| The explosion of online reviews brings both opportunity and challenge to the enterprises.The reviews are posted by consumers independently right after their buying,which can give enterprise quick feedback and an effective channel of communication.However,the online reviews is of great volume with a lot of low-quality reviews inside.Thus,it is of great challenge to make full use of the reviews.In this research,we try to mine the online reviews from an enterprise perspective,to mine customer requirement for product improvement.Our research is composited of four parts:(1)Explore the way of Chinese short text sentiment analysis.According to the analysis object,the sentiment can be inferred from different aspects,feature,sentence,paragraph and chapter.And by the length,we can divide the review into short text and long text.Online reviews are posted by thousands of consumers independently,which is short and flexible,especially in Chinese.In this research,we build the feature lexicon and sentiment word lexicon from the review text,and the language form text is translated into quantitative data for further analysis.(2)Construct the attributes of review helpfulness from an enterprise perspective.The abandon reviews is a treasure buried in the soil.To catch the essence of the online reviews,we need to measure the helpfulness.Our survey shows that there are significant difference between customer perspective and enterprise perspective towards the judgment of helpfulness.Enterprise try to get feedback from the reviews,especially the negative opinions,which is a good source of product improvement.Based on literature and a questionnaire investigation,three kinds of attributes is built to predict the helpfulness of online reviews.Empirical analysis finds that the helpful ratio is as low as 8%and helpfulness is negative.to the length and number of sentence.Compared with the quantity,enterprise care more about the quality of the reviews,which describe product features in detail.(3)Putting forward the method of prioritization of customer requirement based on conjoint analysis.Conjoint analysis is widely used in marketing,which estimates the structure of a consumer's requirements(e.g.part worth,importance weights,ideal points)given his/her overall evaluations of a set of alternatives that are pre-specified in terms of levels of different features.In this research,taking the pros and cons as the performance of features,and the sentiment of text as customer utility,part worth model is selected to infer the range of different product features and weight is calculated based on that.Our study of 55,000 helpful online reviews from JD.com shows that battery,signal,camera and appearance dominant other features,while price and edition account little for utility.(4)Putting forward the way of product improvement analysis based on Kano model.According to Kano model,different features act in different ways,which can be classified into one-dimensional features,must-be features and attractive features.Must-be features should be improved to the basic level and one-dimensional and attractive features should be improved under the limit of cost.Our study shows that battery is one-dimensional feature,function,music,price,screen,feeling,appearance,logistics and system are attractive features,the others are must-be features. |
PDF Full Text Request