Included are 100,000 random tweet IDs from the investigated dataset. This is as per Twitter's data distribution policy. An example of training CIDER along a weather and sentiment linguistic axis has been attached.