• RCV1-v2 dataset is described at RCV1-v2 info
• Because it is required signing Agreement to obtain data, this repo contains no data. Personally, it only took me 1 day to be granted access to the data.
• Unfortunately, the downloaded file is neither in handy format, nor there are obvious ways to extract desirable info from it, while all I expect is a csv file contains something likes ID, text, labels. As a result, this script serves that intent.

Basically, just run:

python main.py path_to_dir

where path_to_dir is the absolute path to the directory containing file rcv1.tar.xz. It would output 2 csv files at path_to_dir:

• rcv1_v2.csv: your main interested data
• rcv1_v2_topics_desc.csv: description about topics

The content in column text are raw text in xml format. It can be parsed easily with xml.etree.ElementTree.XML(text)