SRS-Stories: Vocabulary-constrained multilingual story generation for language learning

2025-12-20

AI Summary

This study addresses the challenge of generating personalized reading materials for language learning: producing natural, grammatically correct, and semantically coherent texts that strictly incorporate target new words and spaced-repetition-scheduled review words, while respecting learners’ known vocabulary constraints. We propose the first multilingual story generation framework that deeply integrates lexical-constrained text generation with a spaced repetition system (SRS). Our approach comprises fine-tuned large language models (LLMs), three novel lexical-constraint decoding strategies, a cross-lingual (English/Chinese/Polish) generation architecture, and a cognitively grounded SRS-driven dynamic scheduling algorithm. Experiments demonstrate that our method significantly outperforms baseline constrained beam search across grammaticality, coherence, and lexical usage quality—achieving state-of-the-art performance in both human evaluation and automated metrics.

Technology Category

• Natural Language Processing: Generation
• Planning, Routing, and Scheduling: Planning with Language Models
• Machine Learning: Multimodal Learning

Application Category

• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Web Mining and Content Analysis: Mining multimedia, multimodal, multilingual, cross-lingual Web data

Abstract

In this paper, we use large language models to generate personalized stories for language learners, using only the vocabulary they know. The generated texts are specifically written to teach the user new vocabulary by simply reading stories where it appears in context, while at the same time seamlessly reviewing recently learned vocabulary. The generated stories are enjoyable to read and the vocabulary reviewing/learning is optimized by a Spaced Repetition System. The experiments are conducted in three languages: English, Chinese and Polish, evaluating three story generation methods and three strategies for enforcing lexical constraints. The results show that the generated stories are more grammatical, coherent, and provide better examples of word usage than texts generated by the standard constrained beam search approach

Problem

• Generates personalized stories for language learners using known vocabulary
• Teaches new vocabulary through context while reviewing recently learned words
• Optimizes vocabulary learning with a Spaced Repetition System in multiple languages

Innovation

• Uses large language models for personalized story generation
• Integrates Spaced Repetition System to optimize vocabulary learning
• Enforces lexical constraints across multiple languages effectively