Judul penelitian: Validasi Informasi COVID-19 Berbahasa Indonesia Menggunakan Pendekatan Berbasis NLI dan KG
Video ini merupakan rekaman seminar hasil penelitian tesis Magister Informatika, Sekolah Teknik Elektro dan Informatika, Institut Teknologi Bandung atas nama Arief Purnama Muharram (NIM: 23521013) yang dilaksanakan pada hari Senin, 20 Mei 2024, pukul 13:00 hingga 14:00 WIB. Seminar tersebut membahas pendekatan NLI dan KG untuk memvalidasi informasi COVID-19 dalam Bahasa Indonesia.
Masifnya misinformasi di Internet telah mendorong perkembangan sistem pengecekan fakta otomatis untuk menanggulangi permasalahan tersebut. Umumnya, sistem pengecekan fakta otomatis menggunakan pendekatan NLI untuk memverifikasi kebenaran klaim berdasarkan bukti yang ada. Namun, kinerja model NLI sering kali mengalami stagnasi pada titik tertentu karena kurangnya konteks pengetahuan tertentu saat pelatihan model. Penelitian ini memperkenalkan penggunaan KG bersama dengan NLI untuk meningkatkan kinerja model dalam memvalidasi informasi COVID-19 dalam Bahasa Indonesia. Selain itu, penggunaan KG juga bertujuan untuk membuat model lebih sadar (aware) terhadap pengetahuan saat ini.
Jika terdapat pertanyaan seputar penelitian ini, silakan sampaikan melalui email ariefpurnamamuharram@gmail.com. Sesi tanya jawab dalam seminar ini tidak direkam.
Catatan: Penelitian ini mendapatkan masukan dan mengalami revisi penting setelah seminar. Hasil akhir penelitian dapat diakses di arxiv.org/abs/2409.00061.
---
Research title: Validation of COVID-19 Information in Indonesian Language Using a Natural Language Inference and Knowledge Graph-based Approach
This video is a recording of a thesis research seminar by Arief Purnama Muharram (Student ID: 23521013), a Master's student in Informatics at the School of Electrical Engineering and Informatics, Bandung Institute of Technology. The seminar was held on Monday, May 20, 2024, from 1:00 PM to 2:00 PM WIB. It discusses the application of Natural Language Inference (NLI) and Knowledge Graphs (KG) in validating COVID-19 information in Indonesian language.
The widespread misinformation on the Internet has driven the development of automated fact-checking systems to address this issue. Typically, these systems use the NLI approach to verify the truthfulness of claims based on existing evidence. However, the performance of NLI models often plateaus due to a lack of specific contextual knowledge during model training. This research introduces the use of KG in conjunction with NLI to enhance model performance in validating COVID-19 information in Indonesian. Additionally, KG aims to make the model more aware of current knowledge.
If you have any questions regarding this research, please feel free to reach out via email at ariefpurnamamuharram@gmail.com. Note that the Q&A session from the seminar was not recorded.
Note: This research received feedback and underwent significant revisions after the seminar. The final results of the research can be accessed at arxiv.org/abs/2409.00061.