Deep learning text classification method integrated with shallow semantic pre-judgment mode

Abstract

The invention discloses a deep learning text classification method integrated with a shallow semantic pre-judgment mode and the method comprises the steps: firstly, carrying out the conventional CNN deep learning training of a text corpus, which comprises word embedding, convolution, pooling, and mode output; secondly, using a domain vocabulary dictionary as superficial layer semantic vocabularies, and calculating a superficial layer semantic pre-judgment mode based on the superficial layer semantic vocabularies; performing dual-mode fusion on a shallow semantic pre-judgment mode and a deep learning decision-making mode to serve as a final decision-making mode of a SDG-CNN model. so that a loss function is constructed and parameter optimization is implemented by using the decision mode. According to the method, the defects that a traditional deep learning model lacks background knowledge and semantic information in the model optimization process and the information mode is single are overcome, and the performance of the deep learning text classification model is improved.

Description

technical field [0001] The invention relates to the field of deep learning and text classification, in particular to a deep learning text classification method integrating shallow semantic prediction modalities. Background technique [0002] Text classification refers to the process of predicting the category belonging to a large amount of unstructured text corpus according to a given classification system. With the breakthrough of deep learning technology, the deep learning model represented by convolutional neural network has achieved good results in text classification. But in general, the accuracy and reliability are far from reaching the practical level, which is caused by the lack of prior knowledge of deep learning. Because the deep learning model driven by big data can only find statistical conclusions in the data set, it is difficult to effectively use prior knowledge. Integrating prior knowledge into deep learning models is an idea to solve the bottleneck of deep...

AI Technical Summary

Problems solved by technology

[0003] The present invention provides a deep learning text classification method SDG-CNN (Semantic Decision Guide Convolutional Neural Network) integrating shallow semantic prediction mode, which overcomes the lack of background knowledge and semantic information in the process of model optimization of traditional deep learning models, Flaws of a single information modality

Images