Toxic-Comment-Classification-

classify comments into types of of toxicity like threats, obscenity, insults, and identity-based hate

data:https://www.kaggle.com/c/jigsaw-toxic-comment-classification-challenge/data

LSTM model outline: Layer (type) Output Shape Param #

input_2 (InputLayer) (None, 300) 0

---

embedding_2 (Embedding) (None, 300, 128) 2560000

---

lstm_layer (LSTM) (None, 300, 50) 35800

---

global_max_pooling1d_2 (Glob (None, 50) 0

---

dropout_3 (Dropout) (None, 50) 0

---

dense_3 (Dense) (None, 10) 510

---

dropout_4 (Dropout) (None, 10) 0

---

dense_4 (Dense) (None, 6) 66

CNN+BiLSTM+Glove

---

Layer (type) Output Shape Param # Connected to

input_9 (InputLayer) (None, 300) 0

---

embedding_20 (Embedding) (None, 300, 300) 6000000 input_9[0][0]

---

conv1d_19 (Conv1D) (None, 300, 30) 27030 embedding_20[0][0]

---

dropout_14 (Dropout) (None, 300, 30) 0 conv1d_19[0][0]

---

bidirectional_7 (Bidirectional) (None, 300, 400) 369600 dropout_14[0][0]

---

global_average_pooling1d_7 (Glo (None, 400) 0 bidirectional_7[0][0]

---

global_max_pooling1d_7 (GlobalM (None, 400) 0 bidirectional_7[0][0]

---

concatenate_7 (Concatenate) (None, 800) 0 global_average_pooling1d_7[0][0] global_max_pooling1d_7[0][0]

---

dense_14 (Dense) (None, 6) 4806 concatenate_7[0][0]

Total params: 6,401,436 Trainable params: 6,401,436 Non-trainable params: 0