False complaint detection method based on dual-channel feature contrast learning

The invention discloses a complaint and report false information detection method based on two-channel feature comparative learning, and belongs to the field of artificial intelligence. Dividing the original data into a label data set and a label-free data set, and inputting the label-free data set into a pre-training model for pre-training; performing word segmentation processing on the data in the training set, and establishing a word list to obtain corresponding vectors; inputting the corresponding vectors into a pre-training model with dropout to generate corresponding text features, and inputting the text features into a dual-channel classifier to obtain different two-dimensional vector features; obtaining different two-dimensional vector features through two channels, and obtaining a KL divergence loss value; performing comparative learning on the text features and the classifier features to obtain comparative loss and cross entropy loss, and performing back propagation on the total loss value and updat