id
int64 0
19
| arxiv_id
stringlengths 11
12
| page
int64 1
234
| bounding_box
sequencelengths 4
4
| latex_content
stringlengths 217
28.9k
| extracted_content
sequencelengths 1
85
| similarity_score
float64 0.36
1
| table_image
unknown | page_image
unknown |
|---|---|---|---|---|---|---|---|---|
0 | 1405.2377v1 | 4 | [
99.43800354003906,
121.62852478027344,
515.7630004882812,
146.0369873046875
] | "\\begin{table}[H]\n \\centering\n \\scriptsize\n \\begin{tabular}{|p{3cm}|p{3cm}|p{4cm}|p{(...TRUNCATED) | [["Domain","Raw Features","Response","Data Set Cardinality"],["Australian Credit Scor-\ning","16","D(...TRUNCATED) | 0.462025 | "iVBORw0KGgoAAAANSUhEUgAABoMAAABjCAIAAADfDu7wAAB2bUlEQVR4nO29d1wUyfP43QssOUdBkJxBFBFBESOCOaCIWVHP7GH(...TRUNCATED) | "iVBORw0KGgoAAAANSUhEUgAACU4AAA0oCAIAAADF/wRIAAEAAElEQVR4nOzdZ3wVZeI/7AkQOoQqRXqXIiBIUYqKCNgVxc4q9sK(...TRUNCATED) |
1 | 1405.2377v1 | 5 | [
121.80699920654297,
121.62850952148438,
493.3940124511719,
170.34600830078125
] | "\\begin{table}[H]\n \\centering\n \\scriptsize\n \\begin{tabular}{|p{3cm}|p{2cm}|p{2cm}|p{(...TRUNCATED) | [["Statistic","Average","Minimum","Maximum","Standard Devi-\nation"],["Predictive Accuracy of\nRando(...TRUNCATED) | 0.383961 | "iVBORw0KGgoAAAANSUhEUgAABc8AAADECAIAAADGc06JAACeRklEQVR4nOzdZ1wUV/s4/Ivee++9gyCoqIBdsXexYdcIsWOP3ia(...TRUNCATED) | "iVBORw0KGgoAAAANSUhEUgAACU4AAA0oCAIAAADF/wRIAAEAAElEQVR4nOzdZ5wUVcI37EPOOaPknFFQQRHEhDmLOaGumLOurht(...TRUNCATED) |
0 | 2207.05295v2 | 6 | [
51.43559646606445,
106.33673095703125,
297.42755126953125,
170.03900146484375
] | "\\begin{table}[t]\n\\centering\n\\resizebox{\\columnwidth}{!}{\n\\begin{tabular}{c|c|c|c|c|c}\n (...TRUNCATED) | [["Data","Number of rows","TabSynDex Score",null,null,null],[null,null,"% of real data treated as sy(...TRUNCATED) | 0.545251 | "iVBORw0KGgoAAAANSUhEUgAAA9kAAAEACAIAAAAoeCJzAADYoElEQVR4nOydeSAV3f/4z7WTvUhEm4hSkiXJ4yGVR6seWpBKWrR(...TRUNCATED) | "iVBORw0KGgoAAAANSUhEUgAACZAAAAxgCAIAAADBWzgHAAEAAElEQVR4nOzdd2AU1b7A8TNbsumF9NAChI7SQelI7yAgimChCoi(...TRUNCATED) |
0 | 1911.00623v2 | 10 | [
45.827999114990234,
180.93701171875,
443.06201171875,
283.5989990234375
] | "\\begin{table}[htbp]\n% \\caption{Comparison of traditional machine learning algorithms. Notation: (...TRUNCATED) | [["Algorithm","Model size","Optimization","Training complexity","Inference complexity"],["Decision t(...TRUNCATED) | 0.715429 | "iVBORw0KGgoAAAANSUhEUgAABjYAAAGcCAIAAAD1Yg71AAEAAElEQVR4nOy9dVzU2ff4f2aIIQQURETBwMZYu0DsxK61de1YY+3(...TRUNCATED) | "iVBORw0KGgoAAAANSUhEUgAAB5gAAAtACAIAAABC8WhNAAEAAElEQVR4nOz9Z3wVxf/4f086IaH3XqX3Jk26oEgHQaQXwQYWEAX(...TRUNCATED) |
1 | 1911.00623v2 | 13 | [
50.15800094604492,
122.1409912109375,
434.0989990234375,
336.260009765625
] | "\\begin{table}[htbp]\n% \\caption{DNN resource requirements modeling. ASIC: Application-Specific In(...TRUNCATED) | [["Work","Platform","Framework","Metric","Measured\nfeatures","Regression\nmodel","Relative\nerror"](...TRUNCATED) | 0.76588 | "iVBORw0KGgoAAAANSUhEUgAABgEAAANaCAIAAAAvY3s0AAEAAElEQVR4nOydd1xUR/fwzy5L7yggUlQERKLYokGw9941do0VrLG(...TRUNCATED) | "iVBORw0KGgoAAAANSUhEUgAAB5gAAAtACAIAAABC8WhNAAEAAElEQVR4nOzdd4DUxP/4/7leaUcVODrSO4IIghRBpChSBKUqoCI(...TRUNCATED) |
2 | 1911.00623v2 | 11 | [
52.99300003051758,
111.1820068359375,
431.26300048828125,
264.3190002441406
] | "\\begin{table}[htbp]\n\\caption{Comparison of popular CNNs.}\n\\centering\n\\scriptsize\n\\begin{ta(...TRUNCATED) | [["Metric","AlexNet\n[96]","VGG-16\n[157]","GoogLeNet\n[167]","ResNet-18\n[69]","ResNet-50\n[69]","I(...TRUNCATED) | 0.389231 | "iVBORw0KGgoAAAANSUhEUgAABesAAAJmCAIAAABDjkzvAAEAAElEQVR4nOzdZ1wU1/cw8LOV3gQERYoi0rEgqKDYsMQWjWCPGo2(...TRUNCATED) | "iVBORw0KGgoAAAANSUhEUgAAB5gAAAtACAIAAABC8WhNAAEAAElEQVR4nOzdd3wUxf/48bn0hBSSAAFC6L0KSBXpIkV6UXpHUUF(...TRUNCATED) |
3 | 1911.00623v2 | 16 | [
49.22999954223633,
111.18197631835938,
435.0270080566406,
272.28900146484375
] | "\\begin{table}[tbp]\n\\caption{Comparison of lightweight CNNs.}\n\\centering\n\\scriptsize\n\\begin(...TRUNCATED) | [["Metric","MobileNet\nV1-1.0[79]","MobileNet\nV2-1.0[152]","Squeeze-\nNet[85]","Squeeze-\nNext-1.0-(...TRUNCATED) | 0.364123 | "iVBORw0KGgoAAAANSUhEUgAABgkAAAKGCAIAAAD/P+wgAAEAAElEQVR4nOydd1xTSffwTxIIEDqIoBRRREERV1AUULGt+th77x1(...TRUNCATED) | "iVBORw0KGgoAAAANSUhEUgAAB5gAAAtACAIAAABC8WhNAAEAAElEQVR4nOzdV3wU1f/4/5OE9EAglNBLRHpHojSpIiAdREGagoA(...TRUNCATED) |
4 | 1911.00623v2 | 19 | [
45.827999114990234,
119.05322647094727,
450.36199951171875,
583.1997528076172
] | "\\begin{table}\n\\begin{threeparttable}[tb]\n\\centering\n\\tabcolsep=0.04cm\n\\caption{The chronol(...TRUNCATED) | [["Year","Approach","Keywords","Quantization1 Benchmark",null],[null,null,null,"Parameter\nForward B(...TRUNCATED) | 0.545965 | "iVBORw0KGgoAAAANSUhEUgAABlMAAAdBCAIAAAAP3PsKAAEAAElEQVR4nOzdZ1wU1/cw8LOVKr0KCCKgIKJgB9SIXWI0sURj72K(...TRUNCATED) | "iVBORw0KGgoAAAANSUhEUgAAB5gAAAtACAIAAABC8WhNAAEAAElEQVR4nOz9eaAP5f/4/19ndxYc27EcS3ZCWVpsKWshFFFCEiE(...TRUNCATED) |
0 | 1508.05417v2 | 9 | [
315.7550048828125,
580.4850158691406,
556.8099975585938,
746.3389892578125
] | "\\begin{table}[!b]\\scriptsize\n\\centering\n\\caption{Simulation Parameters}\n\\begin{tabular}{ l (...TRUNCATED) | [["Size of active region (W × L)","0.1 × 5 (µm)"],["Temperature (T )","298 (K)"],["Relative permi(...TRUNCATED) | 0.817975 | "iVBORw0KGgoAAAANSUhEUgAAA8UAAAKZCAIAAAAeax4DAAEAAElEQVR4nOy9d1wTy/f4fULviIAUEaUogkoRFLBd28WCiqCC2Cs(...TRUNCATED) | "iVBORw0KGgoAAAANSUhEUgAACZAAAAxgCAIAAADBWzgHAAEAAElEQVR4nOzdZXwUV/vw8RPbGAkhQpAESHANnuJO0RaXQnEtRUr(...TRUNCATED) |
0 | 2105.00030v1 | 4 | [
312.1571014404297,
78.635986328125,
562.8569763183593,
342.24798583984375
] | "\\begin{table}[t]\n\\caption{Summary of curatorial actions in annotation schema}\n\\centering\n\\be(...TRUNCATED) | [["Curatorial Action","Examples"],["Initial review and planning","Look at deposited files, determine(...TRUNCATED) | 0.985605 | "iVBORw0KGgoAAAANSUhEUgAAA+wAAAQfCAIAAAAGGvTmAAEAAElEQVR4nOyddUBUTdfAh0aQEFFCBQNRRCwUDMDADlQsQEUFscV(...TRUNCATED) | "iVBORw0KGgoAAAANSUhEUgAACZAAAAxgCAIAAADBWzgHAAEAAElEQVR4nOz9eZQk53nf+T7P+0ZErrV19b6vaCyNrbERO0mQlEi(...TRUNCATED) |
End of preview. Expand
in Dataset Viewer.
Arxiv-tables Dataset
Dataset Summary
The Arxiv-tables dataset is a collection of tables extracted from scientific papers published on arXiv, primarily focused on ML papers. It includes both the LaTeX source of the tables and their corresponding rendered images from the PDF versions of the papers.
Supported Tasks
This dataset can support several tasks, including but not limited to:
- Table structure recognition
- LaTeX to image generation for tables
- Image-to-LaTeX conversion for tables
- Table information retrieval
Dataset Structure
The dataset is organized as follows:
arxiv_tables/
├── images/
│ └── [arxiv_id]/
│ ├── table_0.png
│ ├── table_1.png
│ └── ...
├── pages/
│ └── [arxiv_id]/
│ ├── page_0.png
│ ├── page_1.png
│ └── ...
└── gt.json
Each arXiv paper has its own directory containing:
- Individual table images
- Full page images
- A ground truth JSON file with table information
Data Instances
Each instance in the dataset corresponds to a table and includes:
- Table image
- Full page image containing the table
- LaTeX source of the table
- Extracted content from the PDF
- Bounding box coordinates
- Page number
- Similarity score between LaTeX and extracted content
Data Fields
The gt.json file contains the following fields for each table:
id: Unique identifier for the tablepage: Page number where the table appearsbounding_box: Coordinates of the table in the PDFlatex_content: Original LaTeX code of the tableextracted_content: Content extracted from the PDF versionsimilarity_score: Similarity between LaTeX and extracted contenttable_image: Path to the cropped table imagepage_image: Path to the full page image
Dataset Creation
Curation Rationale
This dataset was created to provide a large-scale resource for researchers working on table understanding in scientific documents.
Source Data
The source data comes from arXiv, a free distribution service and open-access archive for scholarly articles. The dataset focuses on papers in the biology domain.
Annotations
The dataset does not contain manual annotations. The ground truth is automatically generated by matching LaTeX source code to the rendered PDF content.
Other Known Limitations
- The matching between LaTeX source and PDF content is based on similarity scores and may not be perfect.
- Some tables may be missed or incorrectly extracted due to limitations in the extraction process.
- The dataset is limited to papers that have both LaTeX source and PDF versions available on arXiv.
- Downloads last month
- 2