Fashion Feature Extraction And Analysis Of Online Merchandise Based On Temporal Text

With the approaching of "new-retail" era,shopping online becomes increasingly common,and it has more and more effect on people's daily life.Shopping sites provide numerous information,as well as consumer's transaction records.Massive web information conceal loads of valuable messages.So it has high significance to study the problems.This paper studies the text description of online commodity as starting point.The temporal feature is also considered.The commodity's feature text representation method is defined.This paper extracts the general feature of commodity from temporal text by using text mining techniques,combines the sales data with text description.We calculate the popular index weight of the characteristic words,so that we can obtain fashion features of target category of goods.In order to reduce the data redundancy caused by the feature extraction process,this paper proposes a new semantic similarity calculation method of Chinese words for pruning operation.By using the HowNet ontology knowledge base,the information content(IC)is also taking account.The words deduplication based on semantic weight could greatly reduce the computational complexity.Furthermore,we propose a temporal feature model and algorithm for online commodity's popular features.To check the validity of the algorithm,a large number of real commodity transaction data and text descriptions are collected from the electricity platform,which were used as the experimental data sets.The numerical results are compared with the search words rank at the same time,which suggested that the method has higher accuracy and certain practicality.