The difference between Tutorials A "use word embeddings in the pretrained bert model" and Tutorials C "bert+lstm/blstm-crf/enc-dec focus model models"

[LXM-Emily]

Hello, your work is very excellent. However, due to my poor understanding, I have some questions when reading the code, and would like to ask for for your help. That is, BERT is used in Tutorial A to obtain word embedding, while BERT+ BLSTM model in Tutorial C is also used to obtain word embedding. What is the difference between the two? And if I want to use bert to get word embeddings and use blstm as the basic model, should I use tutorial A or tutorial C? I will be very happy if you can reply me.

[gongel]

Tutorial A is to extract BERT embedding from dataset and has two steps. If you train BERT + LSTM for two tasks, use Tutorial C is better.

[LXM-Emily]

Thank you for your reply and advice vey much. Can I understand it this way? Tutorial A takes the mean of word vectors for the same words in the dataset, and each word corresponds to a word vector, while Tutorial C takes a sentence as a unit and inputs bert to obtain the word vector of this sentence. Thank you again!