LOC and MISC type training

[ShaoQiBNU]

PER 和 ORG 的 python feature_pu_model.py --dataset conll2003 --flag PER 可以正常得到结果，但是LOC和MISC始终是 Precision: 0, Recall: 0.0, F1: 0，得不到结果

[JackySnake]

我也遇到了这个问题，请问您解决了吗？

[XINGXIAOYU]

I guess this is probably caused by the class imbalance problem. You should set different class balance rate for LOC and MISC.

[JackySnake]

I guess this is probably caused by the class imbalance problem. You should set different class balance rate for LOC and MISC.

Thank you. It is truly caused by the class weight(descript in Loss definition , p5).
I want to know the class weight of each class in experiment of the paper, could you provide the accurate values for reference?
And what the range of the class weight?
If the corverage of entity class lexicon is good ,the class weight should be a bigger value, is that right?

[pmouren]

请问大家有解决这个权重参数的设置问题么，我想在中文数据集上进行复现也遇到了同样的问题，但是不知道按照什么依据来设置合理的参数值