search

jinglescode / textual-similarity-universal-sentence-encoder

Extract embeddings and group sentences with universal sentence encoder package from TensorFlow.js.
https://jinglescode.github.io/textual-similarity-universal-sentence-encoder
Apache License 2.0
42 stars 10 forks source link

Request for Information #2

Open nikhilkomakula opened 2 years ago

nikhilkomakula commented 2 years ago

Hello Hong Jing,

Thanks for the great article.

I am using your article as a reference to implement semantic similarity in my application and would like to ask few questions regarding the same:

  1. I am assuming that as part of installing TensorFlow and USE, we are actually downloading the entire model locally within the application. If so, I still see some api calls to https://storage.googleapis.com. Any idea?
  2. It is really taking long time when requesting word embeddings for more than like 100 sentences. At times, my application is also breaking with weird errors. Is there a way to better optimise it? The number of sentences requesting for word embeddings might even go upto like 2000 or more.

Thanks in advance.

Best Regards, Nikhil