A Deep Learning Text Classification Method Integrating Shallow Semantic Representation Vectors

Abstract

The invention discloses a deep learning text classification method integrating shallow semantic representation vectors. The method includes: first training word embedding vectors on text corpus, and secondly using domain vocabulary dictionary as shallow semantic vocabulary to generate text based on shallow semantic vocabulary A shallow semantic vector representation of each word in the corpus. Next, the weighted splicing of the two word vectors is input into the CNN model as a new word vector for feature extraction and model training to build a text classifier. The invention solves the defect that the big data-driven word vector expression lacks lexical features and knowledge representation, and it is difficult to truly understand the semantic information of the words, so that the model has richer feature expressions and higher classification performance.

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 representation vectors. Background technique [0002] Text classification refers to the process of predicting category attribution for a large amount of unstructured text corpus according to a given classification system. With the breakthrough of deep learning technology, word embedding technology represented by word2vec and deep learning model represented by convolutional neural network have achieved good results in text classification. But overall, the accuracy and reliability are far from practical levels, which is caused by the lack of prior knowledge of deep learning. Because big data-driven deep learning models can only find statistical conclusions in the data set, it is difficult to effectively utilize prior knowledge. Integrating prior knowledge into deep learning models is an id...

AI Technical Summary

Problems solved by technology

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

Images