The dataset for paper "Why Do We Click: Visual Impression-aware News Recommendation", ACM MM 2021
Dataset
- The filename corresponds to the newsid in the original dataset.
- The download link for our dataset: https://drive.google.com/file/d/1gx0OzN7qSuyRlvN0cfVUjB4tmoKvvQk1/view?usp=sharing
- imageList.npy: The newsid list of successfully crawled images in the format of [newsid,...].
Utils
- data_generator.py:For generating IM-MIND dataset (news text and cover images are required).
- word_feature_generator.py:For generating visual impression representations of words.
- global_feature_generator.py:For generating global impression representations.