This repository contains several rudimentary notebooks exploring the Text Summarization task - from the extractive, as well as the abstractive point of view.
Download the wikihowAll.csv from https://github.com/mahnazkoupaee/WikiHow-Dataset
This directory contains a news article from The Guardian for manual evaluation
This directory contains notebooks for different scenarios and algorithms used
Every notebook should be good on its own, meaning that apart from changing the directory paths, every notebook is self-containable and can be run from scratch
-
huggingface_wikihow_summarization - makes use of the summarization pipeline of huggingface, which is then used from scratch
-
LSTM-with-attention-wikihow.ipynb - uses a LSTM Encoder/Decoder Model with Decoder Attention
to train on short articles and summaries -
extractive-summarization-simple.ipynb - implement a simple word frequency based extractive method for text summarization and evaluates
-
extractive-summarization-page-rank.ipynb - implements an extractive method based on the page rank algorithm
All the references and results are to be listed in the project report