| E-commerce is a new cross-discipline that integrates computer science,marketing,management,law and modern logistics.After years of development,the current development speed of e-commerce in China is obvious to all,and has become the leader in global e-commerce.With the rapid development of e-commerce and the rapid increase in the number of e-commerce platforms,the e-commerce information is becoming richer 'and the number of web pages is also exploding.It is difficult for consumers to compare products between different e-commerce platforms.Based on the above background,in recent years,the social demand for cross-platform e-commerce information retrieval and intelligent recommendation systems has become increasingly strong.The construction of cross-platform e-commerce information retrieval and intelligent recommendation system involves real-time online crawling of heterogeneous e-commerce platform data,information reconstruction and intelligent product recommendation based on user selection preferences.Based on the research progress in related fields at home and abroad,this article takes the mobile phone sales webpages of multiple e-commerce platforms as research examples,focuses on how to implement cross-e-commerce platform mobile phone sales information search and intelligent recommendation,and focuses on the theme of information extraction and intelligent recommendation.The technology has carried out related research,the main research content and research results are as follows:(1)Aiming at the problem of cross-platform product information extraction,this article uses the keyword-weighted Shark-PageRank algorithm to determine the topic webpage queue.In this process,the topic is described by adding a dynamic topic library generation method,so that the topic crawler has a better accuracy than Shark The PageRank algorithm has been further improved.On this basis,a template-based method for automatic extraction of web page information is proposed.This method first uses the topic information positioning method based on structural semantic entropy to locate the attribute information of the products to be extracted in the topic web page,and then summarizes the extraction path to construct Template library,and finally use the template library to quickly and accurately extract product information.The experimental results show that the template-based web page information extraction method proposed in this paper has a certain improvement in the accuracy of the web page information extraction task and the extraction time.(2)For the problem of smart product recommendation,from the perspective of sentiment analysis of product reviews,this article first extracts product subject words from user reviews through the LDA theme model,and then focuses on the location of different words in the review text and different parts of the review text.The contribution degree of the emotional tendency judgment under the theme is different.A Bi-LSTM+mixed attention mechanism model that integrates the feature of the subject word is proposed to calculate the emotional value of user reviews under different themes,and finally integrate the emotional value of all reviews of the product.The intelligent recommendation model is used to calculate the sentiment value of commodities under a specific theme,and the sentiment value of commodities under different themes is used as a recommendation basis,and the smart recommendation of commodities is implemented in combination with user selection preference features.The experimental results show that the Bi-LSTM+hybrid attention mechanism model that combines the features of the subject words proposed in this paper has further improved accuracy,recall and F value compared with the previous method.(3)Based on the above research results,combined with the Web front-end development technology,this paper has carried out the prelim inary design and implementation of the prototype system for cross-platform information search and product intelligent recommendation. |
PDF Full Text Request