Averaging the vector representation of words found in the passage or sentence.

General pretrained embeddings and embeddings from specific domains are less effective in classifying texts.

Some researchers suggest using embeddings from the passage rather than averaged embedding to be used.