【作者】 王磊；

【导师】 包尔固德；

【作者基本信息】 北京交通大学 ， 软件工程（专业学位）， 2022， 硕士

【摘要】 随着社会的发展,人们越来越认识到学习对于未来发展的重要性,对孩子教育的投入也在逐渐增大。然而市面上火热的拍照搜题类辅助学习软件却可能对于孩子的学习起到反作用,孩子可能会直接把搜索得到的答案抄写,并不会去尝试学习如何解答。本系统的搜题解与传统的拍照搜题不同,传统的拍照搜题是从数据库中将原题和答案找出来直接展示给用户,而搜题解是从库中找出与搜索题目解题思路一致的题目并展示题解,用户只有通过学习题解才可以解决自己的问题,从而避免了抄答案等情况,达到举一反三的学习效果。此外,随着城镇化的发展,教育展现出了教育不平衡与不充分,教育资源主要集中于发达地区和富裕家庭,而对于贫困地区和家庭来说,教育资源相对缺乏,无法满足学生的教育需求,本系统可以缓解此类现象。为了解决以上问题,本文从小学数学入手,利用人工智能技术来开发一款新的辅助学习系统。在本文中,利用外部OCR技术、自研文本纠错技术、自研题解匹配技术实现了本系统最重要的拍照搜题解功能和文本搜题解功能。为解决文本纠错数据集稀缺的问题,设计了生成伪数据的构造规则,在模型训练时,先使用伪数据训练基模型,然后基于基模型在真实数据上训练最终模型,最后通过后处理的方式提高纠错准确率,使得准确率达到80.5%。题解匹配算法分为了召回和排序两阶段,分别使用了Sentence bert模型和Bert模型,其整体准确率达到84.5%,召回率达到74.3%,超出预期。此外,在题解匹配前,将文本纠错算法应用于OCR结果纠错,使得匹配结果得到有效提升。学生在课下做完作业,常常不知道自己是否做对,因此系统提供了人工在线批改作业,使得批改结果更可靠更人性化。同时,为了方便学生在课下对当堂课程进行巩固练习,系统提供了单元同步学习功能。此外,为了方便用户检索资料,系统还为用户提供了资料搜索功能并展示热门下载资料。通过以上功能构建了辅助学习系统,它不仅使得偏远地区的孩子也可以享受到和城里孩子一样的教育资源,而且也可以帮助学生解决部分学习上的问题。在本项目的实现过程中,使用了主流框架SSM搭建后台服务,Vant框架搭建微信小程序界面。本人主要参与了文本纠错算法设计与实现、题解匹配算法设计与实现,个人中心模块、作业管理模块与单元同步学习模块等模块的设计与实现,以及系统的测试工作。更多还原

【Abstract】 With the development of society,people are more and more aware of the importance of learning for future development,and the investment in children’s education is gradually increasing.However,the popular photo search question assisted learning software on the market may have a negative effect on children’s learning.Children may directly copy the answers they get from the search instead of trying to learn how to answer them.This system is different from the traditional photo search problem.The traditional photo search problem is to find out the original question and answer directly from the database and display it directly to the user.The search solution is to find out the topic consistent with the search question solving method and display the solution.The user can solve his own problems only by learning the exercises solution,so as to avoid copying the answers and so on.In addition,with the development of urbanization,education shows the imbalance and insufficiency of education.Educational resources are mainly concentrated in developed areas and wealthy families.For poor areas and families,educational resources are relatively scarce and can not meet the educational needs of students.This system can alleviate this phenomenon.In order to solve the above problems,this paper starts with primary school mathematics and uses artificial intelligence technology to develop a new auxiliary learning system.In this paper,the most important functions of the system are realized by using external OCR technology,self-developed text error correction technology and self-developed question solution matching technology.In order to solve the problem of scarcity of text error correction data set,the construction rules of generating pseudo data are designed.In model training,the pseudo data is used to train the base model,and then the final model is trained on the real data based on the base model.Finally,the error correction accuracy is improved by post-processing,so that the accuracy reaches 80.5%.The problem solution matching algorithm is divided into two stages:recall and sorting.Sentence Bert model and Bert model are used respectively.The overall accuracy rate is 84.5% and the recall rate is 74.3%,which is higher than expected.In addition,before the problem solution matching,the text error correction algorithm is applied to the OCR result error correction,so that the matching result can be effectively improved.When students finish their homework after class,they often don’t know whether they have done it right.Therefore,the system provides manual online correction of homework,which makes the correction results more reliable and humanized.At the same time,in order to facilitate students to consolidate and practice the current course after class,the system provides the function of unit synchronous learning.In addition,in order to facilitate users to retrieve data,the system also provides users with data search function and displays popular downloaded data.Through the above functions,an auxiliary learning system is constructed,which not only enables children in remote areas to enjoy the same educational resources as urban children,but also helps students solve some learning problems.In the implementation process of this project,the mainstream framework SSM is used to build the background service,and the vant framework is used to build the wechat applet interface.I mainly participated in the design and implementation of text error correction algorithm,problem solution matching algorithm,personal center module,job management module and unit synchronous learning module,as well as the test of the system.更多还原