Wiguna, Edward (2025) Pembangkitan peta pikiran menggunakan metode hierarchical clustering dan transformer untuk artikel ilmiah = Mind map generation using hierarchical clustering and transformer methods for scientific articles. Bachelor thesis, Universitas Pelita Harapan.
![Title [thumbnail of Title]](http://repository.uph.edu/style/images/fileicons/text.png)
Title.pdf
Restricted to Registered users only
Available under License Creative Commons Attribution Non-commercial Share Alike.
Download (177kB)
![Abstract [thumbnail of Abstract]](http://repository.uph.edu/style/images/fileicons/text.png)
Abstract.pdf
Restricted to Registered users only
Available under License Creative Commons Attribution Non-commercial Share Alike.
Download (279kB)
![ToC [thumbnail of ToC]](http://repository.uph.edu/style/images/fileicons/text.png)
ToC.pdf
Restricted to Registered users only
Available under License Creative Commons Attribution Non-commercial Share Alike.
Download (320kB)
![Chapter1 [thumbnail of Chapter1]](http://repository.uph.edu/style/images/fileicons/text.png)
Chapter1.pdf
Restricted to Registered users only
Available under License Creative Commons Attribution Non-commercial Share Alike.
Download (363kB)
![Chapter2 [thumbnail of Chapter2]](http://repository.uph.edu/style/images/fileicons/text.png)
Chapter2.pdf
Restricted to Registered users only
Available under License Creative Commons Attribution Non-commercial Share Alike.
Download (713kB)
![Chapter3 [thumbnail of Chapter3]](http://repository.uph.edu/style/images/fileicons/text.png)
Chapter3.pdf
Restricted to Registered users only
Available under License Creative Commons Attribution Non-commercial Share Alike.
Download (1MB)
![Chapter4 [thumbnail of Chapter4]](http://repository.uph.edu/style/images/fileicons/text.png)
Chapter4.pdf
Restricted to Registered users only
Available under License Creative Commons Attribution Non-commercial Share Alike.
Download (717kB)
![Chapter5 [thumbnail of Chapter5]](http://repository.uph.edu/style/images/fileicons/text.png)
Chapter5.pdf
Restricted to Registered users only
Available under License Creative Commons Attribution Non-commercial Share Alike.
Download (321kB)
![Bibliography [thumbnail of Bibliography]](http://repository.uph.edu/style/images/fileicons/text.png)
Bibliography.pdf
Restricted to Registered users only
Available under License Creative Commons Attribution Non-commercial Share Alike.
Download (310kB)
![Appendices [thumbnail of Appendices]](http://repository.uph.edu/style/images/fileicons/text.png)
Appendices.pdf
Restricted to Repository staff only
Available under License Creative Commons Attribution Non-commercial Share Alike.
Download (1MB)
Abstract
In today's digital era, scientific literature is more accessible than ever through platforms like arXiv, ResearchGate, and ScienceDirect. However, the exponential growth of academic publications, particularly in the medical domain, presents significant challenges for readers to extract relevant information efficiently. Scientific articles are typically dense, filled with complex terminology, and structured in a way that is not always intuitive. This creates a barrier for readers trying to comprehend or locate specific information quickly. One potential solution to this problem is the use of mind maps, which visually organize information into a branched structure, allowing users to better identify, retain, and understand key ideas.
This study proposes a method for automatically generating mind maps from scientific articles using a combination of hierarchical clustering, extractive summarization, and a Transformer-based language model. The core of the system, named biomindBERT, utilizes SciBERT embeddings to represent semantic relationships between sentences. These sentence embeddings are then grouped using hierarchical clustering to capture conceptual hierarchies in the document. Representative sentences from each cluster are selected and structured into a mind map, with the optimal number of clusters determined using the (N+1)/log₂N heuristic to ensure balanced information segmentation. The final output is a structured visual summary that provides a high-level overview of the document’s contents.
Evaluation of biomindBERT involved both automatic metrics (ROUGE and BERTScore) and manual assessment by expert evaluators. The model achieved a ROUGE-1 score of 0.3620, ROUGE-2 of 0.0995, ROUGE-L of 0.1952, and BERTScore-F1 of 0.8429. In manual evaluation, the generated mind maps scored 4.7 for conciseness, 4.6 for coherence, and 4.95 for relevance (out of 5).
/
Di era digital saat ini, literatur ilmiah menjadi semakin mudah diakses melalui berbagai platform seperti arXiv, ResearchGate, dan ScienceDirect. Namun, pertumbuhan eksponensial publikasi akademik, khususnya di bidang medis, menimbulkan tantangan besar bagi pembaca dalam mengekstraksi informasi yang relevan secara efisien. Artikel ilmiah umumnya bersifat padat, dipenuhi dengan terminologi kompleks, dan disusun dengan struktur yang tidak selalu intuitif. Hal ini menjadi hambatan bagi pembaca yang ingin memahami atau menemukan informasi tertentu dengan cepat. Salah satu solusi potensial untuk permasalahan ini adalah penggunaan peta pikiran, yang mampu menyajikan informasi secara visual dalam struktur bercabang, sehingga memudahkan pengguna untuk mengenali, mengingat, dan memahami gagasan utama.
Penelitian ini mengusulkan metode pembangkitan peta pikiran secara otomatis dari artikel ilmiah dengan menggabungkan hierarchical clustering, peringkasan teks ekstraktif, dan model bahasa berbasis Transformer. Sistem inti yang dinamakan biomindBERT menggunakan embedding dari SciBERT untuk merepresentasikan hubungan semantik antar kalimat. Embedding kalimat tersebut kemudian dikelompokkan menggunakan hierarchical clustering guna menangkap hierarki konsep dalam dokumen. Kalimat representatif dari tiap klaster dipilih dan disusun ke dalam peta pikiran, dengan jumlah klaster optimal ditentukan menggunakan pendekatan heuristik (N+1)/log₂N untuk memastikan segmentasi informasi yang seimbang. Hasil akhirnya adalah ringkasan visual terstruktur yang memberikan gambaran umum tingkat tinggi dari isi dokumen.
Evaluasi terhadap biomindBERT dilakukan menggunakan metrik otomatis (ROUGE dan BERTScore) serta penilaian manual oleh evaluator ahli. Model ini memperoleh skor ROUGE-1 sebesar 0,3620, ROUGE-2 sebesar 0,0995, ROUGE-L sebesar 0,1952, dan BERTScore-F1 sebesar 0,8429. Dalam evaluasi manual, peta pikiran yang dihasilkan memperoleh skor 4,7 untuk keringkasan, 4,6 untuk koherensi, dan 4,95 untuk relevansi (dari skala 5).
Item Type: | Thesis (Bachelor) |
---|---|
Creators: | Creators NIM Email ORCID Wiguna, Edward NIM01082210005 edwardewiguna@gmail.com UNSPECIFIED |
Contributors: | Contribution Contributors NIDN/NIDK Email Thesis advisor Samosir, Feliks NIDN0319049302 feliks.parningotan@uph.edu Thesis advisor Diana, Marta NIDK6657777678230102 marta.diana@uph.edu |
Uncontrolled Keywords: | peta pikiran; transformers; hierarchical clustering |
Subjects: | Q Science > QA Mathematics > QA75 Electronic computers. Computer science |
Divisions: | University Subject > Current > Faculty/School - UPH Karawaci > School of Information Science and Technology > Informatics Current > Faculty/School - UPH Karawaci > School of Information Science and Technology > Informatics |
Depositing User: | EDWARD ENGELBERT WIGUNA |
Date Deposited: | 01 Jul 2025 13:42 |
Last Modified: | 01 Jul 2025 13:42 |
URI: | http://repository.uph.edu/id/eprint/69104 |