A Baseline Analysis for Podcast

This repo is for our paper "A Baseline Analysis for Podcast Abstractive Summarization". In this paper, we build several baseline models, and also fine-tuned the current state-of-the-art (SOTA) summarization models on podcast dataset.

Detailed codes and pre-trained models will be released after the TREC 2020 competition.

Local Setup

Tested with Python 3.7 via virtual environment.

Clone the repo, go to the repo folder, setup the virtual environment, and install the required packages:

$ python3 -m venv venv
$ source venv/bin/activate
$ pip install -r requirements.txt

Example

In the example folder, you can find some sample generated podcast summaries from different models.