An Empirical Study On Duplicate Bug Report Identification Using Siamese Cross-Encoder Network

This project replicates "Towards Accurate Duplicate Bug Retrieval Using Deep Learning Techniques" by Jayati Deshmukh, K. M. Annervaz, Sanjay Podder, Shubhashis Sengupta, and Neville Dubash in International Conference on Software Maintenance (ICSM) 2017. We show that even without handling structured information separately, we can achieve comparable performance with respect to the original work.

Dataset

Download and store the dataset into MongoDB from here. If you are using Docker for MongoDB, you can find the docker-compose.yaml file in the root directory.

Installing Packages

We highly encourage to use a virtual environment to run the project. You can find the list of necessary packages in the requirements.txt file in the root directory. Install them by running

pip install -r requirements.txt

Run

Start a jupyter server by running

jupyter notebook

Then open notebooks/siamese-trials/title-descr-eclipse.ipynb in the jupyter app. To see result for different datasets, change the third line in the second cell accordingly.

Result

Check project-report.pdf for detailed analysis of the evaluation and findings.