Melvynkoh94/LSTM_Keras

competition: https://www.kaggle.com/c/jigsaw-toxic-comment-classification-challenge/overview evaluation: 6 class each binary, column-wise ROC AUC, https://www.kaggle.com/c/jigsaw-toxic-comment-classification-challenge#evaluation dataset: dir data/train.csv, data/test.csv technology: deep learning CNN http://cs231n.github.io/convolutional-networks/ lstm: http://colah.github.io/posts/2015-08-Understanding-LSTMs/, 自己写的，扔了就跑 ： https://zhuanlan.zhihu.com/p/35756075 bi-lstm: word embedding: https://towardsdatascience.com/introduction-to-word-embedding-and-word2vec-652d0c2060fa GloVe: embedding/glove.pdf keras: https://keras.io/zh/preprocessing/text/ network: tokenize -> word embedding -> (bi-lstm or lstm) -> CNN (3*3) (or deeper, see lstm_keras.py) -> pooling -> dense net -> 6 binary sigmoid for classify