Analysis And Application Of Brand Reputation For Social Network

With the continuous growth of Internet users,the user opinion collection and brand reputation analysis based on social networks are one of the hot spots for big data applications.For a company,it is possible to obtain the current reputation of the company's products in the market by analyzing user comment data on social networks,to analyze users' concerns about product attributes and performance,to understand users' impressions of the product,and to grasp the right the overall attitude of the brand.The research on brand reputation analysis of social network data is not only an important research topic in the field of natural language processing,but also applied to the "digital marketing" mode of today's era,which has important application value.For the analysis of brand reputation,it is very important to identify the emotions expressed or revealed by consumers in product reviews.Therefore,the extraction evaluation object work is the foundation.Based on this,we complete the reputation analysis of the evaluation object.In order to carry out the reputation analysis from a more macro perspective,the paper analyzes the reputation of predefined aspect categories.The main work of this paper is as follows:1)An evaluation object extraction method based on sequence annotation is proposed.The evaluation object problem is modeled as a sequence labeling problem.We use two methods of general embedding and domain embedding for splicing as input vectors,and by menas of the convolution operations,and use bidirectional long and short-term memory networks to learn context features.Experimental results show that the proposed method can be more targeted and improve the accuracy of evaluation object extraction.2)A method of reputation analysis for evaluation object based on CNN and BiLSTM is proposed.The pre-trained Bert language model is used for word embedding,and the word embedding results are input into the bidirectional long-term and short-term memory network layer to extract the context information.The multi-head attention mechanism and point-wise convolution are used to fully learn the features of reputation evaluation in texts.The experiment shows that the method combining bidirectional long-term and short-term memory network and multi-head attention mechanism is more effective in emotion analysis.3)A multi-output classification method for aspect reputation based on BiGRU and Attention.The sentence is represented by word vector through word embedding technology,which is introduced into bidirectional GRU network to better learn the context information.Then the important information is captured by self-attention,and the data is classified by multi-output.The task of aspect classification and reputation analysis are completed at the same time.Experimental results show that the proposed multi-output classification method can complete the task better.