The noisy labeling problem has been one of the major obstacles for distant
supervised relation extraction. Existing approaches usually consider that the
noisy sentences are useless and will harm the model's performance. Therefore,
they mainly alleviate this problem by reducing the infl