%0 Journal Article
%T 基于深度学习的语义级中文自动校对方法<br>A Semantic Level Chinese Automatic Proofreading Method Based on Deep Learning
%A 邓晨曦
%A 蒋一锄
%A 李合军
%A 彭姣丽
%A 刘曜端
%A 李凌云
%J Computer Science and Application
%P 1373-1381
%@ 2161-881X
%D 2023
%I Hans Publishing
%R 10.12677/CSA.2023.137135
%X <span style=\"font-size:14px;\">中文语法纠错任务是检查和纠正句子中的语法错误，相对于中文拼写错误纠正，中文语法错误纠正面对的错误不仅包括同音字和同形字的错误，还包括多字和少字的情况。本文通过大量的实验验证不同方法的优缺点，基于规则的方法需要消耗大量的人力来构造规则，而基于传统机器学习的方法面临特征提取能力不足的缺点，基于深度学习的方法是目前语法纠错的主要方法，因为语法纠错的文本存在不确定性，所以纠错的结果可能存在多种可能，因此Seq2Seq和预训练语言模型目前取得了较好的效果。</span><br />
<span style=\"font-size:14px;\">The task of Chinese grammar error correction is to check and correct grammatical errors in sentences. Compared with Chinese spelling error correction, Chinese grammar error correction not only includes homophone and homomorphic errors, but also includes redundant and missing characters. This paper verifies the advantages and disadvantages of different methods through a large number of experiments. Rule-based methods need to consume a lot of manpower to construct rules, while traditional machine learn-based methods face the disadvantage of insufficient feature extraction ability. Deep learn-based methods are the main methods for grammar error correction at present. Because there is uncertainty in the text of syntax correction, the result of error correction may have a variety of possible results, so Seq2Seq and the pretrained language model have achieved good results.</span>
%K 深度学习，中文语法纠错，Seq2Seq，预训练语言模型<br>Deep Learning
%K Chinese Grammatical Error Correction
%K Seq2Seq
%K Pre-Trained Language Models
%U http://www.hanspub.org/journal/PaperInformation.aspx?PaperID=68840