This is my BERT implementation (using Pytorch) for the social media corpus (TweetNLP)
Paper: Improved Part-of-Speech Tagging for Online Conversational Text with Word Clusters, Owoputi et al. NAACL 2013
Annotation guideline: http://www.cs.cmu.edu/~ark/TweetNLP/annot_guidelines.pdf