jefferyYu / UMT

Preprocessed Datasets for our Multimodal NER paper
111 stars 13 forks source link

dataset Twitter2017 #6

Closed dwchen123 closed 2 years ago

dwchen123 commented 4 years ago

你好,Twitter2017数据集里的图片是不是没有上传好,我下载了几次都会有部分图片不全 image

jefferyYu commented 4 years ago

Hi there,

For both Twitter-2015 and Twitter-2017 datasets obtained by us, there are a few smaples with very small or broken images. As shown in the begining our running log files, there are a small number of problematic images in the training, dev, test sets. In our experiments, we have replaced these problematic images with a special image which contains a text box "Image Not Found".

Hope it clarifies your concern.