weixin_public_corpus

Why we included this project

For anyone doing Chinese NLP, this is a rare find: a few gigabytes of real WeChat public account articles, already stripped of HTML and stored as one JSON object per line with the account name, ID, title, and body. That format drops straight into pandas or a text pipeline, so it works well for language modeling, topic classification, or sentiment work on contemporary Chinese. The data comes in password-free zip volumes with a preview file, so you can check the content before downloading the whole thing. Keep in mind it is a research corpus, not a maintained tool: you will do your own cleaning, and the collection has not been updated since 2019.