COVID-19 Open Research Dataset

Full-text and metadata dataset of COVID-19 and coronavirus-related scholarly articles optimized for machine readability and made available for use by the global research community.

In response to the COVID-19 pandemic, the Allen Institute for AI has partnered with leading research groups to prepare and distribute the COVID-19 Open Research Dataset (CORD-19). This dataset is a free resource of over 47,000 scholarly articles, including over 36,000 with full text, about COVID-19 and the coronavirus family of viruses for use by the global research community.

This dataset mobilizes researchers to apply recent advances in natural language processing to generate new insights in support of the fight against this infectious disease.

The corpus may be updated as new research is published in peer-reviewed publications and archival services like bioRxiv, medRxiv, and others.


Microsoft provides Azure Open Datasets on an “as is” basis. Microsoft makes no warranties, express or implied, guarantees or conditions with respect to your use of the datasets. To the extent permitted under your local law, Microsoft disclaims all liability for any damages or losses, including direct, consequential, special, indirect, incidental or punitive, resulting from your use of the datasets.

This dataset is provided under the original terms that Microsoft received source data. The dataset may include data sourced from Microsoft.

License Terms

This dataset is made available by the Allen Institute of AI and Semantic Scholar. By accessing, downloading, or otherwise using any content provided in the CORD-19 Dataset, you agree to the Dataset License related to the use this dataset. Specific licensing information for individual articles in the dataset is available in the metadata file. More licensing information is available on the PMC website, medRxiv website, and bioRxiv website.

Volume and retention

This dataset is stored in JSON format and the latest release contains over 36,000 full text articles. Each paper is represented as a single JSON object. View the schema.

Storage Location

This dataset is stored in the East US Azure region. Locating compute resources in East US is recommended for affinity.


When including CORD-19 data in a publication or redistribution, cite the dataset as follows:

In bibliography:

COVID-19 Open Research Dataset (CORD-19). 2020. Version YYYY-MM-DD. Retrieved from COVID-19 Open Research Dataset (CORD-19). Accessed YYYY-MM-DD. doi:10.5281/zenodo.3715505

In text: (CORD-19, 2020)


For questions about this dataset, contact

Data access

Azure Notebooks

The CORD-19 Dataset

CORD-19 is a collection of over 50,000 scholarly articles - including over 40,000 with full text - about COVID-19, SARS-CoV-2, and related corona viruses. This dataset has been made freely available with the goal to aid research communities combat the COVID-19 pandemic.

The goal of this notebook is two-fold:

  1. Demonstrate how to access the CORD-19 dataset on Azure: We connect to the Azure blob storage account housing the CORD-19 dataset.
  2. Walk though the structure of the dataset: Articles in the dataset are stored as json files. We provide examples showing:
  • How to find the articles (navigating the container)
  • How to read the articles (navigating the json schema)

Dependencies: This notebook requires the following libraries:

  • Azure storage (for example, pip install azure-storage)
  • NLTK (docs)
  • Pandas (for example, pip install pandas)

Getting the CORD-19 data from Azure

The CORD-19 data has been uploaded as an Azure Open Dataset here. We create a blob service linked to this CORD-19 open dataset.

from import BlockBlobService

# storage account details
azure_storage_account_name = "azureopendatastorage"
azure_storage_sas_token = "sv=2019-02-02&ss=bfqt&srt=sco&sp=rlcup&se=2025-04-14T00:21:16Z&st=2020-04-13T16:21:16Z&spr=https&sig=JgwLYbdGruHxRYTpr5dxfJqobKbhGap8WUtKFadcivQ%3D"

# create a blob service
blob_service = BlockBlobService(

We can use this blob service as a handle on the data. We can navigate the dataset making use of the BlockBlobService APIs. See here for more details:

The CORD-19 data is stored in the covid19temp container. This is the file structure within the container together with an example file.

        0001418189999fea7f7cbe3e82703d71c85a6fe5.json        # filename is sha-hash
        PMC1065028.xml.json                                  # filename is the PMC ID
biorxiv_medrxiv/                                             # note: there is no pmc_json subdir

Each .json file corresponds to an individual article in the dataset. This is where the title, authors, abstract and (where available) the full text data is stored.

Using metadata.csv

The CORD-19 dataset comes with metadata.csv - a single file that records basic information on all the papers available in the CORD-19 dataset. This is a good place to start exploring!

# container housing CORD-19 data
container_name = "covid19temp"

# download metadata.csv
metadata_filename = 'metadata.csv'
import pandas as pd

# read metadata.csv into a dataframe
metadata_filename = 'metadata.csv'
metadata = pd.read_csv(metadata_filename)

That's a lot to take in at first glance, so let's apply a little polish.

simple_schema = ['cord_uid', 'source_x', 'title', 'abstract', 'authors', 'full_text_file', 'url']

def make_clickable(address):
    '''Make the url clickable'''
    return '<a href="{0}">{0}</a>'.format(address)

def preview(text):
    '''Show only a preview of the text data.'''
    return text[:30] + '...'

format_ = {'title': preview, 'abstract': preview, 'authors': preview, 'url': make_clickable}

# let's take a quick look around
num_entries = len(metadata)
print("There are {} many entries in this dataset:".format(num_entries))

metadata_with_text = metadata[metadata['full_text_file'].isna() == False]
with_full_text = len(metadata_with_text)
print("-- {} have full text entries".format(with_full_text))

with_doi = metadata['doi'].count()
print("-- {} have DOIs".format(with_doi))

with_pmcid = metadata['pmcid'].count()
print("-- {} have PubMed Central (PMC) ids".format(with_pmcid))

with_microsoft_id = metadata['Microsoft Academic Paper ID'].count()
print("-- {} have Microsoft Academic paper ids".format(with_microsoft_id))
There are 51078 many entries in this dataset:
-- 42511 have full text entries
-- 47741 have DOIs
-- 41082 have PubMed Central (PMC) ids
-- 964 have Microsoft Academic paper ids

Example: Read full text

metadata.csv doesn't contain the full-text itself. Let's see an example of how to read that. Locate and unpack the full text json and convert it to a list of sentences.

# choose a random example with pdf parse available
metadata_with_pdf_parse = metadata[metadata['has_pdf_parse']]
example_entry = metadata_with_pdf_parse.iloc[42]

# construct path to blob containing full text
blob_name = '{0}/pdf_json/{1}.json'.format(example_entry['full_text_file'], example_entry['sha'])  # note the repetition in the path
print("Full text blob for this entry:")

We can now read the json content associated to this blob as follows.

import json
blob_as_json_string = blob_service.get_blob_to_text(container_name=container_name, blob_name=blob_name)
data = json.loads(blob_as_json_string.content)

# in addition to the body text, the metadata is also stored within the individual json files
print("Keys within data:", ', '.join(data.keys()))

For the purposes of this example, we are interested in the body_text, which stores the text data as follows:

"body_text": [                      # list of paragraphs in full body
        "text": <str>,
        "cite_spans": [             # list of character indices of inline citations
                                    # e.g. citation "[7]" occurs at positions 151-154 in "text"
                                    #      linked to bibliography entry BIBREF3
                "start": 151,
                "end": 154,
                "text": "[7]",
                "ref_id": "BIBREF3"
        "ref_spans": <list of dicts similar to cite_spans>,     # e.g. inline reference to "Table 1"
        "section": "Abstract"

The full json schema is available here.

from nltk.tokenize import sent_tokenize

# the text itself lives under 'body_text'
text = data['body_text']

# many NLP tasks play nicely with a list of sentences
sentences = []
for paragraph in text:

print("An example sentence:", sentences[0])

PDF vs PMC XML Parse

In the above example, we looked at a case with has_pdf_parse == True. In that case the blob file path was of the form:


Alternatively, for cases with has_pmc_xml_parse == True use the following format:


For example:

# choose a random example with pmc parse available
metadata_with_pmc_parse = metadata[metadata['has_pmc_xml_parse']]
example_entry = metadata_with_pmc_parse.iloc[42]

# construct path to blob containing full text
blob_name = '{0}/pmc_json/{1}.xml.json'.format(example_entry['full_text_file'], example_entry['pmcid'])  # note the repetition in the path
print("Full text blob for this entry:")

blob_as_json_string = blob_service.get_blob_to_text(container_name=container_name, blob_name=blob_name)
data = json.loads(blob_as_json_string.content)

# the text itself lives under 'body_text'
text = data['body_text']

# many NLP tasks play nicely with a list of sentences
sentences = []
for paragraph in text:

print("An example sentence:", sentences[0])
Full text blob for this entry:
An example sentence: Double-stranded small interfering RNA (siRNA) molecules have drawn much attention since it was unambiguously shown that they mediate potent gene knock-down in a variety of mammalian cells (1).

Iterate through blobs directly

In the above examples, we used the metadata.csv file to navigate the data, construct the blob file path, and read data from the blob. An alternative is the iterate through the blobs themselves.

# get and sort list of available blobs
blobs = blob_service.list_blobs(container_name)
sorted_blobs = sorted(list(blobs), key=lambda e:, reverse=True)

Now we can iterate through the blobs directly. For example, let's count the number json files available.

# we can now iterate directly though the blobs
count = 0
for blob in sorted_blobs:
    if[-5:] == ".json":
        count += 1
print("There are {} many json files".format(count))
There are 59784 many json files


Data quality issues

This is a large dataset that, for obvious reasons, was put together rather hastily! Here are some data quality issues we've observed.

Multiple shas

We observe that in some cases there are multiple shas for a given entry.

metadata_multiple_shas = metadata[metadata['sha'].str.len() > 40]

print("There are {} many entries with multiple shas".format(len(metadata_multiple_shas)))

There are 1999 many entries with multiple shas

Layout of the container

Here we use a simple regex to explore the file structure of the container in case this is updated in the future.

container_name = "covid19temp"
blobs = blob_service.list_blobs(container_name)
sorted_blobs = sorted(list(blobs), key=lambda e:, reverse=True)
import re
dirs = {}

pattern = '([\w]+)\/([\w]+)\/([\w.]+).json'
for blob in sorted_blobs:
    m = re.match(pattern,
    if m:
        dir_ = m[1] + '/' + m[2]
        if dir_ in dirs:
            dirs[dir_] += 1
            dirs[dir_] = 1

Next steps

View the rest of the datasets in the Open Datasets catalog.