Research On Intelligent Semantic Checking Algorithm For Deep Learning In Article Editing

The development of technology has made the amount of data in all walks of life increasingly large,and the printing industry has also changed.This paper studies the problem of statement correction in the editing of text-based manuscripts in the printing process.This paper designs and implements a semantic checking algorithm based on LSTM neural networks.This algorithm first collects,organizes and constructs a standard corpus data set.Secondly,it uses the word embedding algorithm to map the processed corpus data to the word vector space.Finally,the result obtained by the word embedding algorithm is used to identify the segment and in the statement.The word prediction,the language model of the inspection.The details are as follows:(1)A new word embedding method is designed,and a new word vector model is constructed by this method.The word embedding method is firstly designed according to the collected related text corpus,Chinese language grammatical relationship and the shortcomings of existing word embedding algorithm order and global collinearity.Secondly,a matrix of feature columns with phrases as words is established by the frequency of statistical words-phrases.The clustering algorithm reduces the dimensions of the matrix and maps the words into a low-dimensional word vector space to construct a new one.Word vector model.Finally,another word vector model is constructed using the existing word embedding algorithm.The two word vector models are compared and analyzed.Although the word vector model generated by clustering the feature matrix according to the frequency of the related text statistic-the phrase is slightly better than the existing word embedding algorithm in global collinearity,the clustering is based on The overall effect of the word and word group embedding method is slightly inferior to the word vector model constructed by the existing word embedding algorithm.(2)Using the two word vector models to construct a joint prediction model based on the LSTM-based language model.The model first uses the LSTM-based language model constructed by two word embedding algorithms to predict the text,and then combines the results to generate the final prediction result.The experimental results of using the language model for word prediction accuracy show that the accuracy of the joint prediction model is higher than that of the two single models,and it is a stable and effective method for solving the word check of the sentences in the relevant topic texts.Semantic check method.