Rabbani’s research on English-to-Bangla machine translation, Bangla speech and sentence recognition, handwritten text, and low-resource language technology.
6 related publications · 2 projectsResearch overview
Bangla NLP, Machine Translation, and Speech Recognition
My Bangla language-technology research explores practical methods for machine translation, sentence classification, speech recognition, handwriting recognition, and text extraction from images. The program addresses a shared constraint: Bangla language systems often have fewer standardized datasets and tools than high-resource languages, while real use includes regional dialects, varied sentence structures, handwritten text, and documents captured as images. The work therefore combines linguistic rules, statistical analysis, supervised machine learning, pattern recognition, and application development.
Two English-to-Bangla translation systems use the main or principal verb to normalize varied English sentences into a simpler subject–verb–object structure. The later PVBMT study combines grammatical rules with corpus handling for prepositional phrases and idioms and evaluates syntactic and semantic quality against existing online systems. For text analytics, a supervised-learning study classifies interrogative and exclamatory Bangla sentences drawn from stories, blogs, and conversations. A systematic review maps methods and open problems in Bangla speech recognition, with particular attention to regional dialects, dataset limitations, and computational requirements.
Related work expands the input modalities. A structural stroke-matching method analyzes handwritten Bangla characters and words, while an OCR application extracts English text from images and translates it into Bangla or Hindi. Together, these publications connect core language processing with usable translation and recognition tools. The research highlights why evaluation should include dialect, domain, writing style, input quality, and human interpretation—not only a single aggregate accuracy score.
Questions
The problems guiding this research.
Low-resource robustness
Which models and datasets can represent regional speech, informal text, and domain variation without hiding poor performance behind an overall average?
Language across modalities
How can text, speech, handwriting, and camera-captured documents share practical recognition and translation workflows for Bangla users?
Research method
From a measurable signal to a defensible result.
- 01
Model linguistic structure
Represent verbs, syntax, semantics, sentence classes, and writing patterns in forms that reflect Bangla language behavior and the intended task.
- 02
Build representative data
Use text, speech, handwriting, and image sources that expose dialect, domain, script, and capture-quality variation rather than one narrow sample.
- 03
Compare approaches
Evaluate rule-based, statistical, supervised-learning, OCR, and pattern-recognition methods with task-appropriate baselines and error analysis.
- 04
Test usable output
Assess whether translations and recognitions preserve meaning for readers, not only whether an automated score or classifier label appears favorable.
P01
Onubadok
Verb Based Approach for English to Bangla Machine Translation- A desktop and web-based application for translation any English sentence to Bangla sentence.
↗P02English to Bangla Translator
A desktop application for translation simple English Sentence to Bangla Sentence using tense structure.
↗Journal article · 2016
PVBMT: A Principal Verb based Approach for English to Bangla Machine Translation
Masud Rabbani, Kazi Md. Rokibul Alam, Muzahidul Islam, Yasuhiko Morimoto. “PVBMT: A Principal Verb based Approach for English to Bangla Machine Translation”. International Journal of Computer Vision and Signal Processing, 6(1), 1–9. (2016).
Conference paper · 2014
A New Verb based Approach for English to Bangla Machine Translation
Masud Rabbani, Kazi Md. Rokibul Alam, Muzahidul Islam. “A New Verb based Approach for English to Bangla Machine Translation”. 2014 International Conference on Informatics, Electronics & Vision (ICIEV), 1–6. (2014). https://doi.org/10.1109/ICIEV.2014.6850684.
Conference paper · 2024
Natural Language Processing for recognizing Bangla speech with regular and regional dialects: A survey of algorithms and approaches
Paramita Basak Upama, Parama Sridevi, Masud Rabbani, Kazi Shafiul Alam, Munirul Haque, Sheikh Iqbal Ahamed. “Natural Language Processing for recognizing Bangla speech with regular and regional dialects: A survey of algorithms and approaches”. 2024 IEEE 48th Annual Computers, Software, and Applications Conference (COMPSAC), 312–319. (2024). https://doi.org/10.1109/COMPSAC61105.2024.00051.
Conference paper · 2021
An Automated Approach for Bangla Sentence Classification Using Supervised Algorithms
Md. Hasan Imam Bijoy, Mehedi Hasan, Abdur Nur Tusher, Md. Mahbubur Rahman, Md. Jueal Mia, Masud Rabbani. “An Automated Approach for Bangla Sentence Classification Using Supervised Algorithms”. 2021 12th International Conference on Computing Communication and Networking Technologies (ICCCNT), 1–6. (2021). https://doi.org/10.1109/ICCCNT51525.2021.9579940.
Conference paper · 2015
A New Stroke Matching based Approach to Recognize Bangla Handwritten Text
Masud Rabbani, Kazi Md. Rokibul Alam, Muzahidul Islam, Yasuhiko Morimoto. “A New Stroke Matching based Approach to Recognize Bangla Handwritten Text”. 2015 18th International Conference on Computer and Information Technology (ICCIT), 501–506. (2015). https://doi.org/10.1109/ICCITechn.2015.7488122.
Conference paper · 2021
SOT: An Application-based Research for Translate Natural Language from Image
Masud Rabbani, Ali Md Musfick Jamil, Thaharim Khan, Md Ishrak Islam Zarif. “SOT: An Application-based Research for Translate Natural Language from Image”. Advances in Electrical and Computer Technologies, 711, 171–178. (2021). https://doi.org/10.1007/978-981-15-9019-1_15.
Responsible translation
Promising research still needs careful validation.
Language technology becomes useful when it supports the forms of Bangla that people actually read, write, and speak. This requires attention to regional dialects, informal language, handwritten variation, limited training data, and interfaces that let users review uncertain output.
Future work benefits from larger shared datasets, transparent benchmarks, native-speaker evaluation, dialect-aware testing, and direct comparison with modern multilingual language and speech models while preserving privacy and cultural context.
Continue exploring