KLASIFIKASI SENTIMEN BERBASIS ASPEK PADA ULASAN FORE COFFEE MENGGUNAKAN PENDEKATAN HYBRID LEXICON DAN FINE-TUNED INDOROBERTA

KUSUMA, AHMAD FIRZA SURYA (2026) KLASIFIKASI SENTIMEN BERBASIS ASPEK PADA ULASAN FORE COFFEE MENGGUNAKAN PENDEKATAN HYBRID LEXICON DAN FINE-TUNED INDOROBERTA. S1 thesis, Universitas Mercu Buana Jakarta.

Preview	Text (HAL COVER) 01 Cover.pdf Download (1MB) | Preview
	Text (BAB I) 02 Bab 1.pdf Restricted to Registered users only Download (190kB)
	Text (BAB II) 03 Bab 2.pdf Restricted to Registered users only Download (191kB)
	Text (BAB III) 04 Bab 3.pdf Restricted to Registered users only Download (443kB)
	Text (BAB IV) 05 Bab 4.pdf Restricted to Registered users only Download (1MB)
	Text (BAB V) 06 Bab 5.pdf Restricted to Registered users only Download (154kB)
	Text (DAFTAR PUSTAKA) 08 Daftar Pustaka.pdf Restricted to Registered users only Download (160kB)
	Text (LAMPIRAN) 09 Lampiran.pdf Restricted to Registered users only Download (1MB)

Abstract

The rapid growth of the specialty coffee industry in Indonesia has created intense competition, in which customer reviews serve as a crucial data source for evaluating service quality. However, general sentiment analysis is often insufficient to provide in-depth insights, as customers frequently express varying opinions toward different aspects within a single review. This study designs an Aspect-Based Sentiment Analysis (ABSA) model for Fore Coffee customer reviews using a Hybrid approach between Lexicon and transformer-based language models. The initial data labeling stage is performed using a Lexicon-based method through the InSet dictionary, which is then used as pseudo-labels for fine-tuning the IndoRoBERTa language model to classify sentiment. The developed model is able to identify sentiment on five specific aspects, namely Product, Service, Price, Ambiance, and Facilities, through a combination of Lexicon scores and IndoRoBERTa predictions using Hybrid weighting. Evaluation results show that the Hybrid approach is able to produce aspect-based sentiment classification that is consistent with the Indonesian language context, with a tangible contribution particularly in coverage�gap cases where the Lexicon approach fails to provide a sentiment assessment due to limited vocabulary coverage, while accuracy scores measured against the Lexicon's own labels must be interpreted with caution given the potential for evaluation circularity inherent in the weak supervision scheme used. The results of this analysis are expected to help Fore Coffee management determine improvement strategies based on more targeted customer feedback. Fore Coffee, Pseudo-labeling. Keywords: Aspect-Based Sentiment Classification, IndoRoBERTa, Lexicon-based, Perkembangan industri kopi kekinian di Indonesia menciptakan persaingan yang ketat, di mana ulasan pelanggan menjadi sumber data krusial untuk evaluasi kualitas Pelayanan. Namun, analisis sentimen secara umum seringkali belum cukup memberikan wawasan mendalam karena pelanggan sering menyampaikan opini yang berbeda-beda terhadap aspek yang berbeda dalam satu ulasan. Penelitian ini merancang model Aspect-Based Sentiment Analysis (ABSA) untuk ulasan pelanggan Fore Coffee dengan pendekatan Hybrid antara Lexicon dan model bahasa berbasis transformer. Tahap pelabelan data awal dilakukan menggunakan pseudo-label pendekatan Lexicon-based untuk proses melalui kamus InSet, yang kemudian digunakan sebagai fine-tuning model IndoRoBERTa dalam mengklasifikasikan sentimen. Model yang dikembangkan mampu mengidentifikasi sentimen pada lima aspek spesifik, yaitu Produk, Pelayanan, Harga, Suasana, dan Fasilitas, melalui kombinasi skor Lexicon dan prediksi IndoRoBERTa dengan pembobotan Hybrid. Hasil evaluasi menunjukkan bahwa pendekatan Hybrid mampu menghasilkan klasifikasi sentimen berbasis aspek yang konsisten dengan konteks Bahasa Indonesia, dengan kontribusi nyata terutama pada kasus-kasus di mana pendekatan Lexicon gagal memberikan penilaian sentimen akibat keterbatasan cakupan kosakata (coverage-gap), sementara nilai akurasi terhadap label Lexicon sendiri perlu diinterpretasikan secara hati-hati mengingat potensi evaluation circularity pada skema weak supervision yang digunakan. Hasil analisis ini diharapkan dapat membantu manajemen Fore Coffee dalam menentukan strategi perbaikan berdasarkan umpan balik pelanggan yang lebih terarah. Kata kunci: Klasifikasi Sentimen Berbasis Aspek, IndoRoBerTa, Lexicon-based, Fore Coffee, Pseudo-labeling

Item Type:	Thesis (S1)
NIM/NIDN Creators:	41522010173
Uncontrolled Keywords:	Klasifikasi Sentimen Berbasis Aspek, IndoRoBerTa, Lexicon-based, Fore Coffee, Pseudo-labeling
Subjects:	000 Computer Science, Information and General Works/Ilmu Komputer, Informasi, dan Karya Umum > 000. Computer Science, Information and General Works/Ilmu Komputer, Informasi, dan Karya Umum > 004 Data Processing, Computer Science/Pemrosesan Data, Ilmu Komputer, Teknik Informatika 100 Philosophy and Psychology/Filsafat dan Psikologi > 130 Paranormals/Paranormal > 133 Parapsychology and Occultism/Parapsikologi dan Ilmu Ghaib > 133.3 Divinatory Arts/Kepercayaan Kepada Tahayul > 133.32 Fortune-Telling by Crystals and Stones; Dowsing; Fortune-Telling by Cards, Tea Leaves and Coffee Grounds, Oracles and Sibyls/Mencari Peruntungan Lewat Kristal dan Batu; Pencelupan; Mencari Peruntungan Lewat Kartu, Daun Teh and Kopi, Oracle and Siby 200 Religion/Agama > 260 Christian Social Theology/Teologi Sosial Kristen > 268 Religious Education/Pendidikan Agama Kristen, Pengajaran Agama Kristen > 268.7 Services/Pelayanan
Divisions:	Fakultas Ilmu Komputer > Informatika
Depositing User:	khalimah
Date Deposited:	04 Sep 2026 04:23
Last Modified:	04 Sep 2026 04:23
URI:	http://repository.mercubuana.ac.id/id/eprint/103620

Actions (login required)

View Item