Toward Smarter Code Review: Evaluating LLM-Based Importance Filtering for Pull Request Comments

This is the HW of 722, we fine tuned a bert base model to predict if a comment in GitHub PR is useful. 1500data.xlsx is the file that we three annotated 1500 comments sampled from 495 projects, over 4 million PRs from Github Platform. finetuned_BERT_comments_detector.ipynv is the code used to finetune bert model. fintuned_BERT_epoch_5.ipynb is the fifth model trained by our code which is also the model performed better. senEmailNoti.ipynb is the file used to predict a comment usefulness and send email to users with a email title to notify stakeholders if the comment in a PR is important or not. Interview Plan.pdf is the file we used to interview users.