Design And Implementation Of Expert System For Cloud Product Information Consultatio

With the development and popularity of cloud computing,China’s cloud service industry is developing rapidly and the market is continuously expanding.Currently,cloud service products have become more sophisticated and diversified.To learn about a cloud product,users usually obtain information through search engines or manual consultation,but these methods cannot guarantee the speed and quality of information retrieval.At the same time,the existing cloud product information consulting system usually adopts the form of structured knowledge base to answer users’ questions,which has a large limitation in the coverage of questions,and the construction of structured knowledge base needs to consume a lot of human cost.It is a great challenge to integrate and manage the existing cloud product information and provide users with comprehensive and efficient cloud product information consulting services with the help of more advanced technologies.For the above background,the specific research work of this dissertation is as follows:(1)Construction of a knowledge base for cloud product documentation.In order to integrate and manage various existing cloud product information,the product documents on the official websites of cloud service providers are crawled and processed through the Scrapy framework,and stored in the Elastic Search.To address the problems of small extraction granularity and easy splitting of key information in traditional keyword extraction methods,a multi-feature fusion keyword extraction algorithm is proposed to calculate the importance of double-word phrases by internal closeness and contextual relevance features,and fuse and rank the results with locationweighted keyword extraction.The experimental results show that the algorithm extracts keywords with higher accuracy and recall,and is more consistent with the manual tagging results,laying the foundation for the subsequent coarse recall of documents in the question answering algorithm.(2)Design of question answering algorithm based on multi-document reading comprehension.In order to solve the problems of effectiveness and efficiency of multi-document reading comprehension,a combined coarse recall and fine recall question answering algorithm is proposed.In the coarse recall phase,Elastic Search is used for fast recall of answer-related documents to narrow the answer search space.In the fine recall phase,a multi-document reading comprehension model is proposed for answer extraction,which introduced a co-attention mechanism to enhance the interaction between questions and documents,used a joint multi-task training approach to achieve two-way feedback between tasks,and improves the robustness of the model through hierarchical fine-tuning and hard negative sampling.The model’s robustness is improved through hierarchical fine-tuning and negative sampling of difficult cases.The experimental results show that the recall rate of the document coarse recall method can reach more than 98%,and the reading comprehension model outperforms the baseline model in both EM and F1 indexes,which provides algorithmic support for the implementation of the intelligent Q&A function of system.(3)Design and implementation of an expert system for the consultation of cloud product information.Firstly,to ensure the rationality of the system functions,this dissertation analyzed the functional requirements from various perspectives,including those of general users,expert users,and system administrators.Secondly,it identified the five functional modules of portal management,intelligent question and answer,expert answer,extended service,and management and maintenance.Finally,based on Bootstrap,Django and other technologies,it realized a cloud product expert system with intelligent question answering and diversified service modes.It was shown from the results of the system test that the system performed well and achieved efficient question-and-answer functions,as well as perfect background management functions.Therefore,this system has more application value in the consultation of cloud product information.