Font Size: a A A

Research And Implementation Of Chinese Grammar Automatic Error Correction System

Posted on:2021-05-01Degree:MasterType:Thesis
Country:ChinaCandidate:J C ZhouFull Text:PDF
GTID:2415330605466979Subject:Software engineering
Abstract/Summary:PDF Full Text Request
In the context of globalization,Chinese has become one of the most popular languages in the world.For Chinese learners,learners do not have rich systematic grammatical knowledge,and under the influence of their mother tongue,it is difficult to accurately identify and correct grammatical errors.In order to fill the shortcomings,the Chinese grammar error correction system becomes very necessary.In the publishing industry,the number of electronic documents has increased significantly and is becoming more and more abundant.The manual error correction method requires more time and effort.Using the Chinese grammar error correction system can greatly reduce labor costs.Aiming at the task of Chinese grammatical error correction,this paper proposes a grammatical error correction scheme based on mixed model.Errors are divided into low-level grammatical errors and high-level grammatical errors.N-gram and CRF models are used to correct low-level grammatical errors.In the face of high-level grammatical error correction,error correction tasks can be handed over to deep neural networks.The core of the Chinese grammar error correction system is the realization of the error correction function,which mainly depends on the use of a variety of deep models,such as rnn_attention,rnn_crf,conv_seq2seq,seq2seq_attention,transformer,bert.Each model can run independently.The error correction dataset used in the experiment is derived from the training corpus published by the 2018 NLPCC.The parallel corpus is obtained through preprocessing,and 30,000 sentences are taken as the test set.The rest is used as the training corpus.The division method uses a random division method.Each sentence and the corresponding correct sentence form a sample pair,and the corpus is used to train a deep neural network model.In addition,each model can independently preprocess data,be trained,and predict,and the grammatical error correction effect is significant.The construction and optimization of the grammar error correction module has been Completed.On the basis of the realization of the interactive function,the Chinese grammar automatic error correction system in B/S mode is implemented,which is applied to the Flask framework.Flask is a lightweight and specifiable framework.The Chinese grammar automatic error correction system constructed in this topic can realize the functions of loading custom confusion sets and turning off word granularity correction and other functions,which can correct common grammatical errors in Chinese text.
Keywords/Search Tags:Grammatical error correction, BERT, Transfomer, DNN
PDF Full Text Request
Related items