About the dataset
#2
by
sushant-m-nair
- opened
Hi,
I find this dataset quite good. I can have a one-stop place to get lots of links of AI Papers. However, I have a suggestion - the content column data needs to be cleaned. A lot of invalid characters (due to the presence of tables and images) are present making the content require preprocessing. It would be nice for us to be able to skip preprocessing.
Anyways, a great initiative of which I am truly appreciative.
Thanks.
This repo is great for collecting URLs. I have an idea that we can use these URLs, and extract text from PDF in a clean way using the code in this repo:
https://github.com/sushantnair/ArXiv-PDF-to-Text/tree/main
Please check it out.
sushant-m-nair
changed discussion status to
closed