---
license: other
license_name: public-domain
license_link: LICENSE
---
# MedCPT Introduction

**MedCPT generates embeddings of biomedical texts that can be used for semantic search (dense retrieval)**. The model contains two encoders:
- [MedCPT Query Encoder](https://huggingface.co/ncbi/MedCPT-Query-Encoder): compute the embeddings of short texts (e.g., questions, search queries, sentences).
- [MedCPT Article Encoder](https://huggingface.co/ncbi/MedCPT-Article-Encoder): compute the embeddings of articles (e.g., PubMed titles & abstracts).

**This repo contains the MedCPT Query Encoder.**

**MedCPT has been pre-trained by an unprecedented scale of 255M query-article pairs from PubMed search logs**, and has been shown to achieve state-of-the-art performance on several zero-shot biomedical IR datasets. In general, there are three use cases:
1. Query-to-article search with both encoders.
2. Query representation for clustering or query-to-query search with the [query encoder](https://huggingface.co/ncbi/MedCPT-Query-Encoder).
3. Article representation for clustering or article-to-article search with the [article encoder](https://huggingface.co/ncbi/MedCPT-Article-Encoder).

For more details, please check out our [paper](https://arxiv.org/abs/2307.00589) (Bioinformatics, 2023). Please note that the released version is slightly different from the version reported in the paper.

# Case 1. Using the MedCPT Query Encoder

```python
import torch
from transformers import AutoTokenizer, AutoModel

model = AutoModel.from_pretrained("ncbi/MedCPT-Query-Encoder")
tokenizer = AutoTokenizer.from_pretrained("ncbi/MedCPT-Query-Encoder")

queries = [
	"diabetes treatment", 
	"How to treat diabetes?", 
	"A 45-year-old man presents with increased thirst and frequent urination over the past 3 months.",
]

with torch.no_grad():
	# tokenize the queries
	encoded = tokenizer(
		queries, 
		truncation=True, 
		padding=True, 
		return_tensors='pt', 
		max_length=64,
	)
	
	# encode the queries (use the [CLS] last hidden states as the representations)
	embeds = model(**encoded).last_hidden_state[:, 0, :]

	print(embeds)
	print(embeds.size())
```
The output will be:
```bash
tensor([[ 0.0413,  0.0084, -0.0491,  ..., -0.4963, -0.3830, -0.3593],
        [ 0.0801,  0.1193, -0.0905,  ..., -0.5380, -0.5059, -0.2944],
        [-0.3412,  0.1521, -0.0946,  ...,  0.0952,  0.1660, -0.0902]])
torch.Size([3, 768])
```
These embeddings are also in the same space as those generated by the MedCPT article encoder. 

# Case 2. Semantically searching PubMed with your query

We have provided the embeddings of all PubMed articles generated by the MedCPT article encoder at https://ftp.ncbi.nlm.nih.gov/pub/lu/MedCPT/pubmed_embeddings/.
You can simply download these embeddings to search PubMed with your query.

# Acknowledgments

This work was supported by the Intramural Research Programs of the National Institutes of Health, National Library of Medicine.

# Disclaimer

This tool shows the results of research conducted in the Computational Biology Branch, NCBI/NLM. The information produced on this website is not intended for direct diagnostic use or medical decision-making without review and oversight by a clinical professional. Individuals should not change their health behavior solely on the basis of information produced on this website. NIH does not independently verify the validity or utility of the information produced by this tool. If you have questions about the information produced on this website, please see a health care professional. More information about NCBI's disclaimer policy is available.

# Citation

If you find this repo helpful, please cite MedCPT by:

```bibtext
@article{jin2023medcpt,
  title={MedCPT: Contrastive Pre-trained Transformers with large-scale PubMed search logs for zero-shot biomedical information retrieval},
  author={Jin, Qiao and Kim, Won and Chen, Qingyu and Comeau, Donald C and Yeganova, Lana and Wilbur, W John and Lu, Zhiyong},
  journal={Bioinformatics},
  volume={39},
  number={11},
  pages={btad651},
  year={2023},
  publisher={Oxford University Press}
}
```