MARKDOWN ORDERING ON JUPYTER NOTEBOOK USING PAIRWISE METHOD ON AI4CODE DATASET

• Main Author: Naufal Sudrajat, Zaidan
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/78314
• Institution: Institut Teknologi Bandung

The development of Machine Learning research has resulted in its use in various fields. Microsoft Fx and GitHub Copilot, which utilize Machine Learning to assist programmers, have initiated much research on using machine learning in the software engineering pipeline. Google launched a competition to build a model to order Jupyter Notebook with jumbled Markdown Cells to find the relationship between markdown (comments) and programming languages. The task to order Jupyter Notebook with jumbled Markdown Cells will be called Markdown Ordering in this final project paper. This final project aims to build a Markdown Ordering model using the pairwise method to order Markdown Cells on the Jumbled Jupyter Notebook. This final project uses the AI4CODE dataset, which contains a collection of Jupyter Notebook Kaggle containing cell markdowns in English and code cells in Python. This final project focuses on experiments to build machine learning models using the pairwise method, referred to in research (Manku & Paul, 2022), and determines each model's performance. The performance of each model built will be compared to find the best performance model. Based on the experimental results, using the Pairwise-BERT-Softmax model can increase the performance of the baseline model by 10%. This model is the best model built in this final project.