Penilaian Pra-Verifikasi Sertifikasi Halal Melalui Ekstraksi Informasi Berbasis Named-Entity Recognition

Rannuan, Nabhyla Niagara (2026) Penilaian Pra-Verifikasi Sertifikasi Halal Melalui Ekstraksi Informasi Berbasis Named-Entity Recognition. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5025221286-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (6MB) | Request a copy

Abstract

Sertifikasi halal di Indonesia menghadapi tantangan kapasitas akibat tingginya volume pengajuan dari program Sertifikasi Halal Gratis (SEHATI), sementara verifikasi dokumen memerlukan ketelitian tinggi karena informasi produk dan bahan tersebar di berbagai bagian dalam satu dokumen. Kondisi ini mendorong kebutuhan akan sistem pendukung yang mampu mengenali informasi penting, membentuk relasi produk–bahan, dan mengidentifikasi tingkat titik kritis halal bahan secara otomatis. Penelitian ini mengembangkan sistem pra-verifikasi tiga tahap pada dokumen Sistem Jaminan Produk Halal (SJPH) sebagai dokumen pengajuan sertifikasi halal. Tahap pertama adalah ekstraksi entitas menggunakan Named Entity Recognition (NER). Tahap kedua adalah deteksi dan koreksi kesalahan anotasi relasi produk–bahan. Tahap ketiga adalah klasifikasi tingkat titik kritis halal bahan ke dalam lima kelas risiko. Model BERT–BiLSTM–CRF menghasilkan F1-score tertinggi sebesar 0,9938 pada tahap ekstraksi entitas. Koreksi anotasi produk–bahan menggunakan pendekatan knowledge graph mencapai Accuracy@1 sebesar 0,5086 untuk produk dan 0,5426 untuk bahan. Klasifikasi titik kritis halal menggunakan metode Ensemble Voting dengan Random Under-Sampling menghasilkan F1-score sebesar 0,7791. Sistem yang dikembangkan menunjukkan potensi dalam mendukung proses pra-verifikasi dokumen sertifikasi halal, meskipun perluasan data referensi dan penambahan fitur masih diperlukan untuk meningkatkan akurasi.
==================================================================================================================================
Halal certification in Indonesia faces capacity challenges due to the high volume of submissions generated by the Free Halal Certification Program (SEHATI), while document verification requires considerable accuracy because product and ingredient information is distributed across different sections of a document. This condition highlights the need for a supporting system capable of automatically identifying essential information, constructing product–ingredient relationships, and determining ingredient halal criticality levels. This study develops a three-stage pre-verification system for Halal Product Assurance System (SJPH) documents submitted for halal certification. The first stage involves entity extraction using Named Entity Recognition (NER). The second stage focuses on detecting and correcting product–ingredient relationship annotation errors. The third stage classifies ingredient halal criticality into five risk levels. The BERT–BiLSTM–CRF model achieved the highest F1-score of 0.9938 in the entity extraction stage. Product–ingredient annotation correction using a knowledge graph approach achieved Accuracy@1 scores of 0.5086 for products and 0.5426 for ingredients. Halal criticality classification using Ensemble Voting with Random Under-Sampling produced an F1-score of 0.7791. The proposed system demonstrates potential for supporting the pre-verification of halal certification documents, although expanding reference data and incorporating additional features are necessary to further improve performance.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	BERT–BiLSTM–CRF, Knowledge Graph, Named-Entity Recognition, Sertifikasi Halal, Titik Kritis Halal, BERT–BiLSTM–CRF, Halal Certification, Halal Critical Point, Knowledge Graph, Named-Entity Recognition
Subjects:	T Technology > T Technology (General) > T58.62 Decision support systems
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55201-(S1) Undergraduate Thesis
Depositing User:	Nabhyla Niagara Rannuan
Date Deposited:	25 Jun 2026 07:13
Last Modified:	25 Jun 2026 07:13
URI:	http://repository.its.ac.id/id/eprint/134073

Actions (login required)

View Item