Answers are grounded in this paper and cite their sources.
English learners often lack two things at once: regular conversation with real partners and grammar practice that matches their individual weaknesses. This paper reviews recent LLMbased and adaptive English-learning systems and proposes an integrated web platform that combines friend-gated language-partner discovery, real-time chat and audio/video calls, and an LLM-driven adaptive grammar module. The platform uses React.js, Node.js with Express.js, MongoDB and the Stream SDK. The adaptive module analyses learner text, categorises grammatical errors, maintains an error-weighted learner profile through an exponential-movingaverage update, and selects the next exercise by maximising a utility function that balances weakness relevance, difficulty fit and repetition. A reciprocal compatibility score ranks candidate language partners. The paper formalises these components mathematically, presents architecture, workflow, sequence and data-model diagrams, and defines an evaluation protocol that compares adaptive practice with a fixed-sequence baseline. No experimental results are claimed: the numerical illustration uses synthetic inputs only, and implementation and empirical testing are required to establish whether the approach improves learning outcomes.
No takes yet. Share an insight, caveat, or question.
Sonone et al. (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: