A Study On E-Commerce Deep Web Query Interface Mining

With the rapid development in the recent 20 years, E-commerce has a remarkable achievement. It changes the traditional commerce mode and pattern with an unprecedented way, and deeply influences the human understanding about the commerce. What's more, it changes the life and production mode of people. As the network size is rapidly developing world wide, the Web sites and the number of web pages on Internet are growing explosively. In the meantime, the continuous improvement of network techniques impels more Web databases can be visited through the network inquiry entrance, it is called Deep Web in academia. According to the relevant research institute and business organization, the Deep Web in World-Wide-Web storages tremendous data, and the E-commerce data accounts for a large proportion. These e-commerce data have the very high commercial value. However, due to the independence of web form and content development of these e-commerce websites, the web data presents isomerism. Consequently mining valuable information and data from the e-commerce websites automatically is considered as a very challenging task.In order to solve the problem that how to be understanded the deep web search interface by the computer semantically.The relationship among label and search element or among search element themselves in e-commerce deep web query interface was analyzed fully. Additionally, the vision feature in deep web query interface was analyzed from poisition feature, layout feature and appearance feature. The heuristic rules were summarized through observing a large number of e-commerce deep web query interface. A method that mining the deep web query interface semantic model based on the visual feature and WordNet was presented according to the above analysis. Firstly, the data in the deep web query interface page need to be dealt with beforehand. Then the visual block tree can be built using the VIPs, the label can be identified using the clustering. Finally the label tree can be built using the sementic relationship among the label in the deep web query interface, the label tree and search forms were matched according to the visual feature. The accuracy that mining the semantic information was improved using the method.In addition, data from UIUC website is used to compare the traditional algorithm's and the proposed algorithm's performance in accuracy and robustness. The experiments'result shows that the proposed algorithm played better in this two metrics. Finally, the limitations and shortcomings of the proposed method and the future study are described in detail.