基于深度学习的智能中文文本检校方法
|
摘要
基于新华社在用商业校对软件的日常使用情况,以及采编业务需求两方面信息,针对商业校对软件应用中存在的不足,提出了一种基于深度学习的智能文本检校方法。该方法利用自然语言处理技术,特别是句法分析、语言模型等技术对用户的输入文本进行深度分析和挖掘,并引入语义分析技术,通过计算句子中各词语的语义相关程度来提高错误识别率,有效减少了采编人员的工作量,并提高了稿件的检校准确率。
Based on the daily usage of business proofreading software and relative demands from journalists and editors, this paper proposes a Chinese text proofreading method based on deep learning framework. In this method, natural language proceeding technologies such as syntax parsing and language model are used to conduct depth analysis and mining on user texts. Followed by semantic analysis technology, this method enhanced its error recognition rate by calculating the semantic relatedness among words and phrases in sentences, thus to cut the workload for editors and enhance the accuracy of text checking.
Based on the daily usage of business proofreading software and relative demands from journalists and editors, this paper proposes a Chinese text proofreading method based on deep learning framework. In this method, natural language proceeding technologies such as syntax parsing and language model are used to conduct depth analysis and mining on user texts. Followed by semantic analysis technology, this method enhanced its error recognition rate by calculating the semantic relatedness among words and phrases in sentences, thus to cut the workload for editors and enhance the accuracy of text checking.
引文
[1]韩文颖.面向问答的中文语法错误自动检测方法研究[D].哈尔滨工业大学,2015.
[2]人民网.新华社公布规范用语[EB/OL].(2015-11-05)[2017-08-03].http://culture.people.com.cn/n/2015/1105/c87423-27782404.html
[3]王维智.基于特征提取和特征选择的级联深度学习模型研究[D].哈尔滨工业大学,2015.
[4]马金山,刘挺,李生.基于n-gram及依存分析的中文自动查错方法[C]//Advances in Computation of Oriental Languages--Proceedings of the,International Conference on Computer Processing of Oriental Languages.2003.
[5]万家资讯.“奥马巴”和“奥巴马”傻傻分不清,多家媒体负责人受罚[EB/OL].(2015-12-18)[2017-08-03].http://365jia.cn/news/2015-12-08/E52F49B964AEC41A.html
[2]人民网.新华社公布规范用语[EB/OL].(2015-11-05)[2017-08-03].http://culture.people.com.cn/n/2015/1105/c87423-27782404.html
[3]王维智.基于特征提取和特征选择的级联深度学习模型研究[D].哈尔滨工业大学,2015.
[4]马金山,刘挺,李生.基于n-gram及依存分析的中文自动查错方法[C]//Advances in Computation of Oriental Languages--Proceedings of the,International Conference on Computer Processing of Oriental Languages.2003.
[5]万家资讯.“奥马巴”和“奥巴马”傻傻分不清,多家媒体负责人受罚[EB/OL].(2015-12-18)[2017-08-03].http://365jia.cn/news/2015-12-08/E52F49B964AEC41A.html