Customer Needs Mining Based On Product Performance Lexicon And Opinion Analysis In Product Development

In the process of product development,customer needs from the user domain determine product positioning,which is the first step to the success of the product.With the rapid improvement of manufacturing productivity,customer needs now have the characteristics of rapid change,high degree of personalization,and various forms of expression.Traditional customer needs research methods have been unable to meet the current market due to its low efficiency and high manual intervention.In the era of big data,how to use data-driven means to efficiently and objectively acquire and analyze customer needs has become an urgent problem in product development.Therefore,the methods of customer needs acquiring and analysis were studied,and the method of customer needs modeling driven by review data was proposed.Firstly,in order to identify product performance,the object of comment in product reviews,the product performance lexicon was put forward.Based on matching rules,product performance words can be accurately identified.In order to create the lexicon,a product performance word extraction index with combined features was proposed.According to the characteristics of online product reviews,this index integrated degreecentricity feature,dependent syntactic feature and domain feature,and can be used in target words extracting.Through the experiment on the mobile phone product review data,it has been proved that the index can effectively distinguish the product performance words from the interference words.Secondly,opinion mining from product reviews was studied.According to whether containing evaluation object,the opinion sentences in reviews were divided into explicit opinion sentences and implicit opinion sentences.For explicit opinion sentences,an explicit opinion mining method based on matching rules was proposed.This method used the strategy of fully mining and opinion pruning,hence the dependency syntax rules and the pointwise mutual information supplementary rules were designed to fully mine customer opinions,and then improve the precision by opining pruning.For implicit opinion sentences,an implicit opinion mining method based on semantic similarity was proposed.This method defined explicit and implicit opinion sentences that express the same opinion as conjugate opinion sentences.Implicit opinion mining can be done by finding the explicit conjugate of implicit opinion sentences.Experimental results showed that the above two methods were effective.Finally,in order to establish the customer needs model,a K-means viewpoint clustering method based on Word2 vec word vectors was designed to summarize the mined customer opinions and gain customer needs.Through improving the classic Kano model,the product review based Kano analysis method was proposed for qualitative and quantitative analysis of customer needs.To normatively express the customer needs model,the matter-element theory was introduced.At last,in the experiment,a customer needs model was established based on a mobile phone product reviews,which also proved the feasibility of the above method.