PromCSE
Code for "Improved Universal Sentence Embeddings with Prompt-based Contrastive Learning and Energy-based Learning (EMNLP 2022)" (by YJiangcm)
SimCSE
[EMNLP 2021] SimCSE: Simple Contrastive Learning of Sentence Embeddings https://arxiv.org/abs/2104.08821 (by princeton-nlp)
The number of mentions indicates the total number of mentions that we've tracked plus the number of user suggested alternatives.
Stars - the number of stars that a project has on GitHub. Growth - month over month growth in stars.
Activity is a relative number indicating how actively a project is being developed. Recent commits have higher weight than older ones.
For example, an activity of 9.0 indicates that a project is amongst the top 10% of the most actively developed projects that we are tracking.
Stars - the number of stars that a project has on GitHub. Growth - month over month growth in stars.
Activity is a relative number indicating how actively a project is being developed. Recent commits have higher weight than older ones.
For example, an activity of 9.0 indicates that a project is amongst the top 10% of the most actively developed projects that we are tracking.
PromCSE
Posts with mentions or reviews of PromCSE.
We have used some of these posts to build our list of alternatives
and similar projects. The last one was on 2022-05-03.
-
State of the Art in Sentence Embeddings
To answer your question about sentence embedding SOTA, it is not s-Bert and hasn't been for a while. SimCSE officially takes the crown since it's been presented at a conference, though according to paperswithcode's benchmark leaderboard there are other papers on arxiv that report higher performance on STS and similar tasks such as DCPCSE. Having tried both of these for my use case I found SimCSE to be better but YMMV.
SimCSE
Posts with mentions or reviews of SimCSE.
We have used some of these posts to build our list of alternatives
and similar projects. The last one was on 2022-05-03.
-
BERT-Based Clustering on a Corpus of Genre Samples Kinda Sucks. Why?
Base BERT sentence embeddings are just not good for a couple of reasons and there's some research papers that show this. You can try SimCSE, Google's USE or SBERT as mentioned previously and you'll get better output. It's just an inherent flaw to base BERT that it can't produce good sentence embeddings. Papers have shown you probably will get better scores using GloVe embeddings from scratch than base BERT.
-
State of the Art in Sentence Embeddings
To answer your question about sentence embedding SOTA, it is not s-Bert and hasn't been for a while. SimCSE officially takes the crown since it's been presented at a conference, though according to paperswithcode's benchmark leaderboard there are other papers on arxiv that report higher performance on STS and similar tasks such as DCPCSE. Having tried both of these for my use case I found SimCSE to be better but YMMV.
What are some alternatives?
When comparing PromCSE and SimCSE you can also consider the following projects:
nlu - 1 line for thousands of State of The Art NLP models in hundreds of languages The fastest and most accurate way to solve text problems.
inltk - Natural Language Toolkit for Indic Languages aims to provide out of the box support for various NLP tasks that an application developer might need
DiffCSE - Code for the NAACL 2022 long paper "DiffCSE: Difference-based Contrastive Learning for Sentence Embeddings"
AnnA_Anki_neuronal_Appendix - Using machine learning on your anki collection to enhance the scheduling via semantic clustering and semantic similarity
BERTopic - Leveraging BERT and c-TF-IDF to create easily interpretable topics.