Increasing Text Filtering Accuracy with Improved LSTM

keywords: Texting filtering, LSTM, word vector, CBOW, dropout, CNN, DAN
How to eliminate useless information in the vast network information and retain effective information is a problem that needs to be continuously explored in the field of deep learning. This paper conducts text classification on the network evaluation frequently encountered in daily life, mainly to screen out the meaningless comments published by Internet users, to have access to more useful information. In this paper, a text filtering model was constructed based on word vector and Long Short-Term Memory (LSTM) and improved by adding Deep Averaging Networks (DAN) and convolutional neural network (CNN). The major improvement of the LSTM & DAN model was to retain the original word vector information and to improve the accuracy of the text classification model without increasing hyperparameter and model structure complexity. The LSTM & CNN model mainly combines the advantages of convolutional neural network in exploring the deep information of text, which was an improvement over the original LSTM. It was proved by experiments that this improvement is meaningful. Compared with the shallow neural network, the accuracy has been greatly improved.
reference: Vol. 42, 2023, No. 6, pp. 1491–1517