arXiv cs.CL· \v{S}\'arka Uramov\'a, Petar Kirilov Petrov, Josef Jon, Ond\v{r}ej Bojar, Michal Nov\'ak·· 15 小时前AI 评分31
ImgMT：基于维基百科的图文翻译评测数据集
ImgMT: Wikipedia-based Dataset For Evaluating Text-in-Image Translation
AI 导读
研究者发布 ImgMT，一个从 Wikimedia Commons 本地化 SVG 插图中挖掘的多语言图文翻译数据集与基准，含 35,612 组对齐图像对、1,855 个图像集，覆盖 129 种语言、20 多种书写系统和 4,886 个语言对。
正文 · 原文
Abstract:Text-in-image translation aims to translate all textual content embedded in an image. In Text Image Translation (TIT), the output is translated text; In-Image Machine Translation (IIMT) additionally renders the translation back into the image. We introduce ImgMT, a massively multilingual dataset and benchmark mined from localized SVG illustrations on Wikimedia Commons, focusing on diagrams, maps, and infographics. ImgMT contains 35,612 aligned image pairs grouped into 1,855 image sets, spanning 129 languages, 20+ writing systems, and 4,886 language pairs.
We propose a multi-level evaluation protocol that measures source-side text-region detection and OCR accuracy alongside target-side translation quality. We benchmark several multimodal large language models and a dedicated TIT system in English->X and X->English settings, and find that the source language and writing system substantially affect TIT quality. Finally, we show that ImgMT is not well suited for isolating the contribution of visual information in multimodal translation.
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2610.35416 [cs.CL] |
| (or arXiv:2610.35416v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2610.35416 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Šárka Uramová [view email]
[v1]
Mon, 26 Oct 2026 13:26:57 UTC (2,762 KB)
来源：arXiv cs.CL · arxiv.org