Playing word2vec with pre-trained model

Introduction

There are many pre-trained models provided by facebook, google or research institutes (e.g., Academic SINICA in Taiwan). Learners can download these pre-trained models, playing with them, even doing research following Garg el al., (2017).

Downloading pre-trained models

English models

• Facebook FastText https://fasttext.cc/docs/en/crawl-vectors.html
• Google News https://code.google.com/archive/p/word2vec/
• NYTimes model provided by Garg et al.,(2o17) http://stanford.edu/~nkgarg/NYTembeddings/
• https://nlp.stanford.edu/projects/histwords/

Chinese models

• CKIP http://ckip.iis.sinica.edu.tw:8080/license/ total 1.2GB
• https://github.com/Embedding/Chinese-Word-Vectors
• https://github.com/facebookresearch/fastText/blob/master/docs/crawl-vectors.md

Reference

1. Garg, N., Schiebinger, L., Jurafsky, D., & Zou, J. (2017). Word Embeddings Quantify 100 Years of Gender and Ethnic Stereotypes, 115(16). https://doi.org/10.1073/pnas.1720347115
2. Xu, H., Xiao, D., Wu, L., & Wang, C.-J. (2018). The Hidden Shape of Stories Reveals Positivity Bias and Gender Bias. Retrieved from https://arxiv.org/abs/1811.04599