This repo contains the data sets of our paper:
Each line consists of a comment text and its quadruple, separated by "####", and there are multiple quadruples for a sample. Each tetrad is formally named 'Category Aspect Opinion Sentiment'. Aspect and Opinion are represented by the character index in which they appear in the original text, e.g. "10:18" means the 10th to 18th character of the original text." -1,-1" indicates the implicit case. The "0", "1", "2" in the Sentiment represents negative, neutral, and positive, respectively.