Implementasi Pendekatan Named-Entity Recognition Dalam Pembangunan Knowledge Graph Sirah Nabawiyah

Prayoga, Genta Putra (2026) Implementasi Pendekatan Named-Entity Recognition Dalam Pembangunan Knowledge Graph Sirah Nabawiyah. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5025221040-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (9MB) | Request a copy

Abstract

Sirah Nabawiyah memuat informasi mengenai tokoh, peristiwa, waktu, dan lokasi dalam sejarah Islam. Namun, penyajiannya sebagai teks naratif menyebabkan hubungan antar entitas sulit ditelusuri secara sistematis. Penelitian ini bertujuan membangun knowledge graph Sirah Nabawiyah berbahasa Indonesia menggunakan Named-Entity Recognition (NER), Neo4j, dan Social Network Analysis (SNA) untuk penelusuran dan analisis hubungan antarentitas.
Penelitian dilakukan melalui ekstraksi teks menggunakan OCR, prapemrosesan, pemecahan teks menjadi chunk, pelabelan semi-otomatis, serta pelatihan NER berbasis IndoBERT dengan strategi iterative self-training. Orientasi peran semantik digunakan untuk menentukan entitas Person, Event, Location, dan Time. Entitas hasil ekstraksi dinormalisasi melalui normalisasi alias, dihubungkan berdasarkan pola relasi, diperkaya dengan periodisasi, dan disimpan dalam Neo4j. Evaluasi mencakup tiga skenario NER, SNA, dan enam kueri fungsional.
Hasil terbaik dicapai IndoBERT uncased dengan augmentasi data dan F1-score mikro sebesar 0,9756. Knowledge graph yang dibangun memuat 1.192 simpul dan 728 relasi. Proyeksi jaringan antar tokoh terdiri atas 137 simpul dan 1.853 sisi serta menghasilkan delapan komunitas dengan modularitas Louvain sebesar 0,2831. Muhammad menempati posisi tertinggi pada seluruh ukuran sentralitas. Keenam kueri fungsional berhasil dijalankan, menghasilkan jawaban tidak kosong, dan dapat ditelusuri ke sumber, meskipun sebagian relasi masih memerlukan verifikasi terhadap teks sumber.
=====================================================================================================================================
Sirah Nabawiyah contains information about people, events, times, and locations in Islamic history. However, its narrative form makes relationships among entities difficult to trace systematically. This study aims to construct an Indonesian Sirah Nabawiyah knowledge graph using Named-Entity Recognition (NER), Neo4j, and Social Network Analysis (SNA) for relational exploration and analysis.
This research involved OCR-based text extraction, preprocessing, segmentation into chunks, semi-automatic labeling, and IndoBERT-based NER training through iterative self- training. A semantic-role orientation was used to define Person, Event, Location, and Time entities. Extracted entities were normalized through alias normalization, connected using relation patterns, enriched with periodization, and stored in Neo4j. Evaluation comprised three NER scenarios, SNA, and six functional queries.
The best result was achieved by uncased IndoBERT with data augmentation, obtaining a micro F1-score of 0.9756. The knowledge graph contained 1,192 nodes and 728 relationships. The projected person network comprised 137 nodes and 1,853 edges, forming eight communities with a Louvain modularity of 0.2831. Muhammad ranked highest across all centrality measures. All six functional queries ran successfully, returned nonempty results, and were traceable to the source, although some relationships still require verification against the source text.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	IndoBERT, Knowledge Graph, Named-Entity Recognition, Sirah Nabawiyah, Social Network Analysis
Subjects:	T Technology > T Technology (General) > T57.5 Data Processing
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55201-(S1) Undergraduate Thesis
Depositing User:	Genta Putra Prayoga
Date Deposited:	25 Jul 2026 08:48
Last Modified:	25 Jul 2026 08:48
URI:	http://repository.its.ac.id/id/eprint/137567

Actions (login required)

View Item