Kaggle Competition: Jigsaw Unintended Bias in Toxicity Classification

一些优化方向思路，仅供参考:D

• EDA

• Preprocess

  可参考之前比赛的思路，解决OOV问题

  • BPE
  • TTA

• Model

  • Sequence model: bilstm, HAN...
  • Bert fine tune

• Loss

  • Focal loss (根据identity决定alpha)

• Metrics

  • 实现比赛AUC指标评测方法

• Argument

  • Adversarial Training
  • 训练identity分类器(40W examples), 对jigsaw上个比赛数据进行分类,抽取identity负样本
  • VFAE

• Tricks

  • multi-task
  • sample_weights

• Managed with git-lfs (X): can not upload new objects to public fork
• Dealing with OOV and data imbalanced problem

detect toxic comments ― and minimize unintended model bias

根据评论数据文本，判断是否为 toxicity （ toxicity is defined as anything rude, disrespectful or otherwise likely to make someone leave a discussion. ） ，并减小模型的unintended bias，使模型输出结果更加公正

模型搭建位于models文件夹下，bert模型位于keras_layers/keras_bert

• Word2v + bpe + 2xbiLSTM
• ELMo + 2xbiLSTM
• Word2v + DGCNN
• bert fine-tune

• Sample Weight

  • Subgroup及Subgroup负样本Loss加权

• Custom object function

  • Rank Loss、Focal Loss

• Data Augmentation

  • 扩充Subgroup sample
  • 平衡Subgroup 正负样本比例

在trainer当中选择训练模型, (TODO: 添加命令行参数)

nohup python -u trainer.py > train.log 2>&1 &