E5 Base Korean

🚀 upskyy/e5-base-korean

This model is a fine - tuned model on korsts and kornli, derived from [intfloat/multilingual - e5 - base](https://huggingface.co/intfloat/multilingual - e5 - base). It maps sentences and paragraphs into a 768 - dimensional dense vector space, which can be used for semantic textual similarity, semantic search, paraphrase mining, text classification, clustering, etc.

🚀 Quick Start

This model is a fine - tuned version of intfloat/multilingual - e5 - base on korsts and kornli datasets. It can map sentences and paragraphs into a 768 - dimensional dense vector space, enabling various NLP tasks such as semantic similarity calculation and text classification.

✨ Features

• Multilingual Support: Supports multiple languages including Korean, making it suitable for cross - language NLP tasks.
• High - Dimensional Embeddings: Maps text to a 768 - dimensional dense vector space for accurate semantic representation.
• Versatile Applications: Can be used for semantic textual similarity, semantic search, paraphrase mining, text classification, clustering, etc.

📦 Installation

First, you need to install the sentence - transformers library:

pip install -U sentence-transformers

💻 Usage Examples

Basic Usage

from sentence_transformers import SentenceTransformer

# Download from the 🤗 Hub
model = SentenceTransformer("upskyy/e5-base-korean")

# Run inference
sentences = [
    '아이를 가진 엄마가 해변을 걷는다.',
    '두 사람이 해변을 걷는다.',
    '한 남자가 해변에서 개를 산책시킨다.',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 768]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities.shape)
# [3, 3]

Advanced Usage

Without sentence - transformers, you can use the model like this: First, you pass your input through the transformer model, then you have to apply the right pooling - operation on - top of the contextualized word embeddings.

from transformers import AutoTokenizer, AutoModel
import torch


# Mean Pooling - Take attention mask into account for correct averaging
def mean_pooling(model_output, attention_mask):
    token_embeddings = model_output[0] # First element of model_output contains all token embeddings
    input_mask_expanded = attention_mask.unsqueeze(-1).expand(token_embeddings.size()).float()
    return torch.sum(token_embeddings * input_mask_expanded, 1) / torch.clamp(input_mask_expanded.sum(1), min=1e-9)


# Sentences we want sentence embeddings for
sentences = ["안녕하세요?", "한국어 문장 임베딩을 위한 버트 모델입니다."]

# Load model from HuggingFace Hub
tokenizer = AutoTokenizer.from_pretrained("upskyy/e5-base-korean")
model = AutoModel.from_pretrained("upskyy/e5-base-korean")

# Tokenize sentences
encoded_input = tokenizer(sentences, padding=True, truncation=True, return_tensors="pt")

# Compute token embeddings
with torch.no_grad():
    model_output = model(**encoded_input)

# Perform pooling. In this case, mean pooling.
sentence_embeddings = mean_pooling(model_output, encoded_input["attention_mask"])

print("Sentence embeddings:")
print(sentence_embeddings)

📚 Documentation

Model Details

Full Model Architecture

SentenceTransformer(
  (0): Transformer({'max_seq_length': 512, 'do_lower_case': False}) with Transformer model: XLMRobertaModel 
  (1): Pooling({'word_embedding_dimension': 768, 'pooling_mode_cls_token': False, 'pooling_mode_mean_tokens': True, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False, 'include_prompt': True})
)

Evaluation

Semantic Similarity

• Dataset: sts - dev
• Evaluated with EmbeddingSimilarityEvaluator

Framework Versions

• Python: 3.10.13
• Sentence Transformers: 3.0.1
• Transformers: 4.42.4
• PyTorch: 2.3.0+cu121
• Accelerate: 0.30.1
• Datasets: 2.16.1
• Tokenizers: 0.19.1

🔧 Technical Details

This model is fine - tuned on korsts and kornli datasets based on the intfloat/multilingual - e5 - base model. It uses a XLMRobertaModel as the backbone and a Pooling layer for post - processing. The pooling operation is set to mean pooling, which can effectively aggregate the information of contextualized word embeddings.

📄 License

This project is licensed under the MIT license.

📖 Citation

BibTeX

@article{wang2024multilingual,
  title={Multilingual E5 Text Embeddings: A Technical Report},
  author={Wang, Liang and Yang, Nan and Huang, Xiaolong and Yang, Linjun and Majumder, Rangan and Wei, Furu},
  journal={arXiv preprint arXiv:2402.05672},
  year={2024}
}

@inproceedings{reimers-2019-sentence-bert,
    title = "Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks",
    author = "Reimers, Nils and Gurevych, Iryna",
    booktitle = "Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing",
    month = "11",
    year = "2019",
    publisher = "Association for Computational Linguistics",
    url = "https://arxiv.org/abs/1908.10084",
}