Chinese-Word-Vectors

Why we included this project

If you work on Chinese NLP, this collection is a useful place to start. It bundles more than a hundred pre-trained word vector sets, covering dense SGNS and sparse PPMI representations trained on corpora from Baidu Encyclopedia and Wikipedia to People's Daily, financial news, Zhihu, and Weibo. Features span word, ngram, character, radical, and dependency variants, so you can match the representation to your task instead of training from scratch. The repo also ships the CA8 analogical reasoning dataset and an evaluation toolkit for judging how well a set handles Chinese morphological and semantic relations. Teams without large Chinese corpora or GPU time get a tested starting point, and the accompanying ACL 2018 paper makes the methodology citable.