Research On Text Semantic Smoothness Calculation Based On Neural Network Model

With the sharp increase of text data,the intelligent processing and analysis of text has become one of the important ways to solve the problem of information overload,and the semantic research of text is the key.As an important direction of text semantic researclh,the study of the smoothness of text semantics is relatively rare.It is undeniable that the study of textuality is also crucial for machine translation,intelligent correction of articles,etc.However,how to judge whether sentences are fluent and whether the organization of sentences is reasonable is a very important and challenging research topic.The traditional manual processing method not only requires a large amount of labor cost and time resources,but also has large inaccuracies and differences.If we can construct an effective model by studying the semantics of textuality and let the computer judge whether a sentence is fluent in a human-like way,it will hopefully alleviate the existence of traditional manual corrections,such as the inability to correct a large amount of text,inefficiency and consistency.Evaluation and other issues.This paper proposes a semantic fluency calculation method based on dependency syntax analysis,which calculates the fluency by combining the syntactic sufficiency of the sentence stem and the semantic fluency of the sentence details.This paper also proposes a semantic smoothness calculation method based on a combination of neural network model and machine learning algorithm.The method first classifies the article,uses word2vec to represent the word vector,and then adds the convolutional neural network(CNN)layer to the model.The word vector is learned,and the word vector and word vector are combined by the attention mechanism,and then learned by the bidirectional long-term memory network model(Bi-LSTM),and finally input to the conditional random field layer(CRF)for calculation and output.In the course of the experiment,the data is processed by the undersampling method for the problem of the positive and negative sample imbalance in the data set of the paper.The experimental results show that compared with the dependency syntax analysis,the model combined with the neural network model and the machine learning algorithm is more suitable for the semantic smoothness calculation of the text.We have carried out experimental verification on a manually annotated data set,preliminary proof of the feasibility and effectiveness of the above model.