Coding
IEEE Award Winning Computer Science Research Paper
-
Python
-
Algorithm Design
-
Formative/Summative Evaluation
From Spring 2022 to Spring 2023, I have served as Research Assistant to a Brigham Young University Computer Science Professor. Under his direction, I developed a multi-document hybrid summarizer. The program takes a collection of product reviews as input, parses through the collection of reviews, and then outputs a newly generated "review" that serves a summary to all of the input.
This hybrid summarizer employs an adapted KL-Divergence algorithm for selecting the most pertinent sentences, and then uses a pre-trained BART model to improve on grammar and syntax. The original concept for using KL Divergence was made by my mentor professor, but he gave me free reign in designing the code and specifics of how to apply the KL Divergence formula.
I coded the summarizer in python on Google Collab. A fellow research assistant found a database of Amazon product reviews to use, and I parsed through it to find 14 review sets to serve as testing data. I selected these review sets to have a wide variety in length to test the limits of the summarizer.
After the summarizer was completed, I researched other summarizers that could be used for comparison, selecting three for testing. All summarizers were tested on our 14 review sets, and ROUGE scoring. Our summarizer was shown to be competitive if not superior to the others. Upon completion of the summarizer, we wrote a report of our findings. I contributed the original drafts of the Introduction, Related Works section, Our Hybrid Summarizer section, Experimental Results, and Conclusion. After revisions from my mentor, we submitted it to the 34th IEEE International Conference on Tools with Artificial Intelligence in late 2022, wherein it won a "Best Student Paper Recognition Award".
Download Minesweeper: