Claim Missing Document
Check
Articles

Found 2 Documents
Search
Journal : Discovery : Jurnal Ilmu Pengetahuan

Application of Graph Theory on Dominant Local Metric Dimension Based Website Using Breadth-First Search Aurelia; Reni Umilasari; Dudi Irawan
Discovery Vol 10 No 2 (2025): October 2025
Publisher : LPPM Universitas Hasyim Asy'ari Lembaga Penelitian dan Pengabdian Masyarakat

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.33752/discovery.v10i2.10073

Abstract

This study presents the implementation of the Breadth First Search (BFS) algorithm to determine the dominant local metric dimension on path graphs through a web-based platform. The system allows users to input the number of vertices, generates a corresponding path graph, computes the dominant local metric basis set, and calculates the minimum distance from each vertex to the basis using BFS. All processes are automated and displayed interactively using the Plotly library within a responsive website interface. The system was built using the Django framework, employing NetworkX for graph operations and Plotly for visualization. The BFS algorithm ensures the shortest distance calculation from a starting vertex to all others, providing a unique distance vector for each vertex that satisfies domination and local resolving set properties. This research shows that BFS is not only effective for theoretical computations but also practical in digital applications. Future development is recommended to support more complex graph structures, such as cycles, trees, or general graphs, to enhance the system's versatility.
Analisis Kinerja Transformer Untuk Named Entity Recognition (NER) Menggunakan IndoBERT Pada Transkrip Video Politik Berbahasa Indonesia Muchtar, Nabillah Ufairoh; Deni Arifianto; Reni Umilasari
Discovery Vol 10 No 2 (2025): October 2025
Publisher : LPPM Universitas Hasyim Asy'ari Lembaga Penelitian dan Pengabdian Masyarakat

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.33752/discovery.v10i2.10133

Abstract

Penelitian ini mengkaji kinerja model Named Entity Recognition (NER) berbasis IndoBERT pada transkrip pidato politik berbahasa Indonesia. Dataset terdiri dari 186 transkrip pidato resmi Presiden Republik Indonesia (periode 2014–2024) yang diperoleh dari kanal YouTube Sekretariat Kabinet. Proses meliputi pembersihan teks, normalisasi, tokenisasi WordPiece, pelabelan otomatis menggunakan model cahya/bert-base-indonesian-NER, serta fine-tuning dengan strategi 5-fold cross validation. Evaluasi menggunakan metrik precision, recall, dan F1-score. Hasil menunjukkan IndoBERT mampu mengenali entitas inti (PER, ORG, LOC) dengan performa sangat baik (nilai metrik rata‑rata > 0.98 untuk sebagian besar label), sementara entitas DATE dan CRD sedikit lebih rendah. Analisis distribusi entitas mengungkap dominasi penyebutan organisasi, lokasi, dan angka dalam pidato politik. Temuan ini memperkuat potensi IndoBERT sebagai alat bantu analisis teks politik berbahasa Indonesia