Embedding a text to a vector by pre-trained BERT word embeddings and pooling layers, for the pur[ose of text similarity measuring
