---
license: mit
task_categories:
- text-classification
- summarization
- text2text-generation
- sentence-similarity
- text-generation
language:
- en
tags:
- biology
- chemistry
- medical
size_categories:
- 100M<n<1B
---


# About Dataset

- Dataset name: arXiv academic paper metadata
- Data source: https://arxiv.org/
- Submission date: 1986-04-25 ~ 2025-05-13 (data updated weekly)
- Number of papers: 2,710,806 (as of 2025.5.14)
- Fields included: title, author, abstract, journal information, DOI, etc.
- Data format: json
- Data volume: 4.58G

## About ArXiv

For nearly 30 years, ArXiv has served the public and research communities by providing open access to scholarly articles, from the vast branches of physics to the many subdisciplines of computer science to everything in between, including math, statistics, electrical engineering, quantitative biology, and economics. This rich corpus of information offers significant, but sometimes overwhelming depth.

In these times of unique global challenges, efficient extraction of insights from data is essential. To help make the arXiv more accessible, we present a free, open pipeline on Kaggle to the machine-readable arXiv dataset: a repository of 1.7 million articles, with relevant features such as article titles, authors, categories, abstracts, full text PDFs, and more.

Our hope is to empower new use cases that can lead to the exploration of richer machine learning techniques that combine multi-modal features towards applications like trend analysis, paper recommender engines, category prediction, co-citation networks, knowledge graph construction and semantic search interfaces.

The dataset is freely available via Google Cloud Storage buckets (more info here). Stay tuned for weekly updates to the dataset!

ArXiv is a collaboratively funded, community-supported resource founded by Paul Ginsparg in 1991 and maintained and operated by Cornell University.

The release of this dataset was featured further in a Kaggle blog post here.



See [here](https://www.kaggle.com/datasets/Cornell-University/arxiv) for more information.

## ArXiv On Kaggle 
### Metadata

This dataset is a mirror of the original ArXiv data. Because the full dataset is rather large (1.1TB and growing), this dataset provides only a metadata file in the json format. This file contains an entry for each paper, containing:

id: ArXiv ID (can be used to access the paper, see below)
submitter: Who submitted the paper
authors: Authors of the paper
title: Title of the paper
comments: Additional info, such as number of pages and figures
journal-ref: Information about the journal the paper was published in
doi: [https://www.doi.org](Digital Object Identifier)
abstract: The abstract of the paper
categories: Categories / tags in the ArXiv system
versions: A version history
You can access each paper directly on ArXiv using these links:
```
https://arxiv.org/abs/{id}: Page for this paper including its abstract and further links
https://arxiv.org/pdf/{id}: Direct link to download the PDF
```
Bulk access
The full set of PDFs is available for free in the GCS bucket `gs://arxiv-dataset` or through Google API (json documentation and xml documentation).

You can use for example gsutil to download the data to your local machine.

### List files:
```
gsutil cp gs://arxiv-dataset/arxiv/
```
### Download pdfs from March 2020:
```
gsutil cp gs://arxiv-dataset/arxiv/arxiv/pdf/2003/ ./a_local_directory/
```
### Download all the source files
```
gsutil cp -r gs://arxiv-dataset/arxiv/  ./a_local_directory/
```
Update Frequency
We're automatically updating the metadata as well as the GCS bucket on a weekly basis.

### License
Creative Commons CC0 1.0 Universal Public Domain Dedication applies to the metadata in this dataset. See https://arxiv.org/help/license for further details and licensing on individual papers.

### Acknowledgements
The original data is maintained by ArXiv, huge thanks to the team for building and maintaining this dataset.

We're using https://github.com/mattbierbaum/arxiv-public-datasets to pull the original data, thanks to Matt Bierbaum for providing this tool.
# Ues example

## 1.load data
```
import pandas as pd

df = pd.read_json('arxiv-metadata-oai-snapshot.json', lines=True)
print('论文数量: ', len(df))
df.head()
```

![image/png](https://cdn-uploads.huggingface.co/production/uploads/63101b3d982455e61cd02275/h0q_F_rtdyoO86eBmtc6X.png)

## 2.Contained fields

```py
for col in df.columns:
    print(f' - {col}')
```
```sh
- id ArXiv ID (can be used to access the paper, see below for details)
- submitter Paper submitter
- authors Paper authors
- title Title
- comments Additional information, such as number of pages and figures
- journal-ref Information about the journal where the paper was published
- doi DOI number (digital object identifier) ​​of the paper
- report-no Report number (authors have obtained the report number from their institution before submitting to arXiv)
- categories arXiv system categories/tags
- license License agreement under which the paper is based
- abstract Paper abstract
- versions Version history
- update_date Paper update date
- authors_parsed Paper authors
```

## 3.Annual number of papers submitted
```py
from plotnine import *
from matplotlib.font_manager import FontProperties

font_prop = FontProperties(fname='文泉驿微米黑.ttf')

data = df.created.dt.year.value_counts().sort_index().reset_index()


(
    ggplot(data, aes(x='created', y='count'))
    +geom_bar(stat='identity')
    +labs(title='arXiv论文年度提交量趋势图(1986-2025.3)',
          x='',
          y='')
    + annotate(
        'text',
        x= 1986,
        y= data['count'].max() * 1.05,
        label='Python',
        ha='left',
        va='top',
        size=10,
        color='black',
    )
    +scale_x_continuous(breaks=range(1986, 2026, 3))
    +theme(figure_size=(8, 5),
           text=element_text(family= font_prop.get_name(), size=8, rotation=0),
           plot_title = element_text(size=13, rotation=0, weight='bold'),
           axis_text_x = element_text(size=8, rotation=0))
)
```


![image/png](https://cdn-uploads.huggingface.co/production/uploads/63101b3d982455e61cd02275/MZoVexsRVAtWgXUiiXasO.png)


# 4.Distribution of the number of authors

```py
from plotnine import *
from matplotlib.font_manager import FontProperties

font_prop = FontProperties(fname='文泉驿微米黑.ttf')

data = df['N_authors'].value_counts().sort_index(ascending=True).reset_index()

(
    ggplot(data.head(20), aes(x='N_authors', y='count'))
    +geom_bar(stat='identity')
    +labs(title='arXiv每篇论文作者数量分布图',
          x='',
          y='')
    + annotate(
        'text',
        x= 16,
        y= data['count'].max() ,
        label='Python',
        ha='left',
        va='top',
        size=10,
        color='black',
    )
    +scale_x_continuous(breaks=range(0, 21, 1))
    +theme(figure_size=(8, 5),
           text=element_text(family= font_prop.get_name(), size=8, rotation=0),
           plot_title = element_text(size=13, rotation=0, weight='bold'),
           axis_text_x = element_text(size=8, rotation=0))
)
```



![image/png](https://cdn-uploads.huggingface.co/production/uploads/63101b3d982455e61cd02275/wD0iyPXATOSpPFS3ym7Am.png)

# 5.Top 10 research hotspots

```py
from plotnine import *
import pandas as pd
from matplotlib.font_manager import FontProperties

font_prop = FontProperties(fname='文泉驿微米黑.ttf')
data = df['categories'].value_counts().reset_index()
data.sort_values(by='count', ascending=False, inplace=True)
data['categories'] = pd.Categorical(data['categories'], categories=data['categories'], ordered=True)

(
    ggplot(data.head(10), aes(x='categories', y='count'))  # 注意这里x和y的位置
    + geom_bar(stat='identity')
    + labs(title='arXiv前十大研究热点',
           x='',
           y='')
    + coord_flip()  # 翻转坐标轴以创建水平条形图
    + theme(figure_size=(8, 4),
            text=element_text(family=font_prop.get_name(), size=8, rotation=0),
            plot_title=element_text(size=12, rotation=0, weight='bold'),
            axis_text_x=element_text(size=8, rotation=0))
)
```

![image/png](https://cdn-uploads.huggingface.co/production/uploads/63101b3d982455e61cd02275/XqHq4cTqggsWyzmXa8-0m.png)

# update
This data will be updated monthly

- 2025.4.14