| The Internet of things(IoT)technology is an important research direction in the fields of pervasive computing and information technology.Its applications include various scenarios such as smart cities,smart logistics,smart wear,and smart homes.With the improvement of the microelectronics technology and computer software,smart home systems and various smart devices have gradually entered people’s daily lives.By providing different types of services,they have gradually become good assistants and necessary choices to improve the user life quality.At the same time,artificial intelligence(AI)algorithms represented by machine learning and deep learning have achieved unprecedented development in image,text and other fields,bringing new opportunities for further development of smart home systems.Through the unremitting efforts of scientific researchers,the system and technology of command rationality detection have been developed and advanced in all aspects.However,there are still many problems and challenges.For example,the existing command rationality detection methods mainly occupy environmental parameters as intermediate variables,which leads to the existing ways that detect the command rationality of smart devices being not comprehensive,low in automation,and insufficient in flexibility.In addition,the existing methods do not combine the latest research results of AI such as deep learning.The original methods of constructing conflict dictionaries and machine learning have a lot of disadvantages such as high labor costs and complex processes,resulting in high system redundancy.This thesis combines the actual application requirements of the smart home system.In view of the above problems and deficiencies,the following aspects are studied:(1)Propose a detection system for the rationality of smart home system commands based on natural language processing(NLP).The system uses the knowledge of the knowledge graph(KG)as the source of information to learn and analyze the functional sentences of various smart devices,and.uses NLP related technologies to determine the command conflicts between various functions.The proposed system defines the conflicts of various smart devices which can be also used to analyze and detect the rationality of commands related to smart devices.(2)Extending from the above-mentioned key technologies of text classification,this thesis studies opinion mining as well,which is one of the sub-problems of text classification,and proposes a spatiotemporal multi-feature extraction opinion mining algorithm based on simple recurrent unit(SRU),attention mechanism,and dilation convolution.The algorithm is also explained from the four language levels of words,grammar,semantics,and opinions according to the characteristics of natural language.(3)In order to verify the proposed system and algorithm,the web crawler technology is used to process massive network resources,and the functional statement text resources for training text classification neural network model and the command set for system verification are localized to form a data set,which is convenient for the further research.Through the research,this thesis provides a practical method for the command rationality detection and system application by occupying deep learning algorithms,which also broadens the application scopes of the command rationality system.The experimental results thoroughly show that the proposed system can take advantages of the related technologies of NLP to effectively monitor the rationality of commands in the background of smart home.The system framework can be applied to many fields and has broad application prospects.In addition,the opinion mining algorithm proposed in this thesis is verified by the Internet Movie Database(IMDb)public data set,all evaluation indicator values(e.g.accuracy,recall,precision)of the algorithm have been improved.It has large application and research significance in both fields of text classification and command rationality detection system. |
