It's done automatically by looking for category names (and synonyms) in abstracts. This mostly works although it's not 100%. However, if you click on the paper, you can add/remove them in the "Tasks" section.
Sharing a project I’ve been working on in spare time: Papers with Code.
Papers with Code links the latest machine learning papers on ArXiv with code on GitHub. It allows you to see “trending research” allowing you to keep up-to-date with what’s popular in the ML community, as well as the most popular historical papers. Additionally it shows the framework the implementation was coded in: TensorFlow/PyTorch/MXNet/etc. Lastly, you can search for papers and choose a code implementation of your choice.
Links are automatically scraped from arxiv papers and github repos, so in some cases e.g. the SNIPER repo there are three entries because this library implements three papers.
Let me know your thoughts, and if there’s any other feature ideas you have for the site,
Just checking the references in this article. The offshore tax claim seems to be referenced to "Guido Fawkes" (https://order-order.com/ "Guido Fawkes Blog of plots, rumours and conspiracy") which seems to be a collection of conspiracy articles about all kinds of stuffs, written by this guy: https://en.wikipedia.org/wiki/Paul_Staines
A shocking amount of disinformation (“fake news”) is also created by and spread from smaller, fringe Web communities that have relatively outsized influence on the greater Web.
We set out to measure just how this influence flows in a systematic and methodological manner, analyzing how URLs from 45 mainstream and 54 alternative news sources are shared across 8 months of Reddit, 4chan, and Twitter posts. Highlights:
1. Reddit and 4chan post mainstream news URLs at over twice the rate than Twitter does
2. Alternative news URLs spread much faster than mainstream URLs, perhaps an artifact of automated bots
3. 4chan was also the most successful at “reviving” old stories
We found that Twitter does have heavy influence on the spread of fake news. The_Donald and /pol/ are responsible for around 6% of mainstream news URLs over 4.5% of alternative news URLs on Twitter.
Hopefully we'll learn more about the underlying genetics and have better risk assessment tests in the future to catch these earlier.