Claim Missing Document
Check
Articles

Found 1 Documents
Search
Journal : Journal Software, Hardware and Information Technology

Penerapan Natural Language Processing (NLP) dengan Metode Cosine Similarity pada Sistem E-Monev untuk Pencarian Program Pembangunan Daerah Ansyori, Moch Faizal; Ahmad Heru Mujianto
Journal Software, Hardware and Information Technology Vol 5 No 2 (2025)
Publisher : Jurusan Sistem Informasi Universitas Islam Negeri Alauddin Makassar

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.24252/shift.v5i2.183

Abstract

The Evaluation and Monitoring System (E-Monev) is a digital tool utilized for the periodic oversight of performance achievements and budget absorption within Regional Government Organizations (OPD). A primary challenge in its implementation lies in accurately identifying relevant regional development programs based on user text input. Conventional keyword-based search approaches are limited in their ability to comprehensively understand the semantic meaning of text, frequently yielding inaccurate or contextually irrelevant results. This study aims to design and develop a semantic search feature for the E-Monev system. It applies a Natural Language Processing (NLP) approach, employing the Cosine Similarity method and text representation based on BERT (Bidirectional Encoder Representations from Transformers). The data for this research was sourced from the Work Plan (Renja) documents of all OPDs in Bojonegoro Regency. These Renja documents were prepared in accordance with the Decree of the Minister of Home Affairs Number 900.1.15.5-3406 of 2024, which constitutes the Second Amendment to Decree of the Minister of Home Affairs Number 050-5889 of 2021, regulating the verification, validation, and inventory of updates to the classification, codification, and nomenclature of regional development and financial planning. During the analysis, both user input texts and Renja data were converted into embedding vectors, from which their semantic similarity was calculated. Test results demonstrate that the developed system significantly improves search accuracy, achieving a precision value of 0.884.