LitSumm
Introduction
LitSumm is a project that builds upon the existing ExtSumm model proposed in 2019 by Xiao et al. The task is to perform extractive summarization on academic literature (papers, articles).
My research team has re-implemented this model, and made several corrections. We used a visualization technique to provide examples of model activation on sample articles. We also defined a new metric to evaluate academic literature that takes sentence ordering into accounts.
Results
| ROUGE-1 | ROUGE-2 | ROUGE-L |
|---|---|---|
| 46.26 | 17.95 | 30.63 |