METHOD AND DEVICE FOR PROCESSING INFORMATION, AND STORAGE MEDIUM

A method for processing information, includes: performing word segmentation on first corpus data to obtain second corpus data containing at least one word; processing, based on a first pretrained model, the at least one word contained in the second corpus data, to obtain a first eigenvector sequence corresponding to the second corpus data; processing, based on a second pretrained model, characters contained in the first corpus data, to obtain a second eigenvector sequence corresponding to the first corpus data; and obtaining a target eigenvector based on the first eigenvector sequence and the second eigenvector sequence. In the disclosure, by characterizing corpus data with vectors based on words and characters respectively, multiple types of features are fused, so that training samples acquired for classification are more abundant, and the precision and the effect of generalization of a model obtained by training are significantly improved.