Bangla Sentence Transformer

🚀 Bangla Sentence Transformer

Sentence Transformer is a state-of-the-art natural language processing (NLP) model that can encode and transform sentences into high-dimensional embeddings. This technology enables powerful insights and applications in various fields such as text classification, information retrieval, and semantic search.

This model is fine-tuned from stsb-xlm-r-multilingual and is now available on Hugging Face! 🎉🎉

🚀 Quick Start

✨ Features

• Supports multiple languages including Bengali (bn) and English (en).
• Suitable for the task of sentence similarity.
• Based on sentence-transformers technology for feature extraction.

📦 Installation

Using this model becomes easy when you have sentence-transformers installed:

pip install -U sentence-transformers

💻 Usage Examples

Basic Usage

from sentence_transformers import SentenceTransformer
sentences = ['আমি আপেল খেতে পছন্দ করি। ', 'আমার একটি আপেল মোবাইল আছে।','আপনি কি এখানে কাছাকাছি থাকেন?', 'আশেপাশে কেউ আছেন?']

model = SentenceTransformer('shihab17/bangla-sentence-transformer')
embeddings = model.encode(sentences)
print(embeddings)

Advanced Usage

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 = ['আমি আপেল খেতে পছন্দ করি। ', 'আমার একটি আপেল মোবাইল আছে।','আপনি কি এখানে কাছাকাছি থাকেন?', 'আশেপাশে কেউ আছেন?']

# Load model from HuggingFace Hub
tokenizer = AutoTokenizer.from_pretrained('shihab17/bangla-sentence-transformer')
model = AutoModel.from_pretrained('shihab17/bangla-sentence-transformer')

# 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)

Calculating Sentence Similarity

from sentence_transformers import SentenceTransformer
from sentence_transformers.util import pytorch_cos_sim


transformer = SentenceTransformer('shihab17/bangla-sentence-transformer')

sentences = ['আমি আপেল খেতে পছন্দ করি। ', 'আমার একটি আপেল মোবাইল আছে।','আপনি কি এখানে কাছাকাছি থাকেন?', 'আশেপাশে কেউ আছেন?']

sentences_embeddings = transformer.encode(sentences)

for i in range(len(sentences)):
    for j in range(i, len(sentences)):
        sen_1 = sentences[i]
        sen_2 = sentences[j]
        sim_score = float(pytorch_cos_sim(sentences_embeddings[i], sentences_embeddings[j]))
        print(sen_1, '----->', sen_2, sim_score)

📚 Documentation

Best MSE: 2.5556

📄 License

No license information provided in the original document.

📖 Citation

If you use this model, please cite the following paper:

@INPROCEEDINGS{10754765,
  author={Uddin, Md. Shihab and Haque, Mohd Ariful and Rifat, Rakib Hossain and Kamal, Marufa and Gupta, Kishor Datta and George, Roy},
  booktitle={2024 IEEE 15th Annual Ubiquitous Computing, Electronics & Mobile Communication Conference (UEMCON)}, 
  title={Bangla SBERT - Sentence Embedding Using Multilingual Knowledge Distillation}, 
  year={2024},
  volume={},
  number={},
  pages={495-500},
  keywords={Sentiment analysis;Machine learning algorithms;Accuracy;Text categorization;Semantics;Transformers;Mobile communication;Information retrieval;Machine translation;Sentence Similarity;Sentence Transformer;SBERT;Knowledge Distillation;Bangla NLP},
  doi={10.1109/UEMCON62879.2024.10754765}}