hotel_reviews_summarization_with_huggingface_transformers
-
Purpose of this project is to compare text summarization of T5 and Bart transformers.
-
76000+ reviews of hotels are used as a dataset for this project taken from here:
https://archive.ics.uci.edu/ml/datasets/OpinRank+Review+Dataset
How to run?
In order to run this, you need anaconda installed on your machine with pytorch and huggingface transformers installed.
Below is the link for installation instructions for Huggingface Transformers:
https://huggingface.co/transformers/installation.html
You can run this project in Jupyter Notebook as I used Google Colaboratory.
There are some code differences as my paths are from Google Drive, so you can change them as required.
- Rouge-l used as evaluation metric is being taken from this GitHub repository:
https://github.com/pcyin/PyRouge
For this awesome implementation of Rouge-l, thanks to Pengcheng YIN