Bangla Writing Feedback Tutor with Spelling and Grammar Error Detection for School Students
Students in Bangladesh and West Bengal write essays and letters in Bangla for exams, but teachers with large classes cannot give detailed feedback on every piece, and mainstream writing assistants offer little or no Bangla support. Build a tutor that marks likely spelling and grammar errors in a student's Bangla text, explains each one in simple Bangla and tracks a student's common mistakes over time. Evaluate it on real student writing marked by teachers.
Intermediate
A small team, or one strong student willing to learn something new.2 semesters · 2 students
7 to ship
4 optional extrasSuggested stack
What you should ship
- A collection of 300 or more anonymised Bangla student paragraphs with teacher-marked errors, gathered with consent
- A synthetic error generator for common spelling confusions and agreement errors, to augment training data
- A spelling error detector built on a Bangla dictionary and a language model
- A token-level grammar error detector fine-tuned from a Bangla pretrained model
- Short explanations in Bangla for each error type, written with a Bangla teacher
- A web editor that highlights errors and shows a student's progress across submissions
- An evaluation of precision and recall per error type on teacher-marked text
If you have time left
- Suggested corrections, not just detection
- Support for informal Bangla written in Latin script
- A teacher dashboard summarising class-wide errors
- Comparing a small fine-tuned model with a large general model
The problem: one teacher, eighty essays
Bangla is spoken by well over 200 million people, yet tools that help students write it well lag far behind those for English. Students in Bangladesh and West Bengal write Bangla essays, letters and answers in school exams. Teachers with large classes mark errors but rarely have time to explain them or track each student's patterns. Common mistakes in spelling (such as confusing similar-sounding letters), conjunct characters and agreement repeat from year to year.
What already exists and the gap
- Bangla keyboards such as Ridmik and Google's Gboard offer word suggestions.
- Mainstream writing assistants focus on English and a few other languages, with little useful Bangla grammar support.
- Research tools for Bangla NLP exist, but few are packaged for students.
The gap is a tutor built for learners: it highlights likely errors, explains them in simple Bangla and helps a student see their own patterns.
Inspiration
BanglaBERT, released by Abhik Bhattacharjee and colleagues at Bangladesh University of Engineering and Technology (BUET) in Findings of NAACL 2022, gave Bangla a strong pretrained language model and benchmark. This project mirrors the Urdu writing and language tools already on Sefism for a different language and script, and builds on that work.
What you build
- Students paste or type their writing and see errors highlighted with a short explanation.
- Teachers see which errors are common in a class and plan lessons around them.
- Researchers get a labelled dataset of learner errors, if consent and anonymisation allow sharing.
The technically interesting part
- Data is the project. Learner error data for Bangla is scarce. Collect real writing with consent, have a teacher mark it, and design an error taxonomy.
- Synthetic errors. Generate realistic errors from clean text by applying common confusions, and test whether training on them transfers to real student errors.
- Script handling. Bangla text has Unicode normalisation issues and conjuncts. Normalise consistently or your model will learn noise.
- Precision first. False alarms discourage learners. Tune for high precision and say what the tool misses.
How to evaluate it
- Precision and recall per error type on held-out teacher-marked text.
- Synthetic versus real: accuracy when trained on synthetic errors only, real errors only, and both.
- Explanation quality: a teacher rates a sample of explanations.
- Learner study: whether students who use it for a few weeks repeat fewer errors, if time allows.
Suggested plan
- Weeks 1 to 5: consent, data collection, error taxonomy and teacher marking.
- Weeks 6 to 10: spelling detector and synthetic error generator.
- Weeks 11 to 18: grammar model fine-tuning and evaluation.
- Weeks 19 to 24: explanations, web editor and progress tracking.
- Weeks 25 to 28: final evaluation, report and demo.
Path to a product
Schools, coaching centres, publishers and edtech companies in Bangladesh and India need Bangla writing tools. A focused tutor with a labelled learner corpus is a strong base for a product or a research collaboration. The project shows end-to-end NLP skills on a low-resource language.
Scope warning
Get consent from students and parents, anonymise everything and follow your university's ethics rules. Focus on detection and explanation; full automatic rewriting can come later. Do not grade students with the tool.
Learn what you need on Sefism
- Machine learning roadmap: fine-tuning and evaluation.
- AI engineering roadmap: serving models behind an app.
- Hugging Face Transformers documentation: fine-tuning a pretrained model.
- Hugging Face Learn: token classification tutorials.
- Speech and Language Processing: background on error detection.
Get help with this idea
Without enough real learner data, the model only learns your synthetic errors. Tauseef can help you plan data collection and an error taxonomy early. If you are new here, create a free account first.
Ideas and guidance, not finished projects
These are project ideas and scoping guidance, published free for students to use as a starting point. I do not build, write, or sell final-year projects, and I do not complete coursework for anyone. Take an idea, make it yours, and build it.
Not sure this one fits you?
I help students pick an idea that matches their skills and their deadline, then scope it down until it is finishable. Guidance only, never done for you.
Stuck on something specific?
Writing only gets you so far. If you want an answer to your situation rather than the general case, book a session and we will work through it together. Sessions are free for approved Sefism members, and a few slots open each week.
Follow along
New writing, resources and project ideas land here first.