Closed lihuiliullh closed 1 year ago
Hi thanks for reaching out! I believe both files for NELL are taken from the MetaR repo and the original NELL-One dataset by Xiong et al. (https://arxiv.org/pdf/1808.09040.pdf). I generated them similarly for the other two datasets.
I believe rel2candidates.json is supposed to contain all possible candidates for the tail node, and e1rel_e2.json is just adjacency list cached for easy access. For the two datasets I added, candidate set is simply all nodes.
May I know what kind of data are stored in file rel2candidates.json, e1rel_e2.json and path graph? Can you share how you process the original triple data step by step?