Resrarch And Implementation Of Spam Review Detection System Based On Nerual Network

With the continuous innovation of the Internet model,the acceleration of the integration of online and offline services,and the accelerating pace of onlineization of public services,e-commerce and professional review websites are rapidly developing,and related product or service reviews are playing an increasingly important role in daily life.Spam review not only affect the normal decision-making of consumers,but also have a negative impact on the business decisions and service direction decisions of merchants and service departments.Spam review recognition is an area of natural language processing,which is of great significance for maintaining the healthy and stable development of online commentary environment.This thesis is mainly based on neural network to research and identify spam review.In this thesis,the problem of the processing of each granularity text and the low recognition rate in the existing spam review recognition is studied,and a new spam review recognition method model is proposed.Specific research results are:1.This thesis constructs a multi-layered spam review recognition model based on various neural networks and multiple attention mechanisms.In view of the fact that the existing model does not consider the chronological semantics and the semantic coding of the sentence without using the attention mechanism,this thesis proposes to use Bi-directional Long Short-Term Memory combined with the Inner-Attention mechanism to encode the sentence semantics,so that the sentence feature vector is generated by the word.Then,using the Bi-directional Gate Recurrent Unit with simpler structure and shorter training duration,the semantic information of the sentence feature vector is encoded by the Inter-Attention mechanism proposed in this thesis,and the text feature vector is generated.Finally,the rationality and effectiveness of the model are verified by comparative experiments.2.Based on the above model,this thesis designs and implements an online spam review recognition system based on related Web technologies such as HTML,JavaScript and Flask,and packages the system into Docker.In this thesis,the overall architecture and modules of the system are designed in detail,and each module and key interfaces are explained in detail according to the application scenarios of the system.