Research And Application Of Text Classification Based On Multi-Semantic Fusion Learning

With the arrival of the Internet era,massive data is flooded on the network,which contains huge value behind the massive information.Therefore,how to accurately process and effectively use the information is one of the hot topics in today’s research.Text classification is an effective method to classify and manage text information.Through in-depth analysis of text semantics,text classification technology can accurately summarize similar texts,simplify and classify complex data,relieve people’s pressure of identifying disorganized data,and help people to effectively identify and utilize information.Nowadays,text classification methods based on deep learning emerge in an endless stream.Text classification models such as Text CNN and Text RNN have come into people’s view.These models have greatly improved the performance of text classification.Text classification is very important for the overall grasp and understanding of text semantics,but most of the existing text classification models only consider the local semantic features of text,ignoring the global features of text,resulting in the model’s understanding of text is not comprehensive and accurate,thus affecting the accuracy of text classification.In addition,in the past,most of the models only considered the semantic features of the context,ignoring other text features,such as grammatical features,label semantic features,etc.,which can help the model to have a comprehensive and accurate understanding of the text and improve the accuracy of text classification.In view of the above problems,this thesis has done the following work:1、A text classification algorithm based on graph neural network and attention mechanism is proposedTo solve the above problems,this thesis proposes a method of semantic feature mining using graph neural network.First,the semantic feature representation of the text is obtained by word embedding.Then,the semantic feature map is established for each text based on the semantic feature of the text,so that each text has its own semantic feature map,which is conducive to reducing the complexity of the structure of the text map and reducing the consumption of memory.After that,making use of the semantic information transmission,through which the contextual semantic information of the text is comprehensively obtained.Finally,the attention mechanism is used to obtain the key semantic information in the text,and classifiers are used to accurately classify the text according to the obtained semantic information.Experiments on authoritative data sets show that this method can effectively improve the accuracy of text classification.2 、 A text classification algorithm based on graph neural network information fusion is proposedThe first method proposed in this thesis is not comprehensive and rich in semantic features and ignores other important textual features.To solve the above problems,this thesis introduces the bert model and the grammatical features of text,strengthens the mining of the semantic features of text,and uses the dependency relation of text clauses to enhance the semantic of text.By using Bert pre-training model,richer semantic features can be obtained,and at the same time,the extraction of grammatical features can be strengthened.Finally,syntactic dependence and semantic relation are combined to improve the accuracy of text classification prediction.A lot of experiments show that this method is effective in improving the performance of text classification.3、Design and implement a text classification system based on the integration of multiple semantic featuresThrough the proposed text classification algorithm based on multiple semantic features,the corresponding text classification system is realized.The text classification system is accurate and the text classification efficiency is high.Through the extensive system testing of the system,it is found that the system can effectively classify the text data accurately and help users quickly find the corresponding text.