ArtELingo

displayName: ArtELingo
license:
- ArtELingo Custom
paperUrl: https://arxiv.org/pdf/2211.10780.pdf
publishDate: "2022"
publishUrl: https://www.artelingo.org/
publisher:
- King Abdullah University of Science and Technology
- University of Notre Dame
- Northeastern University
tags:
- Artistic language

---
# 数据集介绍
## 简介
本文介绍了ArtELingo，这是一个新的基准和数据集，旨在鼓励跨语言和文化的多样性工作。继ArtEmis之后，来自WikiArt的80k艺术品收藏有0.45万个情感标签和仅英文标题，ArtELingo又增加了0.79万个阿拉伯语和中文注释，加上4.8万个西班牙语注释，以评估 “文化转移” 的表现。超过51k的艺术品有3种语言的5个注释或更多。这种多样性使得研究跨语言和文化的异同成为可能。此外，我们研究了字幕任务，发现多样性提高了基线模型的性能。ArtELingo是公开可用的，具有标准拆分和基线模型。我们希望我们的工作将有助于简化未来对多语言和文化意识的人工智能的研究。

## Download dataset
:modelscope-code[]{type="git"}

displayName: ArtELingo
license:
- ArtELingo Custom
paperUrl: https://arxiv.org/pdf/2211.10780.pdf
publishDate: "2022"
publishUrl: https://www.artelingo.org/
publisher:
- King Abdullah University of Science and Technology
- University of Notre Dame
- Northeastern University
tags:
- Artistic language
---
# Dataset Introduction
## Introduction
This paper presents ArtELingo, a novel benchmark and dataset designed to promote cross-lingual and cross-cultural diversity research. Building on ArtEmis, which features 80k artwork collections from WikiArt with 0.45k emotional tags and English-only captions, ArtELingo adds 0.79k annotations in Arabic and Chinese, alongside 4.8k annotations in Spanish, to evaluate model performance on "cultural transfer" tasks. Over 51k artworks have at least 5 annotations across three languages. This diversity enables research into the similarities and disparities across languages and cultures. Additionally, we conduct experiments on the image captioning task and find that enhanced diversity improves the performance of baseline models. ArtELingo is publicly available, with standard data splits and baseline models provided. We hope our work will help streamline future research on multilingual and culturally-aware artificial intelligence.
## Download Dataset
:modelscope-code[]{type="git"}