| Data: Emotion dataset “M”, Train Set “TAS”, Test Set “TSS” |
Result: Review Text Label: “J-S”, “F-S”, “F-G” |
Start |
// Review Text Encoding toward machine understandable word vectors (real valued) |
while each review text RM do |
while each word TM do |
(1) | Word(token) indices allocation |
End while |
End while |
Initializing Hyperparameter |
(2) | embed_dim = 100, 128,300, max_features = 2000, epochs = 7, batch_size = 32, train set = 90%, test size = 10% |
//Deep Learning model training |
while each review text R MTASdo |
(3) | Generate all word embedding vectors in R = [r1, r2, r3, …., rn] |
(4) | Implement Bi-LSTM operation exploiting equations (1)–(13) |
End while |
// Allocating a label to Review Text final depiction |
while each Review Text R MTSSdo |
(5) | Trained(learned) model is built |
(6) | Employ a softmax classifier using Eq. 14, for the classification of output obtained from the Bi-LSTM into “J-S”, “F-S”, “F-G” |
End while |
End |