Entity Extraction pada Chatbot Retrieval-Based untuk Analisis Data Survey Hygiene Factor pada PT PLN (Persero) Unit Pelaksanaan Assessment Center (UPAC)

Niswah, Shof Watun (2026) Entity Extraction pada Chatbot Retrieval-Based untuk Analisis Data Survey Hygiene Factor pada PT PLN (Persero) Unit Pelaksanaan Assessment Center (UPAC). Other thesis, Insttitut Teknologi Sepuluh Nopember.

Text 5026221043-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (7MB) | Request a copy

Abstract

Analisis komentar pada survei Hygiene Factor di PT PLN (Persero) masih dilakukan secara manual sehingga memerlukan waktu yang lama dan menyulitkan eksplorasi informasi dari data kualitatif. Penelitian ini mengembangkan chatbot berbasis retrieval menggunakan framework RASA untuk mendukung analisis hasil survei melalui integrasi entity extraction dan Large Language Model (LLM). Proses entity extraction menerapkan kombinasi Regular Expression (RegEx), dictionary-based lookup, synonym mapping, dan Fuzzy String Matching berbasis Levenshtein Distance untuk mengenali parameter analisis seperti unit kerja, periode survei, aspek hygiene factor, sentimen, dan ID jadwal. Entitas yang berhasil dikenali digunakan sebagai dasar pengambilan data dari basis data, kemudian hasil analisis disajikan dalam bentuk ringkasan, perbandingan, serta visualisasi distribusi dan tren yang diperkaya menggunakan Gemini 2.5 Flash. Sistem dievaluasi melalui pengujian white-box, black-box, dan human evaluation. Hasil pengujian menunjukkan bahwa konfigurasi RegEx + FuzzyEntityExtractor memperoleh performa terbaik dengan precision 100%, recall 77,61%, dan F1-score 87,39%. Pengujian black-box menghasilkan tingkat keberhasilan sebesar 82,92%, sedangkan human evaluation memperoleh nilai rata-rata di atas 4 pada skala Likert 5 untuk seluruh skenario pengujian. Hasil tersebut menunjukkan bahwa integrasi entity extraction dengan chatbot berbasis retrieval mampu mendukung eksplorasi data survei secara kontekstual, menghasilkan respons yang relevan, serta membantu mempercepat proses analisis data survei.
==============================================================================================================================
Employee Hygiene Factor survey analysis at PT PLN (Persero) is still performed manually, making the analysis process time-consuming and limiting the exploration of qualitative survey data. This study proposes a retrieval-based chatbot using the RASA framework to support survey analysis by integrating entity extraction and a Large Language Model (LLM). The entity extraction module combines Regular Expression (RegEx), dictionary-based lookup, synonym mapping, and Levenshtein Distance-based Fuzzy String Matching to identify analytical parameters, including work unit, survey period, hygiene factor aspect, sentiment, and schedule ID. The extracted entities are used to retrieve relevant survey records from the database, while the analytical responses, including summaries, comparisons, distribution charts, and trend visualizations, are refined using Gemini 2.5 Flash. The proposed system was evaluated through white-box testing, black-box testing, and human evaluation. Experimental results show that the RegEx + FuzzyEntityExtractor configuration achieved the best performance with 100% precision, 77.61% recall, and an F1-score of 87.39%. Black-box testing obtained an overall success rate of 82.92%, while human evaluation produced average scores above 4 out of 5 across all evaluation scenarios. These findings indicate that integrating entity extraction into a retrieval-based chatbot effectively supports contextual survey data exploration, delivers relevant responses, and accelerates qualitative survey analysis.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Chatbot sebagai Alat Eksplorasi Data, Framework Rasa, Regex Entity Extractor, Dictionary-Based Lookup, Chatbot for Exploratory Data Analysis, Rasa Framework, Regex Entity Extractor, Dictionary-Based Lookup
Subjects:	T Technology > T Technology (General) T Technology > T Technology (General) > T59.7 Human-machine systems.
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Information System > 57201-(S1) Undergraduate Thesis
Depositing User:	Shof Watun Niswah
Date Deposited:	29 Jul 2026 03:08
Last Modified:	29 Jul 2026 03:08
URI:	http://repository.its.ac.id/id/eprint/138988

Actions (login required)

View Item