Evaluation of Structural Hyperparameters for Text Classification with LSTM Networks

A Method for MBTI Classification Based on Impact of Class Components

“…After that, we create two models using Bi-LSTM and CNN architectures. Our goal was not to find the optimal architecture for MBTI classification, as in [55], but to prove that the proposed method improves results with different architectures. In addition, since LSTM architectures are trained to recognize patterns across time, and CNN architectures recognize patterns across space, weighting parameters could lead to insights into the behavior of compound class labels.…”