Whats the difference between nnSent and get_sentence_embeddings_from_pre-trained_models?

[tqx94]

Hi,

If I want to find sentences related to sentences Im querying, which command should I use? NNsent./fasttext nnSent model.bin corpora [k] directly or get_sentence_embeddings in https://github.com/epfml/sent2vec/blob/master/get_sentence_embeddings_from_pre-trained_models.ipynb and then manually find the cosine similarity?

If i were to use nnsent command: ./fasttext nnSent model.bin corpora [k]:

• for the corpora that I upload, do I need to preprocess the texts by using stanford jar to match the pretrained models? Because in the get_sentence_embedding ipynb, i have to tokenise the sentences to match the models.
• Do i need to clean the input text to be queried too?

Thanks!

[mpagli]

If I want to find sentences related to sentences Im querying, which command should I use?

Both options are fine.

for the corpora that I upload, do I need to preprocess the texts by using stanford jar to match the pretrained models?

Yes, now there is a simpler option than this jar, check this official python wrapper: https://github.com/stanfordnlp/stanfordnlp

Do i need to clean the input text to be queried too

Yes