DIY-MKG: An LLM-Based Polyglot Language Learning System

Sep 29, 2026

Upload by James

Existing language learning tools, even those powered by Large Language Models (LLMs), often fail polyglot learners by not building connections across vocabularies in multiple languages, offering limited personalization of learning pace, and causing harmful cognitive offloading. This work presents Do-It-Yourself Multilingual Knowledge Graph (DIY-MKG), an open-source system for personalized vocabulary knowledge graphs expanded with LLM-suggested related words. It adds rich annotation, adaptive LLM-driven quiz generation, and a feedback loop through incorrect-question flags, with evaluations showing reliable multilingual expansion and highly accurate quizzes.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages11

Reading Time28 min

AccessFree to download

Abstract

1 Introduction

2 Related Work

DIY-MKG: An LLM-Based Polyglot Language Learning System Kenan Tang UCSB [kenantang@ucsb.edu](kenantang@ucsb.edu) Yanhong Li University of Chicago [yanhongli@uchicago.edu](yanhongli@uchicago.edu) Yao Qin UCSB [yaoqin@ucsb.edu](yaoqin@ucsb.edu) arXiv :2507 .0 1872v 1 [ cs .CL] 2 Jul 2025 Abstract Existing language learning tools, even those powered by Large Language Models (LLMs), often lack support for polyglot learners to build linguistic connections across vocabularies in multiple languages, provide limited customization for individual learning paces or needs, and suffer from detrimental cognitive offloading. To address these limitations, we design Do-It-Yourself Multilingual Knowledge Graph (DIY-MKG), an open-source system that supports polyglot language learning. DIY-MKG allows the user to build personalized vocabulary knowledge graphs, which are constructed by selective expansion with related words suggested by an LLM. The system further enhances learning through rich annotation capabilities and an adaptive review module that leverages LLMs for dynamic, personalized quiz generation. In addition, DIY-MKG allows users to flag incorrect quiz questions, simultaneously increasing user engagement and providing a feedback loop for prompt refinement. Our evaluation of LLM-based components in DIY-MKG shows that vocabulary expansion is reliable and fair across multiple languages, and that the generated quizzes are highly accurate, validating the robustness of DIY-MKG. 1 1 Introduction Large Language Models (LLMs) have demonstrated superior multilingual capabilities (Gemini Team, 2025 ; Hurst et al., 2024) . Therefore, LLMs have been used extensively in education applications, assisting language learners in improving capabilities such as reading, writing, and translation (Han et al., 2023a,b ; Chu et al., 2025) .2 However, less attention has been paid to using LLMs to assist vocabulary acquisition, a fundamental task for language learning (Barcroft, 2004) . 1 [https://github.com/kenantang/DIY-MKG](https://github.com/kenantang/DIY-MKG) 2In this paper,“language learners” refers specifically to learners who are learning a foreign language (L2, L3, etc.) . Figure 1: DIY-MKG allows the language learner to construct a multilingual knowledge graph to help with vocabulary acquisition. Linguistic connections between words, such as synonyms and cognates, enhance memorization of vocabulary knowledge. This is a screenshot taken from the DIY-MKG interface. Currently available commercial software, such as Duolingo, usually includes a full-fledged system to help vocabulary acquisition. Treating the user’s current vocabulary as a word list, a commercial software tracks the learning progress, reminds the user to review words, and provides interactive lessons and quizzes to assist vocabulary acquisition. Despite the maturity of commercial software, LLMs open up exciting new possibilities for vocabulary acquisition, which have not yet been incorporated into popular commercial software. With the strong capabilities of LLMs in mind, we identify three limitations of existing systems. First, existing systems do not exploit linguistic connections across words in multiple languages (examples in Figure 1) . Secondly, existing systems are not customized to the pace of diverse language learners (Section 3.3), as the systems heavily rely on predefined lessons and quizzes. Thirdly, when LLMs are involved in language learning, many systems lack a wellstructured design that balances the contributions of both the language learner and the LLM to the learning process. When the learner blindly accepts the material generated by an LLM, the learning process is hampered by excessive cognitive offloading (Kosmyna et al., 2025), where the learner fails to engage in independent problem-solving or critical thinking while interacting with the LLM. To address these three limitations, we propose Do-It-Yourself Multilingual Knowledge Gr

FAQs

What limitations does DIY-MKG address in existing LLM-based language learning tools?

DIY-MKG targets missing cross-language vocabulary connections, insufficient customization of learning pace, and lack of designs that balance learner agency with the LLM to reduce cognitive offloading.

How does DIY-MKG build personalized vocabulary knowledge graphs?

DIY-MKG lets users create and expand a multilingual vocabulary knowledge graph, using selective expansion with related words suggested by an LLM.

How does DIY-MKG improve practice and learning assessment?

DIY-MKG provides rich annotation and an adaptive review module that uses LLMs to generate dynamic, personalized quizzes, plus an option to flag incorrect quiz questions for continuous feedback and prompt refinement.

Sep 29, 2026

Upload by James