Visual News
收藏资源简介:
Visual News是一个大规模的新闻图像标题数据集,包含超过一百万条新闻图像及其相关新闻文章、图像标题、作者信息和其他元数据。该数据集由马里兰大学创建,旨在为新闻图像标题生成任务提供丰富的信息,如事件和实体。数据集内容丰富,涵盖多种新闻来源和主题,支持新闻图像标题生成的研究。创建过程中,数据集经过筛选和处理,确保图像和文本的质量。该数据集的应用领域包括新闻图像标题生成、文本摘要和假新闻检测等,旨在解决新闻图像理解和信息提取的挑战。
Visual News is a large-scale news image captioning dataset containing over one million news images along with their associated news articles, image captions, author information and other metadata. Developed by the University of Maryland, this dataset is designed to provide rich contextual information including events and entities for the news image captioning task. It features abundant content spanning diverse news sources and topics, supporting research on news image captioning. During its construction, the dataset underwent strict filtering and processing to ensure the quality of both images and their accompanying text. Its application scenarios include news image captioning, text summarization, fake news detection and other fields, aiming to address the core challenges in news image understanding and information extraction.

- 1Visual News: Benchmark and Challenges in News Image Captioning马里兰大学 · 2021年



