SISTEM TUTOR PEMBELAJARAN BAHASA JEPANG DENGAN GRAPHRAG BERBASIS NEURO-SYMBOLIC AI

ARISANDY, KRISNA SATYA (2026) SISTEM TUTOR PEMBELAJARAN BAHASA JEPANG DENGAN GRAPHRAG BERBASIS NEURO-SYMBOLIC AI. S1 thesis, Universitas PGRI Madiun.

	Text Halaman Depan.pdf Download (463kB)
	Text Abstrak.pdf Download (123kB)
	Text BAB I.pdf Download (149kB)
	Text BAB II.pdf Download (3MB)
	Text BAB III.pdf Restricted to Repository staff only Download (465kB) | Request a copy
	Text BAB IV.pdf Restricted to Repository staff only Download (3MB) | Request a copy
	Text BAB V.pdf Download (136kB)
	Text Daftar Pustaka.pdf Download (143kB)
	Text Lampiran.pdf Restricted to Repository staff only Download (3MB) | Request a copy

Abstract

Independent learning of elementary Japanese (JLPT N5) is frequently hindered by structural complexities and high prerequisite dependencies among Kanji, vocabulary, and grammar. Utilizing pure Large Language Models (LLMs) as virtual tutors also risks semantic hallucinations due to the absence of explicit curriculum control. This research aims to design and implement an adaptive Japanese virtual tutoring system using a Neuro-Symbolic AI architecture. This hybrid approach integrates the generative flexibility of a local LLM via Llama.cpp with the data validity of a Neo4j Knowledge Graph through Graph Retrieval-Augmented Generation (GraphRAG). The instructional path sequencing is managed dynamically using a topological sort algorithm, supported by Bayesian Knowledge Tracing (BKT) for real-time learner competency tracking, and the SM-2 algorithm for spaced repetition memory management. The backend is developed using FastAPI, while the frontend utilizes SvelteKit, enhanced by a multimodal 3D VRM avatar with integrated Speech-to-Text (Faster Whisper) and Text-to-Speech (Style-BERT-VITS2) features. Layered testing evaluating structural, behavioral, and robustness pillars demonstrates that all primary functionalities operate successfully with precise graph reasoning and absolute mitigation of semantic hallucinations. This study provides a significant scientific contribution to the EdTech domain by anchoring the stochastic nature of LLMs with symbolic graph representations, thereby delivering a secure, academically structured independent learning environment for novice students.

Item Type:	Thesis/Skripsi/Tugas Akhir (S1)
Kata Kunci:	Neuro-Symbolic AI, GraphRAG, Pembelajaran Bahasa Jepang. Neuro-Symbolic AI, GraphRAG, Japanese Language Learning.
Subjects:	L Education > L Education (General) T Technology > T Technology (General)
Divisions:	Fakultas Teknik > Teknik Informatika
Depositing User:	ARISANDY SATYA KRISNA
Date Deposited:	07 Aug 2026 06:51
Last Modified:	07 Aug 2026 06:51
URI:	http://eprint.unipma.ac.id/id/eprint/8490

Available Versions of this Item

• SISTEM TUTOR PEMBELAJARAN BAHASA JEPANG DENGAN GRAPHRAG BERBASIS NEURO-SYMBOLIC AI. (deposited 07 Aug 2026 06:51) [Currently Displayed]

Actions (login required)

View Item