Can Sentence Bert embed an entire paragraph instead of only a sentence? For example, a description of a movie.
If so, what if the word counts exceeded the input limit? I remember Bert can only take up to 512 tokens as inputs?
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
2
Since BERT is designed for a sentence it captures the context in a sentence. However in a paragraph they will be number of sentences and a number of context. it is not a good idea to insert paragraph in BERT. you will not get good results You should tokenize paragraph in sentences using nltk or spacy
-
$\begingroup$ Thank you very much! Is there a way to embed a whole paragraph? should I take the mean of all sentences' embeddings in the paragraph? $\endgroup$ Aug 14, 2022 at 7:46
-
$\begingroup$ Not sure but logically The mean of embeddings should give information about the whole paragraph. $\endgroup$ Aug 14, 2022 at 8:06