This repository is dedicated to GLIMPSE Summarization Project for Deep Natural Language Processing (01VIXSM) 2024/25.
GLIMPSE is a summarization method designed to extract both common and unique viewpoints from multiple peer reviews using a Rational Speech Act (RSA) framework. It enhances transparency and decision-making in academic review processes by balancing informativeness and distinctiveness in generated summaries.
In order to run the whole code we strongly suggest that you use Google Colab with Colab Pro subscription. Otherwise you might face VRAM insufficiency error.
-
To Generate Abstractive and Extractive Candidates and Compute RSA Open GLIMPSE_ASTERISKS.ipynb and run the cells in order (Highly Resource Intensive: Use Colab)
-
To compute ROUGE Scores and extract comptued RSA Scores run all the cells in Scoring.ipynb file. (Low Resource Intensivity)
-
To plot the score graphs and generate score tables, run all the cells in Plots_Tables.ipynb file. (Low Resource Intensivity)