Research On Product New Requirements Discovery Based On Chinese Online Reviews

The era of web 2.0 has created an opportunity for consumers to express their own opinions and ideas freely on the Internet.The comments on all kinds of things on e-commerce and social platforms(such as micro-blog,QQ etc)provided by consumers have formed an important cyber source,that is online reviews.As an important user data,online reviews are consumers' evaluation of products or services based on their own experience,which contains a wealth of valuable information.Producers can clearly understand their feelings or attitudes towards products from users' online reviews,discover some new requirements of the product,so as to improve products and increase the competitiveness of products in the market.Compared with the traditional discovery methods of new requirements of the product,such as market surveys,user interviews,etc.,online reviews have the advantages of low costs and high efficiency.In view of Chinese online reviews,this thesis carry out related researches on the discovery methods of new requirements of the product,aiming at providing better supports for enterprises or producers in product improvement and innovation.There are two problems in the process of finding new requirements of the product by online reviews.Firstly,the exponential growth mode makes the number of reviews huge.How is manual participation in the mining and analysis of online reviews reduced by means of machines.Secondly,some spam and fake reviews can seriously interfere with the accuracy of new requirements of the product.In view of the above problems,this work includes the following aspects.1.Analyzing the online reviews and extracting the main information--feature-opinion pair.The grammatical and semantics analysis of the text of the online reviews is carried out to obtain the dependency syntactic relation between the product features and opinions.The extraction rules of feature-opinion pairs are constructed by means of artificial methods.Combined with LTP(Language Technology Platform)system of Harbin Institute of Technology,then word segmentation,part of speech tagging and dependency parsing are carried out for the online reviews and the analysis results are stored in the form of the XML file.According to the extraction rules and some parsing methods of the XML file,the feature-opinion pairs are extracted automatically.2.It proposes a method for getting the new requirements of the product based on user's emotion.Extracting feature-opinion pairs based on the dependency parsing relations,clustering and counting the feature-opinion pairs to get the value of product feature attention.Considering the impact of related modifiers on opinion words comprehensively,the value of viewpoints is calculated.Obtaining the new requirements according to the value of attention and viewpoints.In addition,recognizing the generic noun phrase in the reviews and acquiring to new requirements which is specific.3.It proposes the concept of mainstream feature-opinion pairs based on big data thinking and use the mainstream feature-opinion pairs as the basic unit of the trustworthiness of reviews to construct a ranking model of the online reviews credibility.Based on the credible and useful online reviews,the user sentiment is quantified fuzzy and combined with KANO model for conversion and evaluation,the new requirements are found from the perspective of product improvement.Some online reviews are obtained by the crawler software,and the the experimental analysis is carried out on the new requirements of the product.The experiment results prove the validity of the methods.This thesis proposed some methods,which can effectively expand the existed approaches of new requirements of enterprises or producers,reduce the cost of requirement elicitation,improve the efficiency and help them to make product improvements and innovations.