Design And Implementation Of Chinese Composition Labeling Platform

With the development of the Internet industry and the popularization of smartphones and tablet computers,online education models have emerged,and major online education platforms have sprung up like bamboo shoots after a rain.Various education and training groups have taken the lead in launching online education models and expanding online users.In the popularization of online education,artificial intelligence technology has also played an important role,providing applications such as intelligent review assignments and online Q&A.In K12 online education,intelligent auxiliary tools for subjects such as English and mathematics are becoming more and more mature,while Chinese subjects still need to be improved,and the automatic correction of Chinese composition is the most important thing.The quality of the algorithm model is often related to the quality of the training set,which requires a large amount of labeled and meaningful data,and the internal Chinese composition of the company often exists in the form of pictures,and there is no convenient and fast labeling tool.There is no unified solution for data set,tasks and labeling results,causing problems such as low human efficiency,uneven quality,and high communication costs.In response to the above problems,a Chinese composition labeling platform came into being,providing project management,task management,data set management,personnel management,image recognition and other functions,coupled with efficient and fast labeling tools,to enable automatic composition correction services.The Chinese composition labeling platform provides a one-stop solution for composition labeling.After the OCR results of the composition pictures in the data set are manually corrected,then they are labeled by the annotator,and finally automatically synchronized to the algorithm team.Task review mechanism,label quality inspection strategy,message reminder and other functions,control the label quality,and generate subdivided label results.Through analysis,the system is divided into several major functional modules: system management module,recognition module,labeling module,review module and general service module.Here,the design and implementation of each part are expanded in detail in accordance with the software engineering life cycle sequence.I am mainly responsible for the development of the system management module,recognition module,labeling module,review module and general service module in the project.The system uses a front-end and back-end separation architecture,using Vue,Element UI and the open source rich text editor Quill to build the front-end,the back-end adopts a micro-service architecture,and Dubbo+Nacos+Spring Boot is used to build the back-end system.The system uses My SQL,Redis and Ali.Cloud OSS as data storage,uses Kafka as message queue,complete the construction of the entire system.The functional modules of this project have been developed and put into use,helping the team to efficiently complete the labeling tasks within the company,providing a large amount of alternative labeling data,and helping the algorithm progress.