Reproduce “MediaSum: A Large-scale Media Interview Dataset for Dialogue Summarization” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Chenguang Zhu, Yang Liu, Jie Mei, Michael Zeng
MediaSum pairs 463,596 public NPR and CNN interview transcripts with source-provided descriptions. Summary-matching words cluster near the beginning in both sources, while NPR also has many near the end. In a UniLM transfer experiment, ROUGE-L F1 on ICSI rose from 17.72 to 18.59 after MediaSum fine-tuning; this finding is limited to the tested model and protocol.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.