Product Requirements Mining Based On Online Reviews

Product requirements mining is an important task for enterprises in the front-end of product innovation.In the increasingly developing Internet environment,online reviews have become an important source of product demand research.Some problems in processing reviews are gradually overcome by technical means.At present,the research of product requirements mining based on online reviews mainly includes the research of product aspects and viewpoint recognition,sentiment classification,viewpoint clustering and summary,which have obtained some achievements in their respective fields,but there are few studies on the whole process to handle online reviews to get product requirements.At the same time,in each individual field,the existing algorithm is not ideal enough.In this thesis,considering all the research of text mining,we put forward a whole process of requirements mining,which includes the product aspects and viewpoint recognition,sentiment classification and the viewpoint clustering and summary.Based on this process,improved algorithms in each step are given and tested by data experiments.In the phase of product aspects and viewpoint recognition,the algorithm based on rule matching is adopted.In the sentiment classification stage,it is implemented based on constructing the affective dictionary and designing the corresponding sentiment score matching algorithm.In the point of view clustering,we first construct the product aspects dictionary,make the synonym clustering for all the product aspects,then analyze the product feature class on the basis of clustering,and take the histogram and scatter graph as the final way of requirements presentation.Finally,the improved algorithm which is validated by data experiment in each process is combined to form a complete product demand mining model based on online reviews,which provides a prototype for future product requirements mining system based on online reviews.This research shows that the algorithm of product aspects and viewpoint mining based on syntactic rules can achieve similar effect with the machine learning algorithm while the artificial cost is lower.The algorithm based on sentiment dictionary has good effect on the structured tags.The requirement expression based on the aspects class can make the product requirements more concentrated.Product requirements mining based on the online reviews can provide a practical and reliable channel for the enterprise to find users' needs.