Kombinasi analisis sentimen dan peringkasan otomatis untuk ulasan novel berbahasa Indonesia menggunakan IndoBERT dan TextRank

Hidayatulloh, Nanda Tiara Sabina (2025) Kombinasi analisis sentimen dan peringkasan otomatis untuk ulasan novel berbahasa Indonesia menggunakan IndoBERT dan TextRank. Sarjana thesis, UIN Sunan Gunung Djati Bandung.

Preview	Text 1_cover.pdf Download (121kB) | Preview
Preview	Text 2_abstrak.pdf Download (182kB) | Preview
Preview	Text 3_skbebasplagiarism.pdf Download (380kB) | Preview
Preview	Text 4_daftarisi.pdf Download (200kB) | Preview
Preview	Text 5_bab1.pdf Download (336kB) | Preview
	Text 6_bab2.pdf Restricted to Registered users only Download (314kB) | Request a copy
	Text 7_bab3.pdf Restricted to Registered users only Download (4MB) | Request a copy
	Text 8_bab4.pdf Restricted to Registered users only Download (1MB) | Request a copy
	Text 9_bab5.pdf Restricted to Registered users only Download (182kB) | Request a copy
	Text 10_daftarpustaka.pdf Restricted to Registered users only Download (178kB) | Request a copy

Abstract

Ulasan pembaca terhadap novel yang tersedia di platform Goodreads sering kali disampaikan dengan narasi yang panjang dan gaya bahasa yang beragam sehingga tidak langsung menyampaikan inti pembahasan. Hal ini menimbulkan tantangan tersendiri dalam memahami opini terkait novel tersebut secara cepat dan efisien. Penelitian ini bertujuan untuk menerapkan model IndoBERT dalam analisis sentimen dan algoritma TextRank dalam peringkasan teks terhadap ulasan novel Laut Bercerita karya Leila S. Chudori. Dilakukan proses fine tuning terhadap model IndoBERT dengan berbagai konfigurasi hyperparameter dan menghasilkan akurasi terbaik sebesar 85.3%. Kemudian, TextRank diuji dengan tiga tingkat compression rate dan menghasilkan ringkasan ekstraktif terbaik pada compression rate 75% dengan nilai F1-score pada ROUGE-1 sebesar 0.793, pada ROUGE-2 sebesar 0.734, serta pada ROUGE-L sebesar 0.786. Analisis sentimen juga dilakukan terhadap data ulasan ringkasan, hasilnya menunjukkan tingkat kecocokan sentimen sebesar 90.24% antara ulasan asli dengan ulasan ringkasan hasil compression rate 75% dan tetap menunjukkan bahwa opini terkait ulasan novel Laut Bercerita karya Leila S. Chudori mayoritasnya adalah positif. Oleh karena itu, kombinasi diantara analisis sentimen menggunakan IndoBERT dan peringkasan teks menggunakan TextRank efektif untuk menyederhanakan dan mengidentifikasi opini dalam ulasan novel, serta tidak menghilangkan informasi opini dalam ulasan secara signifikan. Reader reviews of novels available on the Goodreads platform are often presented in long narratives and varied styles of language, which do not immediately convey the main points of discussion. This poses a challenge in understanding opinions about the novel quickly and efficiently. This study aims to apply the IndoBERT model in sentiment analysis and the TextRank algorithm in text summarization to reviews of the novel Laut Bercerita by Leila S. Chudori. The IndoBERT model was fine-tuned with various hyperparameter configurations, achieving the best accuracy of 85.3%. Then, TextRank was tested with three compression rates and produced the best extractive summary at a compression rate of 75% with an F1-score value of 0.793 on ROUGE-1, 0.734 on ROUGE-2, and 0.786 on ROUGE-L. Sentiment analysis was also conducted on the summary review data, with results showing a sentiment matching rate of 90.24% between the original reviews and the summary reviews at a compression rate of 75%, and still indicating that the opinions regarding Leila S. Chudori's novel Laut Bercerita reviews were predominantly positive. Therefore, the combination of sentiment analysis using IndoBERT and text summarization using TextRank is effective for simplifying and identifying opinions in novel reviews, without significantly losing opinion information in the reviews.

Actions (login required)

View Item