Research On The Diagnosis And Repair Of Short Text Error Based On Web

In this era of highly informationalization,online learning communities play an important role in people's learning.For example,Zhihui,CSDN forums,etc.,people will publish their own questions that are doubtful,and at the same time,they will help answer some of their understanding.The problem.As more and more learners begin to use the online learning community to solve their own learning problems,there are some erroneous natural language descriptions that often interfere with learners' own learning and analysis.In addition,as more and more work is done on the learner's diagnosis,prediction,and recommendation services,for example,through the analysis of questions and answers,the learner's behavior,cognition,and emotions are represented or some learners may resolve The questions are recommended for him to answer,etc.The accuracy of the natural language description is also more demanding,because these tasks are all dependent on the quality of the data used,if there is a problem with the data quality,even if the use of sophisticated analytical methods Not trustworthy.The original textual description of mistakes was mostly detected manually,but the resulting cost was huge and it took a lot of time and effort.Later,the detection of textual error data was mainly based on word matching,and the word errors were found by comparing the thesaurus.In recent years,people have started to integrate into machine learning or deep learning based on a single rule or statistical method to improve the accuracy of the algorithm and the accuracy of the experiment.But overall,these methods are relatively single,and the types of error texts solved are relatively single and have not been implemented.A set of methods for diagnosing and repairing erroneous text data.In order to improve the efficiency and accuracy of automatic diagnosis and repair of short texts,this article will improve some single methods.Through the analysis,the main points have been made as follow:First of all,we classify the possible types of errors in the short texts of the online learning community,and give a reasonable definition of each type of error.Secondly,a method based on the part-of-speech collocation method is used to diagnose the errors of the text grammar structure,and then the grammatical errors are repaired by using the knowledge collocation library.The n-gram probabilistic statistical model,contextual context,and the Chinese fixed collocation method are used to automatically diagnose and repair the sounds.Similar text errors in similarity.Finally,based on knowledge of crawling problems and answers related to machine learning topics as data sets,the following experimental work was performed.First,the diagnosis and repair of text grammar structure problems was firstly based on the word-of-speech method.Then use the theory of evidence to diagnose indeterminate and erroneous cases.The results of relevant experiments show that the accuracy of diagnosis has been significantly improved after the addition of evidence theory.Finally,the knowledge is used to match the library to correct the errors.The experimental results show that there are certain errors in the repair.The second is to compare the method of similarity error diagnosis and repair proposed in this paper with a single method.The experimental results show that the accuracy of the proposed method and recall rate are improved obviously.