Hugging Face's logo

Datasets:
m-a-p
/
PIN-200M

Modalities:
Text
Formats:
json
Languages:
English
Chinese
Size:
10K - 100K
ArXiv:
Tags:
multimodal
interleaved
Libraries:
Datasets
pandas
License:
Dataset card Data Studio Files
xet
Community
1
PIN-200M / README.md
wanng's picture
wanng
Upload README.md
0fb857e verified
preview code
|
raw
history blame contribute delete
34.5 kB
 ---
license: apache-2.0
language:
- en
- zh
configs:
- config_name: pin
data_files:
- split: train
path:
- data/DocLayNet/DocLayNet.jsonl
tags:
- multimodal
- interleaved
size_categories:
- 100M<n<1B
pretty_name: pin-200m
---
# PIN-200M
A mini version of "PIN: A Knowledge-Intensive Dataset for Paired and Interleaved Multimodal Documents"
Paper: https://arxiv.org/abs/2406.13923
This dataset contains around **200M** samples in PIN format, with around **280** TB storage.
🚀 News
[ 2025.09.22 ] !NEW! 🔥 We have completed the final version of the PIN-200M dataset and conducted some simple statistics on it.
[ 2024.12.06 ] !NEW! 🔥 We have updated the quality signals, enabling a swift assessment of whether a sample meets the required specifications based on our quality indicators. Further detailed descriptions will be provided in the forthcoming formal publication. (Aside from the Chinese-Markdown subset, there are unresolved issues that are currently being addressed.)
This dataset contains 200M samples with PIN format.
<img src="assets/intro.png">
## 0 Usage
Download ALL files
```bash
huggingface-cli download m-a-p/PIN-200M --repo-type=dataset --resume-download --local-dir "your_local_path"
```
Download ONLY **Jsonl** files
```bash
huggingface-cli download m-a-p/PIN-200M --repo-type=dataset --resume-download --include "*.jsonl" --local-dir "your_local_path"
```
Decompression
```bash
cat data.tar.part* > data.tar
tar -xvf data.tar
```
## 1 Dataset statistics
| Subsect | Documents (#) | Overall images (#) | Content images (#) | Documents (GB) | Overall images (GB) | Content images (GB) |
|-----------------|-----------|----------------|----------------|---------------------|--------------------------|-----------------------|
| pg19 | 2,611,921 | 2,607,797 | 0 | 12.61 | 1,384.89 | 0.00 |
| OBELICS | 176,600,260 | 175,307,522 | 192,102,531 | 478.82 | 104,613.99 | 108,842.00 |
| mmc4-core-ff | 9,425,497 | 9,332,329 | 15,732,454 | 59.98 | 5,579.99 | 9,659.09 |
| chinese-markdown| 168,323 | 167,989 | 101,026 | 1.34 | 755.08 | 14.69 |
| leetcode | 2,360 | 2,360 | 0 | 0.02 | 1.26 | 0.00 |
| linux-cn | 9,564 | 9,564 | 37,235 | 0.08 | 11.62 | 1.78 |
| DocLayNet | 68,084 | 68,084 | 89,170 | 0.19 | 25.25 | 1.60 |
| PIN-PMC | 5,354,766 | 70,197,548 | 18,527,901 | 208.44 | 45,923.75 | 2,154.95 |
| PIN-Arxiv | 736,497 | 11,356,952 | 4,128,143 | 38.83 | 5,553.01 | 507.99 |
| **PIN-200M** (total) | 194,977,272 | 269,050,145 | 230,718,460 | 800.31 | 163,848.85 | 121,182.10 |
Storage space statistics may have some error, so these values are for reference only.
The following parts may have unknown issues leading to **incomplete data**:
- PIN-PMC: 28, 37, 47
- PIN-Arxiv: 03, 23, 26, 28
- OBELICS: 00013, 00272, 00387, 00389, 00399, 00507, 00512, 00523, 00566, 00568, 00569, 00688, 00712
<img src="assets/distribution.png">
## 2 Data Structure
### 2.1 Subsets
We process 9 subsets, including PIN-Arxiv, PIN-PMC, DocLayNet, Linux-CN, chinese-markdown, OBELICS, MMC4, leetcode, and PG19.
<img src="assets/dataset-example.png">
Note: We do not release the PIN-arXiv subset in the preview version.
### 2.2 Folder Structure
The directory `content images` holds the images mentioned within the markdown text, and `overall images` display the overall visual representation of the markdown files. Moreover, the `JSONL` file encapsulate the textual content along with associated data details.
An example subset:
```
example_dataset/
│
├── content_image/
├── overall_image/
└── example_dataset.jsonl
```
A subset with multiple parts:
```
example_dataset/
│
├── part00/
│ ├── content_image/
│ ├── overall_image/
│ └── part00.jsonl
│
├── part01/
│ ├── content_image/
│ ├── overall_image/
│ └── part01.jsonl
│
... - More similar parts
```
### 2.3 content_image Folder
This folder contains all the content images used in the markdown files.
Note: All images need to be converted to PNG format. The filename should be unique within the folder.
```
content_image/
│
├── 1.png
├── 2.png
...
```
### 2.4 overall_image Folder
This folder contains all the overall images for each sample.
Note: All images need to be converted to PNG format. The filename should be unique within the folder.
```
overall_image/
│
├── 1.png
├── 2.png
...
```
#### 2.5 JSON Lines Format
we provide a detailed example of the annotations included with each data entry.
```
{
"id": 1919,
"meta": {
"language": "en",
"oi_exist": true,
"oi_source": "compiling",
"source_dataset": "example_source (e.g. OBELICS)",
"ori_meta": {
"document_url": "https://www.example.com/2022/02/21/example/",
...
}
},
"doc_id": 1997,
"page_id": 0,
"date_download": "2024-03-01"
},
"license": "CC-BY-4.0",
"quality_signals": {
"doc_length": 100,
...
},
"content_image": [
"content_image/1997-0.png",
"content_image/1997-1.png"
],
"md": "<img src='content_image/1997-0.png'>\n\nThis is a fake sample data line, just for show.\n\nThis is a fake sample data line, just for show.\n\n<img src='content_image/1997-1.png'>\n\nThis is a fake sample data line, just for show.",
"overall_image": "overall_image/1997.png"
}
```
**Field Descriptions:**
**Top-level Keys**
- `id` (number): A globally unique identifier for the data record.
- `meta` (object): A container for document-level metadata.
- `license` (string): The license associated with the sample (e.g., `CC-BY-4.0`).
- `quality_signals` (object): A collection of computed quality indicators for data filtering.
- `md` (string): The Markdown body text.
- `content_image` (string[]): An ordered list of content-level image paths referenced in the body.
- `overall_image` (string or string[]): Path(s) to page- or document-level overall image(s).
**Nested Keys in `meta`**
- `language` (string): The primary language of the document (e.g., en, zh).
- `oi_exist` (boolean): A flag indicating whether a document-level overall image exists.
- `oi_source` (string): The source of the overall image: `ori` (from original dataset) or `compiling` (programmatically generated).
- `source_dataset` (string): Identifies the data origin.
- `ori_meta` (object or null): A snapshot of the metadata from the original dataset.
- `doc_id` (number or string): A document-level identifier to group multiple pages/slices.
- `page_id` (number or null): The page index within a document.
- `date_download` (string, `YYYY-MM-DD`): The date when the source document was collected.
**Quality Signals**
The `quality_signals` field provides a set of computed metrics to allow for rapid data assessment and filtering.
- `image_text_interleaving_count` (ITIF): Measures the frequency of alternation between image and text modalities.
- `text_block_count` (TBC): The total number of text blocks in a sample.
- `total_token_count`: The total number of tokens in the entire Markdown file (tokenized with `meta-llama/Llama-3.2-1B`).
- `doc_length`: The total character count of the entire Markdown file.
- `avg_tokens_per_text_block`: The average number of tokens contained within each text block.
- `avg_text_block_length`: The average character length of each text block.
- **Markup statistics**:
- `bold_char_count`: Count of characters formatted as bold.
- `italic_char_count`: Count of characters formatted as italic.
- `title_count`: Count of title tags.
## 3 Examples of jsonl files
We selected samples consisting of short markdown documents.
### 3.1 An example of DocLynet
Notably, the dataset's overall images are converted from the original dataset's PDFs into PNG format.
```json
{
"id": 20336,
"meta": {
"language": "en",
"oi_exist": true,
"ori_meta": null,
"doc_id": "1407.2292.pdf",
"page_id": "50",
"date_download": "2024-3-24",
"oi_source": "ori",
"source_dataset": "DocLayNet"
},
"quality_signals": {
"image_text_interleaving_count": 7,
"text_block_count": 4,
"avg_text_block_length": 126.75,
"total_token_count": 158,
"bold_char_count": 0,
"italic_char_count": 3,
"title_count": 0,
"avg_tokens_per_text_block": 39.5,
"doc_length": 671
},
"license": "https://cdla.io/permissive-1-0/",
"content_image": [
"content_image/320672.jpg",
"content_image/320673.jpg",
"content_image/320676.jpg",
"content_image/320674.jpg"
],
"overall_image": "overall_image/98bd5c533929933027503b713e5d54698e74be70b6d1fc78db0bf1ed490259bd.png",
"md": "such a function to the sum rule is given by \n\n<img src='content_image/320672.jpg'>\n\nwhich is independent of the continuum threshold s$_{0}$ ..."
}
```
### 3.2 An example of OBELICS
```json
{
"id": 96121,
"meta": {
"language": "en",
"oi_exist": true,
"oi_source": "compiling",
"source_dataset": "OBELICS",
"ori_meta": {
"document_url": "https://mtavari.tv/en/news/47481-imprisoning-political-opponent-was-not-right",
"unformatted_src": "https://api.mtavari.tv/files/styles/news_thumb_lg/public/photos/2021/06/devidiiii.jpg?itok=3mauHu-2",
"src": "https://api.mtavari.tv/files/styles/news_thumb_lg/public/photos/2021/06/devidiiii.jpg?itok=3mauHu-2",
"formatted_filename": "devidiiii",
"original_width": 825,
"original_height": 464,
"format": "jpeg",
"general_meta": {
"url": "https://mtavari.tv/en/news/47481-imprisoning-political-opponent-was-not-right",
"warc_filename": "crawl-data/CC-MAIN-2021-31/segments/1627046154897.82/warc/CC-MAIN-20210804174229-20210804204229-00188.warc.gz",
"warc_record_offset": 376468578,
"warc_record_length": 33539
}
},
"doc_id": 53498,
"page_id": 0,
"date_download": "2024-09-01"
},
"md": "\n\n<img src='content_image/53498-0.png'>\n\nNika Melia, chairman of the United National Movement, continues to hold meetings in Washington...",
"license": "CC-BY-4.0",
"quality_signals": {
"image_text_interleaving_count": 2,
"text_block_count": 3,
"avg_text_block_length": 338.0,
"total_token_count": 219,
"bold_char_count": 0,
"italic_char_count": 1,
"title_count": 0,
"avg_tokens_per_text_block": 73.0,
"doc_length": 1100
},
"content_image": [
"content_image/53498-0.png",
"content_image/53498-2.png"
],
"overall_image": "overall_image/53498-0.png"
}}
```
### 3.3 An example of chinese-markdown
```json
{
"id": 93714,
"meta": {
"language": "zh",
"oi_exist": true,
"ori_meta": null,
"doc_id": 167547,
"page_id": null,
"date_download": "2024-04-30",
"oi_source": "compiling",
"source_dataset": "chinese-markdown"
},
"md": "# BannerView-Android\n\nAndroid Kotlin基于ViewPage2和ViewBinding的轻量级BannerView，功能全面，易定制扩展\n\n# Demo效果图\n\n<img src=\"content_image/167547-1.png\" />\n\n# 使用\n\n- 添加gradle依赖（version=[![](content_image/167547-0.png)](https://jitpack.io/#zicheng2019/banner-android)）\n\n ```groovy...",
"license": "Apache-2.0",
"quality_signals": {
"image_text_interleaving_count": 2,
"text_block_count": 20,
"avg_text_block_length": 452.4,
"total_token_count": 2662,
"bold_char_count": 0,
"italic_char_count": 12,
"title_count": 5,
"avg_tokens_per_text_block": 133.1,
"doc_length": 9169
},
"overall_image": "overall_image/167547.png",
"content_image": [
"content_image/167547-0.png",
"content_image/167547-1.png"
]
}
```
### 3.4 An example of leetcode
```json
{
"id": 723,
"meta": {
"language": "en",
"doc_id": 723,
"page_id": null,
"oi_exist": true,
"source_dataset": "leetcode",
"date_download": "2024-05-05",
"ori_meta": {
"slug": "candy-crush",
"difficulty": "Medium"
},
"oi_source": "compiling"
},
"quality_signals": {
"image_text_interleaving_count": 0,
"text_block_count": 23,
"avg_text_block_length": 354.95652173913044,
"total_token_count": 2493,
"bold_char_count": 7,
"italic_char_count": 5,
"title_count": 5,
"avg_tokens_per_text_block": 108.3913043478261,
"doc_length": 8237
},
"license": "MIT",
"content_image": null,
"md": "# Candy Crush\n\n- slug: candy-crush\n- difficulty: Medium\n\nThis question is about implementing a basic elimination algorithm for Candy Crush.\n\nGiven an `m x n` integer array `board` representing the grid of candy where `board[i][j]` represents the type of candy. A value of `board[i][j] == 0` represents that the cell is empty...",
"overall_image": "overall_image/723.png"
}
```
### 3.5 An example of linux-cn
```json
{
"id": 2976,
"meta": {
"language": "zh",
"doc_id": 8319,
"page_id": null,
"oi_exist": true,
"source_dataset": "linux-cn",
"date_download": "2024-05-06",
"ori_meta": {
"title": "如何在 Linux 中安装最新的 Python 3.6 版本",
"author": "Gabriel Cánepa",
"fromurl": "http://www.tecmint.com/install-python-in-linux/",
"summary": "在这儿，我将阐述如何使用源代码进行安装。这样我们能够获取语言的最新的稳定发行版（3.6），并且提供了一种和 Linux 版本无关的安装方法。",
"pic": "/data/attachment/album/201703/19/053959s707j3taptyzea3t.jpg.thumb.jpg",
"largepic": "/data/attachment/album/201703/19/053959s707j3taptyzea3t.jpg",
"titlepic": true,
"thumb": false,
"islctt": true,
"selector": "",
"translator": "ucasFL",
"reviewer": "",
"editorchoice": false,
"tags": [
"Python",
"Python3"
],
"category": "技术",
"count": {
"commentnum": 2,
"favtimes": 3,
"likes": 0,
"sharetimes": 0,
"viewnum": 22234
},
"comments_data": [
{
"date": "2017-03-19 20:20:39",
"message": "pacman _S python&nbsp; &nbsp; 既可以完成python3.6的安装，啦啦啦、、污啦啦....",
"postip": "183.206.12.222",
"username": "不着调的小男生 [Chrome 57.0|Windows 7]"
},
{
"date": "2017-03-22 04:26:09",
"message": "pyenv用户路过",
"postip": "164.107.165.226",
"username": "来自美国的 Firefox 45.0|Windows 7 用户"
}
],
"related": [],
"excerpt": "在这儿，我将阐述如何使用源代码进行安装。这样我们能够获取语言的最新的稳定发行版（3.6），并且提供了一种和 Linux 版本无关的安装方法。",
"date": "2017-03-19 10:32:00",
"updated": "2017-03-19 10:32:00",
"id": 8319,
"permalink": "/article-8319-1.html"
},
"oi_source": "compiling"
},
"quality_signals": {
"image_text_interleaving_count": 4,
"text_block_count": 44,
"avg_text_block_length": 48.29545454545455,
"total_token_count": 1071,
"bold_char_count": 0,
"italic_char_count": 5,
"title_count": 17,
"avg_tokens_per_text_block": 24.34090909090909,
"doc_length": 2450
},
"license": "CC-BY-NC-4.0",
"content_image": [
"content_image/album_201703_19_053959s707j3taptyzea3t.jpg",
"content_image/album_201703_19_054046z2kzvkwhdxnp9b32.png"
],
"md": "# 如何在 Linux 中安装最新的 Python 3.6 版本\n\n在这篇文章中，我将展示如何在 CentOS/RHEL 7、Debian 以及它的衍生版本比如 Ubuntu（最新的 Ubuntu 16.04 LTS 版本已经安装了最新的 Python 版本）或 Linux Mint 上安装和使用 Python 3.x 。我们的重点是安装可用于命令行的核心语言工具。\n\n然后，我们也会阐述如何安装 Python IDLE － 一个基于 GUI 的工具，它允许我们运行 Python 代码和创建独立函数。\n\n<img src=\"content_image/album_201703_19_053959s707j3taptyzea3t.jpg\" alt=\"\" title=\"\">...",
"overall_image": "overall_image/8319.png"
}
```
### 3.6 An example of mmc-core-ff
```json
{
"id": 128563,
"meta": {
"language": "en",
"oi_exist": true,
"doc_id": 73966,
"page_id": 1,
"source_jsonl": "mmc4-core-ff/docs_no_face_shard_10278_v3.jsonl",
"ori_meta": {
"url": "https://candysdirt.com/tag/professional-real-estate-photography/",
"text_list": [
"Our Tuesday Two Hundred takes us to the lush Lochwood neighborhood of East Dallas.",
"The home at 11324 Carissa Dr. is near greenbelts, wooded areas, hiking trails, and creeks, and the area has mature trees and is close to White Rock Lake.",
"This is a 3-2 with 1,300 square feet, built in 1962.",
"The interior offers an open floorplan, vaulted ceilings, and Midcentury style.",
"The large and open backyard is lovely with shade trees and a relaxing, new patio.",
"It was listed March 4 by John Prell with Creekview Realty for $249,000.",
"Let’s peek inside before it goes under contract!",
"Realtors Can Now Upload 36 Photos to MLS, But Do You Need Those Extra 11 Shots?",
"Some buyers have the hardest time imagining how a home will work for their families, and in the digital age, Realtors are providing tools like virtual open houses and tours to help these buyers fall in love with their listings.",
"So I wasn’t surprised when MetroTex Association of Realtors increased the number of photos agents can upload for each listing from 25 to 36.",
"Of course, some Realtors and real estate professionals see this as a fantastic development.",
"Others?",
"Not so much.",
"This month, there’s a story that’s gone viral by Orlando professional real estate photographer Harry Lim.",
"His post, After Nearly 8 Months, Photos Help Sell Home in 8 Days, is remarkable because it shows the unmistakable difference professional photography makes in the marketing of a house (his before-and-after photos above make that clear!).",
"Here at CandysDirt, we could not agree more!",
"In 2013, contributing writer Karen Eubank wrote an interesting series for us, Outside the Frame, which looked at the importance of professional real estate photography and offered the insights of Dallas’ leading real estate photographers (see parts one, two, three, and four).",
"Today’s real estate buyer is doing their home searches on the Internet, with 90 percent searching online and 89 percent using a mobile search engine.",
"What they find first in their searches are photographs, and the quality of those can make or break a listing.",
"So today, we’ve made a list of five ways professional real estate photographers bring value to the marketing of a house.",
"Here at Candy’s Dirt, we could not agree more!",
"In 2013, contributing writer Karen Eubank wrote an interesting series for us, Outside the Frame, which looked at the importance of professional real estate photography and offered the insights of Dallas’ leading real estate photographers (see parts one, two, three, and four).",
"In the average MLS listing, you’ve got around 25 shots to “sell” a buyer on your property.",
"Poorly planned or executed photography might result in redundant shots (multiple angles of one room where no new visual information is presented), unnecessary photos (you don’t need a photo of the toilet—people know it’s in the bathroom), or just plain bad photos that make the property look boring at best.",
"A professional real estate photographer understands which shots matter and why, and can make judgment calls based on the features of each individual property.",
"“Normally I don’t waste a photo on a laundry room, but a condo unit that has a washer/dryer when the neighboring ones do not might be of great importance compared to the average property,” said Lance Selgo of Unique Exposure Photography.",
"I’m sure you’ve already seen the photo about this poor guy who was just trying to sell his Austin condo and ended up flashing everyone in the process.",
"As it turns out, it was all an accident, says KDFW’s report on the … ahem … matter."
],
"image_info": [
{
"face_detections": null,
"image_id": "73966-0.png",
"image_name": "dcff7634dbf5.jpg",
"matched_sim": 0.2782543898,
"matched_text_index": 2,
"raw_url": "https://candysdirt.com/wp-content/uploads/2016/03/11324-Carissa-a.jpg"
},
{
"face_detections": null,
"image_id": "73966-1.png",
"image_name": "cca863bcf5c4.jpg",
"matched_sim": 0.2481627911,
"matched_text_index": 13,
"raw_url": "https://candysdirt.com/wp-content/uploads/2015/03/Harry-Lim-photography-after-1.jpg"
},
{
"face_detections": null,
"image_id": "73966-2.png",
"image_name": "4ce6f93e1024.jpg",
"matched_sim": 0.2550932765,
"matched_text_index": 14,
"raw_url": "https://candysdirt.com/wp-content/uploads/2015/03/Harry-Lim-photography-before.jpg"
},
{
"face_detections": null,
"image_id": "73966-3.png",
"image_name": "67305f885cd6.jpeg",
"matched_sim": 0.2711390257,
"matched_text_index": 21,
"raw_url": "https://candysdirt.com/wp-content/uploads/2016/02/Woodland-Dr.-Hip-Pocket-1024x768.jpeg"
}
],
"similarity_matrix": [
[
0.2679546475,
0.2456822842,
0.2782543898,
0.2697871029,
0.2253711671,
0.2187553942,
0.2450769991,
0.2209598422,
0.2134867758,
0.2523652911,
0.2162774056,
0.1309889108,
0.1222634017,
0.1528567523,
0.2498115599,
0.1173600107,
0.2460378408,
0.2086704075,
0.2005334944,
0.220557496,
0.1138958782,
0.2460378408,
0.215545997,
0.2039846778,
0.2243762016,
0.1641701609,
0.1935409009,
0.1720048189
],
[
0.193737179,
0.1892189533,
0.1827486455,
0.1972641498,
0.1904278398,
0.175565362,
0.2155833542,
0.2229709327,
0.2141839713,
0.1824499816,
0.1967996359,
0.1211651117,
0.1148867309,
0.2481627911,
0.2460642755,
0.1077519804,
0.2172371447,
0.1969793439,
0.2067773938,
0.2313994616,
0.091579631,
0.2172371447,
0.2276338935,
0.2117997408,
0.2445380539,
0.1727049649,
0.1547327042,
0.0757467374
],
[
0.1592763513,
0.1592113674,
0.1506833732,
0.1602986455,
0.1904094666,
0.1619938463,
0.1853862107,
0.2133037448,
0.2037098408,
0.1507958323,
0.2089106143,
0.1104154736,
0.1209291145,
0.2121700943,
0.2550932765,
0.1198005825,
0.1990599185,
0.1816542447,
0.2058906555,
0.216880396,
0.1026834622,
0.1990599185,
0.1970586479,
0.2115100622,
0.216872409,
0.1638485491,
0.1217981502,
0.0824729875
],
[
0.2534551024,
0.2093835026,
0.1617470533,
0.2067761421,
0.1927422285,
0.1615802497,
0.1912238896,
0.2168802172,
0.2222865969,
0.2208270431,
0.2046597004,
0.1515626013,
0.1259405166,
0.1620695591,
0.2162202597,
0.1032845974,
0.2711390257,
0.2194584012,
0.2196688503,
0.2180148959,
0.1074701771,
0.2711390257,
0.2075627744,
0.1849380136,
0.2361781299,
0.1623896956,
0.1758471876,
0.1574440598
]
],
"could_have_url_duplicate": 0
},
"date_download": "2024-05-12",
"oi_source": "compiling",
"source_dataset": "mmc4-core-ff"
},
"md": "\n\n<img src='content_image/73966-2.png'>\n\nHere at CandysDirt, we could not agree more! In 2013, contributing writer Karen Eubank wrote an interesting series for us, Outside the Frame, ...",
"license": "ODC-BY",
"quality_signals": {
"image_text_interleaving_count": 3,
"text_block_count": 2,
"avg_text_block_length": 1028.0,
"total_token_count": 451,
"bold_char_count": 0,
"italic_char_count": 1,
"title_count": 0,
"avg_tokens_per_text_block": 225.5,
"doc_length": 2140
},
"content_image": [
"content_image/73966-2.png",
"content_image/73966-3.png"
],
"overall_image": "overall_image/73966-1.png"
}
```
### 3.7 An example of PG19
```json
{
"id": 19185,
"meta": {
"language": "en",
"oi_exist": true,
"doc_id": 873,
"page_id": 45,
"split": "train",
"ori_meta": {
"url": "http://www.gutenberg.org/ebooks/18659",
"short_book_title": "The Seventh Manchesters by S. J. Wilson",
"publication_date": 1865
},
"date_download": "2024-05-10",
"oi_source": "compiling",
"source_dataset": "pg19"
},
"md": "The pre-arranged plan was carried out, and after the 127th brigade had made good the high ground east of Thill oy, in face of some opposition, the East Lancs. came through and took up the advance on what had now become a one brigade front. They had not gone far before they encountered the enemy in strength holding Riencourt, and they promptly attacked it...",
"license": "Apache 2.0",
"quality_signals": {
"image_text_interleaving_count": 0,
"text_block_count": 5,
"avg_text_block_length": 1233.4,
"total_token_count": 1409,
"bold_char_count": 0,
"italic_char_count": 0,
"title_count": 0,
"avg_tokens_per_text_block": 281.8,
"doc_length": 6183
},
"content_image": null,
"overall_image": "overall_image/873-45.png"
}
```
### 3.8 An example of PIN-PMC
```json
{
"id": "PMC9832689",
"meta": {
"date_download": "2024-12-14",
"language": "en",
"source": "PMC"
},
"quality_signals": {
"image_text_interleaving_count": 8,
"text_block_count": 95,
"avg_text_block_length": 466.6,
"total_token_count": 13943,
"bold_char_count": 2,
"italic_char_count": 4,
"title_count": 19,
"avg_tokens_per_text_block": 146.76842105263157,
"doc_length": 48469
},
"content_image": [
"content_image/PMC9832689/gr1_lrg.jpg",
"content_image/PMC9832689/gr2_lrg.jpg",
"content_image/PMC9832689/gr3_lrg.jpg",
"content_image/PMC9832689/gr4_lrg.jpg"
],
"md": "# Time series analysis revealed prognostic value of continuous nasopharyngeal SARS-CoV-2 nucleic acid quantification for COVID-19: A retrospective study of >3000 COVID-19 patients from 2 centers\n\n## Abstract\n\nEarly stratification of disease progression remains one of the major challenges towards the post-coronavirus disease 2019 (COVID-19) era...",
"overall_image": [
"overall_image/PMC9832689/p1-main.png",
"overall_image/PMC9832689/p2-main.png",
"overall_image/PMC9832689/p3-main.png",
"overall_image/PMC9832689/p4-main.png",
"overall_image/PMC9832689/p5-main.png",
"overall_image/PMC9832689/p6-main.png",
"overall_image/PMC9832689/p7-main.png",
"overall_image/PMC9832689/p8-main.png",
"overall_image/PMC9832689/p9-main.png",
"overall_image/PMC9832689/p10-main.png",
"overall_image/PMC9832689/p11-main.png",
"overall_image/PMC9832689/p12-main.png"
]
}
```
### 3.8 An example of PIN-Arxiv
```json
{
"id": "1410.4821",
"meta": {
"language": "en",
"source": "Arxiv",
"date_download": "2024-12-03"
},
"quality_signals": {
"image_text_interleaving_count": 0,
"text_block_count": 167,
"avg_text_block_length": 275.89820359281435,
"total_token_count": 12569,
"bold_char_count": 1,
"italic_char_count": 144,
"title_count": 30,
"avg_tokens_per_text_block": 75.26347305389221,
"doc_length": 46418
},
"content_image": [],
"md": "# Convex Optimization in Julia\n\n Madeleine Udell\n\n Karanveer Mohan\n\n David Zeng\n\[email protected]\[email protected]\[email protected]\n\nJenny Hong\n\n Steven Diamond\n\n Stephen Boyd\n\[email protected]\[email protected]\[email protected]\n\n###### Abstract\n\nThis paper describes Convex...",
"overall_image": [
"overall_image/1410.4821/p1-1410.4821v1.png",
"overall_image/1410.4821/p2-1410.4821v1.png",
"overall_image/1410.4821/p3-1410.4821v1.png",
"overall_image/1410.4821/p4-1410.4821v1.png",
"overall_image/1410.4821/p5-1410.4821v1.png",
"overall_image/1410.4821/p6-1410.4821v1.png",
"overall_image/1410.4821/p7-1410.4821v1.png",
"overall_image/1410.4821/p8-1410.4821v1.png",
"overall_image/1410.4821/p9-1410.4821v1.png",
"overall_image/1410.4821/p10-1410.4821v1.png",
"overall_image/1410.4821/p11-1410.4821v1.png"
]
}
```
## 4 License
For data generated or produced by us, please adhere to the Apache 2.0 License.
For data sourced from third parties, compliance with the respective third-party licenses is required.
## Citation
```
@article{DBLP:journals/corr/abs-2406-13923,
author = {Junjie Wang and
Yuxiang Zhang and
Minghao Liu and
Yin Zhang and
Yatai Ji and
Weihao Xuan and
Nie Lin and
Kang Zhu and
Zhiqiang Lin and
Yiming Ren and
Chunyang Jiang and
Yiyao Yu and
Zekun Wang and
Tiezhen Wang and
Wenhao Huang and
Jie Fu and
Qunshu Lin and
Yujiu Yang and
Ge Zhang and
Ruibin Yuan and
Bei Chen and
Wenhu Chen},
title = {{PIN:} {A} Knowledge-Intensive Dataset for Paired and Interleaved
Multimodal Documents},
journal = {CoRR},
volume = {abs/2406.13923},
year = {2024}
}
```