Short Text Mining Research For B2C E-Commerce Platform

With the rapid development of e-commerce,there have been many e-commerce platforms in China,and shopping online has become a daily shopping mode for people.In the big environment of e-commerce,platforms and merchants have gained many development opportunities,but as the market continues to expand and market competition continues to intensify,platforms and merchants will not be able to see issues existing in their own products in time,which will cause they cannot meet the purchasing demand of buyers.And that leads to affecting sales,eventually eliminated by the market.Users face the problem that they have difficulty in choosing a wide variety of goods on the Internet with uneven quality.As a representative online shopping platform,B2 C e-commerce platform is popular among online shopping users.With the occurrence of consumer's behavior,there are a large number of valuable users' information showing up in the B2 C e-commerce platform,such as labels tagged by users in comment system and short questions and answers in interactive system.It is meaningful to dig and analyse these short texts with both characteristics and abundant consumer's behavior information.This paper studies the short text information in the e-commerce platform by text mining,and excavates valuable information to provide intelligence data for e-commerce platform,merchants and users.This paper uses Jingdong,a representative B2 C e-commerce platform in China,as an example.This paper selects two short text information with characteristics in the platform to dig and analyze,one is user's comment labels,another is questions and answers about Jingdong's commodity.In the excavation of short texts of user's comment labels in Jingdong platform,this paper proposes a method for tagged short texts through external word vector expansion.The external word vector is obtained by training the external corpus with Word2 Vec word vector model,and then using clustering analysis for the extended label and short texts.Compared with the traditional vector space model method,the external word vector expansion method can solve the problem which the traditionalmethod can not.It has a better clustering effect for the synonymous tags.In addition,use association rules mining and semantic network analysis for comment tags to find out the interconnectedness among tags and the interconnectedness among the words in tags.In Jingdong's question and answer short text mining,extract the related product features first.In order to ensure the accuracy of feature extraction,the extraction process combines noun extraction,frequent item extraction,synonym merging and other methods,and classify and analyze the extracted product feature words,then use the extracted commodity's feature words as seed words to extract the corresponding words of viewpoint and analyze it by co-occurrence,after that,trained the question and answer data by the Word2 Vec word vector model.Explained the short texts of commodity's question and answer by weighted method.Finally use the meliorated initial clustering center K-Means clustering algorithm to cluster the question short text vectors and analyze the clustering results.