search

UKPLab / sentence-transformers

State-of-the-Art Text Embeddings
https://www.sbert.net
Apache License 2.0
14.83k stars 2.44k forks source link

Add paraphrasing model #545

Open Weilin37 opened 3 years ago

Weilin37 commented 3 years ago

Paraphrasing is very similar to sentence similarity, but with the crucial difference that it's on longer pieces of text.

Model fine-tuned on Microsoft Research Paraphrase Corpus should do it.

nreimers commented 3 years ago

Have a look at these models: https://www.sbert.net/docs/pretrained_models.html#paraphrase-identification

They were trained on 50+ Million paraphrase sentence pairs.

Weilin37 commented 3 years ago

This useful! many thanks.

Is it sensible to just use the semantic similarity examples and just swap out with the paraphrasing model (cluster the embedding). Or is there a different process you'd recommend?

aqsa27 commented 3 years ago

Hi @Weilin37 i am having the same question, any update on this?

nreimers commented 3 years ago

@aqsa27 Could you restate what you exact question is?