python - torchtext如何汇总每个文档的单词向量？(How torchtext aggregates word vectors for each document?)

Question

I am new to torchtext and I know that we can use pre-trained word embeddings for documents using the following code:

(我是torchtext的新手，我知道我们可以使用以下代码对文档进行预训练的单词嵌入：)

text_field.build_vocab(train_data, vectors="glove.6B.100d")

where text_field is a torchtext.data.Field object.

(其中text_field是torchtext.data.Field对象。)

Now my question is that how torchtext aggregates individual word vectors for a document?

(现在我的问题是，torchtext如何聚合文档的各个单词向量？)

Is it using a simple averaging over vectors or something more advanced like LSTM?

(它是使用矢量的简单平均还是更高级的LSTM？)

  ask by sisaman translate from so

Answer 1

等待大神答复