IELTS Speaking Evaluator

IELTS Speaking Evaluator is a Dilijent Systems applied-AI engineering project that pairs examiner-style questions with scoring grounded in a reference question bank.

The challenge

Speaking practice needs an examiner: someone to ask realistic questions and then judge the answer. Judgement is the hard part. Without a reference standard, feedback drifts, and generic AI feedback cannot tell a candidate how their answer compares with what the syllabus actually rewards.

Our solution

• Specialised IBM Watson agents generate examiner-style questions and score candidate answers against reference examples
• True vector RAG: LangChain with MongoDB Vector Search retrieves the most relevant sample questions and model answers
• Scoring is grounded against retrieved reference exemplars from the question bank, not against free-form model judgement
• The same retrieval layer selects the question and supplies the standard the answer is marked against, so practice and marking stay consistent

The result