Datasets:

Dataset Viewer (First 5GB)
Auto-converted to Parquet Duplicate
id
int32
206
400k
title
stringlengths
3
96
text
stringlengths
100
4.76k
url
stringlengths
42
44
wiki_id
int32
337k
19.7M
paragraph_id
int32
0
271
fragment
imagewidth (px)
1.65k
1.65k
full_embeddings
listlengths
70
778
pooled_embeddings
listlengths
9
9
images
dict
206
Dayton Speedway
Dayton Speedway was a <a href="race%20track">race track</a> in <a href="Dayton%2C%20Ohio">Dayton</a>, <a href="Ohio">Ohio</a>, <a href="United%20States">United States</a>.
https://en.wikipedia.org/wiki?curid=19707701
19,707,701
0
[ [ 0.020294189453125, -0.08544921875, -0.0714111328125, 0.120849609375, -0.07501220703125, -0.04150390625, 0.026214599609375, 0.1864013671875, -0.01148223876953125, 0.00759124755859375, -0.11846923828125, 0.054168701171875, -0.0271759033203125, 0.01863098144531...
[ [ -0.01690673828125, -0.08587646484375, -0.0267181396484375, 0.0894775390625, 0.08233642578125, 0.00311279296875, -0.0159759521484375, 0.0511474609375, 0.01544189453125, 0.0034503936767578125, -0.041046142578125, 0.06683349609375, 0.04486083984375, -0.00346755...
{ "author": [], "caption": [], "credit": [], "image": [], "image_url": [], "license": [], "license_url": [], "source_url": [], "type": [], "url": [] }
207
Dayton Speedway
"The track held events for <a href=\"NASCAR\">NASCAR</a>, AAA, MARC, <a href=\"Automobile%20Racing%2(...TRUNCATED)
https://en.wikipedia.org/wiki?curid=19707701
19,707,701
1
[[0.020294189453125,-0.08544921875,-0.0714111328125,0.120849609375,-0.07501220703125,-0.04150390625,(...TRUNCATED)
[[-0.07244873046875,-0.09552001953125,0.01409912109375,0.126953125,0.07415771484375,-0.0507202148437(...TRUNCATED)
{"author":[],"caption":[],"credit":[],"image":[],"image_url":[],"license":[],"license_url":[],"sourc(...TRUNCATED)
208
Dayton Speedway
"The track was opened in June 1934 as a flat \"D shaped\" 5/8 mile dirt track. The original track wa(...TRUNCATED)
https://en.wikipedia.org/wiki?curid=19707701
19,707,701
2
[[0.020294189453125,-0.08544921875,-0.0714111328125,0.120849609375,-0.07501220703125,-0.04150390625,(...TRUNCATED)
[[-0.032958984375,-0.08087158203125,-0.007381439208984375,0.07305908203125,0.071533203125,-0.0163726(...TRUNCATED)
{"author":[],"caption":[],"credit":[],"image":[],"image_url":[],"license":[],"license_url":[],"sourc(...TRUNCATED)
240
Front Page Detective
"Front Page Detective is an American crime drama series which aired on the <a href=\"DuMont%20Televi(...TRUNCATED)
https://en.wikipedia.org/wiki?curid=19708097
19,708,097
0
[[0.020294189453125,-0.08544921875,-0.0714111328125,0.120849609375,-0.07501220703125,-0.04150390625,(...TRUNCATED)
[[-0.050506591796875,-0.1285400390625,-0.01580810546875,0.099365234375,0.10986328125,0.0047645568847(...TRUNCATED)
{"author":[],"caption":[],"credit":[],"image":[],"image_url":[],"license":[],"license_url":[],"sourc(...TRUNCATED)
241
Front Page Detective
"\"Front Page Detective\" stars <a href=\"Edmund%20Lowe\">Edmund Lowe</a> as David Chase, a newspape(...TRUNCATED)
https://en.wikipedia.org/wiki?curid=19708097
19,708,097
1
[[0.020294189453125,-0.08544921875,-0.0714111328125,0.120849609375,-0.07501220703125,-0.04150390625,(...TRUNCATED)
[[-0.05029296875,-0.110107421875,-0.043548583984375,0.0611572265625,0.1483154296875,0.02713012695312(...TRUNCATED)
{"author":[],"caption":[],"credit":[],"image":[],"image_url":[],"license":[],"license_url":[],"sourc(...TRUNCATED)
242
Front Page Detective
"Other cast members were <a href=\"Frank%20Jenks\">Frank Jenks</a> as Lieutenant Rodney, <a href=\"P(...TRUNCATED)
https://en.wikipedia.org/wiki?curid=19708097
19,708,097
2
[[0.020294189453125,-0.08544921875,-0.0714111328125,0.120849609375,-0.07501220703125,-0.04150390625,(...TRUNCATED)
[[-0.032684326171875,-0.08306884765625,0.00421905517578125,0.080810546875,0.169677734375,0.024963378(...TRUNCATED)
{"author":[],"caption":[],"credit":[],"image":[],"image_url":[],"license":[],"license_url":[],"sourc(...TRUNCATED)
243
Front Page Detective
"<a href=\"Jerry%20Fairbanks\">Jerry Fairbanks</a> was the producer and distributor, and Arnold West(...TRUNCATED)
https://en.wikipedia.org/wiki?curid=19708097
19,708,097
3
[[0.020294189453125,-0.08544921875,-0.0714111328125,0.120849609375,-0.07501220703125,-0.04150390625,(...TRUNCATED)
[[-0.041961669921875,-0.08935546875,-0.0164642333984375,0.06329345703125,0.12127685546875,0.01425933(...TRUNCATED)
{"author":[],"caption":[],"credit":[],"image":[],"image_url":[],"license":[],"license_url":[],"sourc(...TRUNCATED)
244
Front Page Detective
"<a href=\"UCLA%20Film%20and%20Television%20Archive\">UCLA Film and Television Archive</a> has 17 ep(...TRUNCATED)
https://en.wikipedia.org/wiki?curid=19708097
19,708,097
4
[[0.020294189453125,-0.08544921875,-0.0714111328125,0.120849609375,-0.07501220703125,-0.04150390625,(...TRUNCATED)
[[-0.038787841796875,-0.11590576171875,-0.0019378662109375,0.1456298828125,0.09124755859375,0.042999(...TRUNCATED)
{"author":[],"caption":[],"credit":[],"image":[],"image_url":[],"license":[],"license_url":[],"sourc(...TRUNCATED)
245
Front Page Detective
"Unlike many other programs which aired on DuMont, the series was produced on film by an outside pro(...TRUNCATED)
https://en.wikipedia.org/wiki?curid=19708097
19,708,097
5
[[0.020294189453125,-0.08544921875,-0.0714111328125,0.120849609375,-0.07501220703125,-0.04150390625,(...TRUNCATED)
[[-0.0523681640625,-0.1055908203125,-0.020416259765625,0.1138916015625,0.1068115234375,0.01455688476(...TRUNCATED)
{"author":[],"caption":[],"credit":[],"image":[],"image_url":[],"license":[],"license_url":[],"sourc(...TRUNCATED)
246
Front Page Detective
"The trade publication \"<a href=\"Variety%20%28magazine%29\">Variety</a>\" described one episode of(...TRUNCATED)
https://en.wikipedia.org/wiki?curid=19708097
19,708,097
6
[[0.020294189453125,-0.08544921875,-0.0714111328125,0.120849609375,-0.07501220703125,-0.04150390625,(...TRUNCATED)
[[-0.00556182861328125,-0.1279296875,-0.056671142578125,0.07781982421875,0.0704345703125,0.008583068(...TRUNCATED)
{"author":[],"caption":[],"credit":[],"image":[],"image_url":[],"license":[],"license_url":[],"sourc(...TRUNCATED)
End of preview. Expand in Data Studio

WikiFragments - Visual Arts Pages with Fragments (WikiFragmentsVA)

WikiFragmentsVA is a domain-specific multimodal dataset focused on the visual arts, derived from Wikipedia (en). It consists of textual paragraphs paired with related images (infoboxes and thumbnails), rendered as unified visual fragments. This dataset extends the base WikiFragments project by providing pre-rendered fragment images and multi-vector embeddings obtained via ColQwen2 v1.0, including optimized pooled representations for efficient retrieval.

image/png Example of a rendered fragment with multiple images and captions.

Dataset Details

Dataset Description

WikiFragmentsVA is a specialized subset of the WikiFragments dataset, curated to cover the Visual Arts domain. To construct this dataset, we recursively navigated Wikipedia categories starting from "Category:Visual arts" and descending up to 5 depth levels.

A multimodal fragment is defined as an atomic knowledge unit consisting of a paragraph from a Wikipedia page and all images that, in the page’s source code, appear above that paragraph. For this dataset, each fragment is rendered into a single image resembling a document layout (images/captions in a grid at the top, paragraph at the bottom) and encoded into multi-vector representations using ColQwen2.

  • Curated by: Nicola Fanelli (PhD Student @ University of Bari Aldo Moro, Italy)
  • Language(s) (NLP): English

License

  • Code: MIT License.
  • Text Data: The Wikipedia text is licensed under CC BY-SA 4.0. When using this dataset, you must provide proper attribution to Wikipedia and its contributors and share any derivatives under the same license.
  • Images: Images are sourced from Wikipedia and Wikimedia Commons. Each image is subject to its own license, which is typically indicated on its original page. Users of this dataset are responsible for ensuring they comply with the licensing terms of individual images. For ease of use, we provide the license and attribution information for each image in the dataset, along with the corresponding URLs to download them at the resolution available on Wikipedia.

Dataset Sources

Uses

Direct Use

This dataset is designed for multimodal retrieval-augmented generation (RAG) in the visual arts domain. It supports:

  • Two-stage retrieval: Using optimized pooled embeddings for fast initial filtering and full multi-vector embeddings for late-interaction re-ranking.
  • Multimodal grounding: Providing rendered visual context to MLLMs for answering complex questions about art history, styles, and artists.
  • Visual Document Retrieval: Evaluating models on their ability to retrieve documents based on visual and textual alignment.

Out-of-Scope Use

  • Real-time systems (the dataset is a static snapshot).
  • Commercial use without verifying individual image licenses via Wikimedia Commons.
  • High-stakes factual applications where real-time verification is required.

Dataset Structure

Each data point represents a multimodal fragment with the following fields:

  • id: Unique identifier.
  • title: Wikipedia page title.
  • text: Cleaned paragraph text.
  • url: Wikipedia page URL.
  • images: Struct containing image PIL objects, captions, licenses, and metadata.
  • fragment: The fragment rendered as a stand-alone image (grid of images + paragraph).
  • full_embeddings: Multi-vector embeddings from ColQwen2 v1.0.
  • pooled_embeddings: Compressed 9-vector representations (special token centroid + 8 content centroids).

Embedding Methodology

We use ColQwen2, which follows a late interaction architecture. Given a query qq and a fragment (document) dd, they are encoded into multi-vector representations EqE_q and EdE_d. The relevance score Sq,dS_{q,d} is computed as:

Sq,d=βˆ‘i∈[∣Eq∣]max⁑j∈[∣Ed∣]Eqiβ‹…EdjS_{q,d} = \sum_{i \in [|E_q|]} \max_{j \in [|E_d|]} E_{q_i} \cdot E_{d_j}

To handle the memory footprint of storing millions of multi-vectors, we implement a token pooling strategy. The document embedding sequence is partitioned into:

  • EdprefE_d^{pref}: Prefix special tokens.
  • EdcontentE_d^{content}: Visual and textual content embeddings.
  • EdsuffE_d^{suff}: Suffix special tokens.

The pooled representation EdpoolE_d^{pool} is then computed as:

Edpool=c(EdprefβˆͺEdsuff)βŠ•CdE_d^{pool} = c(E_d^{pref} \cup E_d^{suff}) \oplus C_d

where c(β‹…)c(\cdot) represents the centroid of the special tokens and CdC_d represents KK centroids (with K=8K=8) obtained by performing hierarchical clustering on the content embeddings.

Dataset Creation

Curation Rationale

The dataset was created to facilitate "ArtSeek," a framework for deep artwork understanding. By focusing on the Visual Arts domain, we provide a high-quality benchmark for evaluating retrieval and reasoning capabilities in a knowledge-rich field where visual and textual context are inseparable.

Source Data

The data is sourced from the English Wikipedia (August 2024 snapshot) and the Kiwix full Wikipedia ZIM dump (January 2024).

Data Collection and Processing

  1. Filtering: Pages were selected by recursively descending from "Category:Visual arts" to a depth of 5.
  2. Extraction: Paragraphs and images were extracted using a modified wikiextractor.
  3. Rendering: Fragments were rendered into images using the FragmentCreator tool, placing images in a grid above the text.
  4. Embedding: We extracted embeddings using ColQwen2 and applied the clustering-based pooling mentioned above to create efficient retrieval indices.

Who are the source data producers?

Text was authored by Wikipedia contributors. Images were contributed to Wikimedia Commons by various users and are subject to individual licenses.

Annotations

There are no manual annotations beyond the original captions associated with images from Wikipedia pages.

Annotation process

N/A.

Who are the annotators?

N/A.

Personal and Sensitive Information

The dataset is derived from public Wikipedia data and is not expected to contain sensitive personal information.

Bias, Risks, and Limitations

  • Coverage Bias: Inherits biases present in English Wikipedia regarding art history (e.g., potential Western-centric focus).
  • Temporal Limitation: Reflects a snapshot in time.
  • Image Quality: Uses lower-resolution images optimized for web rendering from Kiwix.

Recommendations

Users should be aware of the inherited biases from Wikipedia contributors and editorial processes. Verify image licenses via provided URLs for any distribution.

Citation

BibTeX:

@article{fanelli2025artseek,
  title={ArtSeek: Deep artwork understanding via multimodal in-context reasoning and late interaction retrieval},
  author={Fanelli, Nicola and Vessio, Gennaro and Castellano, Giovanna},
  journal={arXiv preprint arXiv:2507.21917},
  year={2025}
}

APA:

Fanelli, N., Vessio, G., & Castellano, G. (2025). ArtSeek: Deep artwork understanding via multimodal in-context reasoning and late interaction retrieval. arXiv preprint arXiv:2507.21917.

Glossary

  • Late Interaction: A retrieval mechanism that computes similarity by summing the maximum dot products between query and document token embeddings.
  • Token Pooling: A technique to reduce the number of vectors stored per document by clustering embeddings into a fixed set of centroids.

Dataset Card Authors

Nicola Fanelli

Dataset Card Contact

For questions, please contact: nicola.fanelli@uniba.it

Downloads last month
2,904

Collection including cilabuniba/wikifragments-visual-arts-embeds

Paper for cilabuniba/wikifragments-visual-arts-embeds