Method Research And Its Application Of Product Usability Analysis Based On User Comments

With the rapid development of mobile Internet era and electronic commerce,the network has greatly changed people's life and expression style.Shopping online has become one of the main consumption method resulting in the massive user comments on the electronic business platform or forum which become the most direct and effective way for enterprises to know the market and consumers' attitudes.mining customer attitudes from the user generated reviews And analysis the availability of product features to provide guidance and basis for the enterprise to develop product improvement and innovation programs,At the same time,from the public opinion to understand the user needs of the method also makes the company's research costs have been controlled.High quality reviews is extracted from the review text,then using text classification method,sentiment analysis,sequential pattern mining to get the information of customer potential demand and product usability from user comments.The main contents of this paper are as follows:1.A method of obtaining high quality of review text is proposed to calculate the features' score based on semantic and sentence environment.Sentence's word and word order similarity algorithm is used to identify duplicate comments or similar comments to avoid redundant information,And word similarity algorithm is used to identify independent of the target product reviews,and the comment time and purchase time difference is calculated to identify false comments.Compared to other method our method shows more advantage in identifying the useful review.2.Bi LSTM-CRF model is constructed to find the mining sequence pattern to extract product feature and segment the multi emotional strength of sentence to single emotional strength review;COBW model is trained separately to obtained sentence vector model and word vector model through which we can get word vector and sentence vector.Then cosine similarity between the word vector and the sentence vector is calculated to find the sentence' most similar emotion word,then the emotional words' emotion score is the product feature in the single emotion intensity sentence's emotion score.3.The feature's emotion score and the frequency of being reviewed time is regarded a normal distribution and the data we can get from the review text is taken as a sample.Estimating the overall distribution through the sample data by Bayes,establish product usability evaluation function mainly taken customer's satisfaction into consideration.A method is Designed to find out the key features which have the bigger influence on product usability according to two eight laws and define the method to calculate the usability factor.Finding out the best order to improve the product's usability quickly,facilitate enterprises to improve product design plan for reasonable control.