Emotion Detection from Text Using Convolutional Neural Networks

The paper under consideration describes an emotion-detection model on text as an example of the working within the Convolutional Neural Network framework model. The data had four sets of emotions i.e., joy, anger, fear and sadness which are highly preprocessed and equalized to achieve the believable training and assessment of the model. This model was trained with 30 epochs and batch size of 32 with the learning rate of 0.001 that gave averaged convergence. An increase in accuracy on each epoch at training with a starting point of 27 percent and ended with 98 percent and validation accuracy had 93 percent. Still in the same note, the loss had decreased to 0.04 as opposed to 1.65 on the training set and 0.19 on the validation loss, which is a good generalization. The evaluation of the performance was done and total accuracy was 93, wherein there was a balance between the precision, recall and <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{F 1}$</tex>-score across classes. Findings reveal the power of CNN-based models to perform the emotion recognition problem in real-life scenarios, which involves text.

Hali tarjima qilinmagan