Research And Design Of Verb Error Detection Algorithm Based On Deep Learning

English essay automatic approval,is a rapid development in recent years.It gradually replaced the teacher manual reading,as an important tool to alleviate the burden of English teachers teaching.At the same time,through the literature research,verbal consistency error and verb tense error is the highest error rate in English composition of the two types of grammatical errors.Therefore,the test results for verb errors can reflect the usefulness of an automated review system.At this stage,the mainstream of the automatic access system has Bingguo,sentence cool and so on.After the investigation,these systems do not satisfy the learner's requirements for verbal consistency errors and verbal tense errors.In this paper,a verb-based grammar error detection algorithm based on depth learning is developed.Through the research and analysis,it is found that the occurrence of verbal consistency errors and verb tense errors is related to the words and phrases appearing in the context,and the LSTM(Long Short-Term Memory)can effectively preserve the context,This paper decides to use LSTM as a training model to model the trained training corpus.At the same time,how to convert the text in the English writing into numerical values for subsequent calculation is also an important step in automatic approval.Most of the mainstream tools use the word pocket model,that is,according to the order of each word in the dictionary to encode.Although this encoding is easy to use,but both the loss of the word position information,but also prone to dimension disaster.Therefore,this paper uses the word embedding model to encode the text,and maps the text information to a low dimension vector space in order,which avoids the position information of the text and avoids the dimension disaster.After that,this paper collects a certain sample of corpus,compares this algorithm with the sentence cool and ice fruit,and verifies the superiority of the algorithm in verbs.In this paper,the correctness of the verbs based on the depth of learning,the accuracy rate of the whole algorithm,the recall rate and the F1 degree are better than the current mainstream automatic reading system.