search

cmavro / ReaRev_KGQA

[EMNLP Findings 2022] ReaRev: Adaptive Reasoning for Question Answering over Knowledge Graphs
https://aclanthology.org/2022.findings-emnlp.181/
MIT License
22 stars 2 forks source link

Is there a webqsp dataset of 10%, 30%, 50% KG after preprocessing? #9

Open firwww opened 1 year ago

firwww commented 1 year ago

Is there a webqsp dataset of 10%, 30%, 50% KG after preprocessing?

cmavro commented 1 year ago

Hi,

Thank you for your interest in our work. We use the incomplete KGs as processed in the Graftnet repo; we downloaded the KGs from there.

firwww commented 1 year ago

Thank you for your answer, do you have a webquestion training log under 50KG, I am difficult to reproduce your results, are the parameters of this experiment the same as the full KG

firwww commented 1 year ago

How to set up a 50KG metaqa? is for i, tpl in enumerate(sample['subgraph']['tuples']): if i % 2 == 0: continue or --fact_drop=0.5? or both？please！

cmavro commented 1 year ago

As far as I recall, we performed hyper-parameter tuning for the incomplete KGs (I have not saved the log files).

For MetaQA, please use if i % 2 == 0: continue.