Design And Implementation Of Online Brand Evaluation System Based On Sentiment Analysis

With the development of globalization,brand plays an important role in the competitiveness of enterprises,how to quantify the value of the brand in order to improve the competitiveness of enterprises is something worth thinking about.The traditional method of brand evaluation is to analyze the brand by means of questionnaires or statistical data such as profit and sales.However,this approach can not be a comprehensive analysis of hundreds of millions of consumers on the brand's views and attitudes,but also can not be intuitive to show the views.The online brand evaluation system based on sentiment analysis is a software to help enterprises to adjust the production by extracting large amounts of consumer brand attribute data view and analyzing consumer attitudes.Sentiment analysis is to judge the attitude of the text,the attitude is divided into praise,derogatory attitude,neutral etc.The main research methods are unsupervised learning and supervised learning.Unsupervised learning model by using unlabeled samples and the test result depends on the feature set.Supervised learning method uses machine learning method to train the labeled sample,and the advantage is strong scalability.The supervised learning method is used in this paper,however,different from the commonly used feature selection techniques,this paper uses the method of fusing the features.The main contributions of this paper:1)Two methods are used to extract features in this paper,and then compare and analyze the results of the classification of the two methods.One approach of extracting features is recognizing name entity;Another method is fusing the shallow and deep features.The experimental results show that the feature extraction method based on the fusion of the depth and shallow feature makes better performance evaluation than the feature extraction method using CRF only.The classification results of two classifiers,support vector machine and logistic regression are compared.Finally,support vector machine is selected as the sentiment classification model.After identifying the emotional attributes of attributes,each attribute of the brand can be given different weights to calculate the brand index,then compared with the brand score calculated by the questionnaire,and continuous optimization to select a weight system which has a higher degree of similarity with the questionnaire score,to be used in the brand evaluation systeme.2)This paper designs and implements a system of B/S Architecture which can be commercialized--an online brand evaluation system based on sentiment analysis.The system establishes the mapping relations between the brand's comment data and attributes,statistics of each index of brand attributes,the higher the rate of praise,the higher the score,and the lower score of the property is the point of concern and optimization;The attributes are given different weights,and the brand index is calculated by weighting,and then rank the brand according to the index.The core modules of the system include data collection module,data preprocessing module,emotion classification module and brand evaluation module.