Research On The Quality Evaluation Model Of Online Reviews Based On The Text Analsisy

With the rapid development of the network shopping market,and related diversity and convenience of shopping platform and application,the online shopping brings great convenience to people's life,more and more people begin to accept and choose this way of life.However,due to the network goods is virtual or untouchable that people can't perceive the quality of product what they want to purchase,a lot of people tend to rely on goods online reviews to make purchasing decisions.While,this situation breeds some unscrupulous businessmen who create a large number of goods comments through various means,such as "high praise cashback",which not only increase consumers the time cost of screening comment,also may cause unnecessary economic loss.Therefore,it becomes the new hot topic about how to quickly identify the high quality ones in the field of online comment.This research mainly embarks from the content of online review,first is to extract feature what factors affecting the quality of online reviews,then build online reviews quality evaluation index and model,finally performance validation of model.The concrete content includes the following five sections:(1)Effectiveness annotations of the comment text.By improving the automatic labeling algorithm based on the length and K-means algorithm,this paper puts forward the Lk-means algorithm to label effectiveness on the comment text,and extract the effectiveness indicator;(2)Feature extraction.Divide online reviews data into two categories,numeric and textual,extract score data from numeric review,extracting such four indicators from text comments,readability,topic relevance and coherence,and complete index from the combination;(3)Construct quality evaluation index system of online reviews.According to the classical WRC information quality evaluation indexes and the improved study found 1R3C data quality evaluation index,and connecting with the features extracted six factors,then put forward the 1W2R3C evaluation index of the present study;(4)Set up the online reviews quality evaluation model.First is to set up online reviews quality evaluation model according to the evaluation index,then divide review data into train set and test set,and use the train set for calculating the weight of each evaluation index in the model and use test set to validate model performance.(5)Performance verification of model.The model performance validation will be conducted in two ways:one is to use the proposed 1W2R3C evaluation indicators,and classical WRC and 1R3C index,make contrast analysis on their modeling;the other one is,based on the index weight of training this model,respectively achieve index weight with introduction expert scoring method and grey correlation correction method,then compare analysis on their modeling;Thus fully validate the good performance of model.In this paper,the research results about online reviews quality evaluation model,which provides some new methods and theoretical basis for further study of online reviews content;and can also provide the corresponding decision support for the general consumers in practice.