Customized bidirectional encoder representations from transformer untuk brt factory online store

Winarta, Audric (2024) Customized bidirectional encoder representations from transformer untuk brt factory online store. Bachelor thesis, Universitas Pelita Harapan.

[thumbnail of Title] Text (Title)
Title.pdf
Available under License Creative Commons Attribution Non-commercial Share Alike.

Download (27kB)
[thumbnail of Abstract] Text (Abstract)
Abstract.pdf
Available under License Creative Commons Attribution Non-commercial Share Alike.

Download (82kB)
[thumbnail of ToC] Text (ToC)
ToC.pdf
Available under License Creative Commons Attribution Non-commercial Share Alike.

Download (88kB)
[thumbnail of Chapter1] Text (Chapter1)
Chapter1.pdf
Available under License Creative Commons Attribution Non-commercial Share Alike.

Download (92kB)
[thumbnail of Chapter2] Text (Chapter2)
Chapter2.pdf
Restricted to Registered users only
Available under License Creative Commons Attribution Non-commercial Share Alike.

Download (188kB)
[thumbnail of Chapter3] Text (Chapter3)
Chapter3.pdf
Restricted to Registered users only
Available under License Creative Commons Attribution Non-commercial Share Alike.

Download (255kB)
[thumbnail of Chapter4] Text (Chapter4)
Chapter4.pdf
Restricted to Registered users only
Available under License Creative Commons Attribution Non-commercial Share Alike.

Download (329kB)
[thumbnail of Chapter5] Text (Chapter5)
Chapter5.pdf
Restricted to Registered users only
Available under License Creative Commons Attribution Non-commercial Share Alike.

Download (80kB)
[thumbnail of Bibliography] Text (Bibliography)
Bibliography.pdf
Available under License Creative Commons Attribution Non-commercial Share Alike.

Download (221kB)
[thumbnail of Appendices] Text (Appendices)
Appendices.pdf
Restricted to Repository staff only
Available under License Creative Commons Attribution Non-commercial Share Alike.

Download (971kB)

Abstract

Penelitian ini bertujuan untuk mengembangkan Bidirectional Encoder Representations From Transformer (BERT) untuk mengotomasi layanan konsultasi pelanggan di BRT Factory Online Store. Model BERT dikustomisasi melalui proses fine-tuning menggunakan dataset yang didapatkan dari Tokopedia BRT Factory online Store. Hasil penelitian mengkaji keakuratan model Chatbot yang dikembangkan untuk memberikan respons terhadap pertanyaan spesifik mengenai suku cadang otomotif. Hasil pengujian menunjukkan bahwa tingkat keakuratan model secara konsisten mencapai 93,667% berdasarkan penilaian dari tiga orang dengan latar belakang pengetahuan tentang suku cadang otomotif yang berbeda. Meskipun nilai ini mencerminkan kinerja yang positif, perlu diperhatikan bahwa variasi dalam pemahaman dan pengetahuan penilai dapat mempengaruhi hasil. Selain itu, terdapat beberapa kesalahan dalam respons yang diberikan oleh Chatbot, termasuk fenomena halusinasi di mana Chatbot menghasilkan informasi yang tidak akurat atau relevan. Oleh karena itu meskipun keakuratan di atas 90% diharapkan untuk interaksi pengguna yang memuaskan, hasil ini menunjukkan perlunya perhatian terhadap konteks pertanyaan dan penyempurnaan lebih lanjut lagi. Penelitian ini memberikan wawasan bagi pengembangan Chatbot ke depannya, terutama dalam meningkatkan kualitas respons dan pemahaman konteks. / This research aims to develop a Bidirectional Encoder Representations From Transformer (BERT) to automate customer consultation services at the BRT Factory Online Store. The BERT model is customized through a fine-tuning process using a dataset obtained from the Tokopedia BRT Factory online Store. The research results examine the accuracy of the Chatbot model developed to respond to specific questions regarding automotive parts. The test results indicate that the model's accuracy level consistently reaches 93.667% based on assessments from three individuals with different backgrounds in automotive parts knowledge. Although this score reflects positive performance, it should be noted that variations in the understanding and knowledge of the assessors can affect the results. Additionally, there are several errors in the responses provided by the Chatbot, including the phenomenon of hallucination where the Chatbot generates inaccurate or irrelevantinformation. Therefore, although an accuracy above 90% is expected for satisfying user interactions, these results indicate the need for attention to the context of the questions and further refinement. This research provides insights for the future development of Chatbots, particularly in improving response quality and context understanding.
Item Type: Thesis (Bachelor)
Creators:
Creators
NIM
Email
ORCID
Winarta, Audric
NIM03082210021
audricw03@gmail.com
UNSPECIFIED
Contributors:
Contribution
Contributors
NIDN/NIDK
Email
Thesis advisor
Yugopuspito, Pujianto
NIDN0324086701
yugopuspito@uph.edu
Uncontrolled Keywords: suku cadang otomotif; Tokopedia; web scraping; Bidirectional Encoder Representations From Transformer; interaksi pengguna; Gemini API
Subjects: Q Science > QA Mathematics > QA75 Electronic computers. Computer science
Divisions: University Subject > Current > Faculty/School - UPH Medan > School of Information Science and Technology > Informatics
Current > Faculty/School - UPH Medan > School of Information Science and Technology > Informatics
Depositing User: Audric Winarta
Date Deposited: 25 Feb 2025 03:37
Last Modified: 25 Feb 2025 03:37
URI: http://repository.uph.edu/id/eprint/67144

Actions (login required)

View Item
View Item