Knowledge Base Question Answering by Transformer-Based Graph Pattern Scoring

[DavidBlavid]

relevant issue: https://github.com/KGQA/leaderboard/issues/172

[xixi019]

@DavidBlavid you need to update the changes to resolve the conflicts

[DavidBlavid]

The merge conflicts have been resolved. Still, the F1 score of the paper seems pretty high, could someone double check?

[xixi019]

I am not sure whether the results could be listed since they remove certain types of questions in the dataset, according to Data Quality in section 4. The original dataset of simple questions consists of more than 10K questions according to https://arxiv.org/pdf/1506.02075v1.pdf, while in the paper they only use 8998 of them. What do you think @LongquanJiang