Research On WEB Information Extraction Method Based On Label

In recent years,with the development of society,the network has become an indispensable part of people's life.And with the development of e-commerce,online shopping has become a very common social phenomenon.At present,most of the current shopping site generated pages dynamically based on the requests from users.However,the data displayed on the web page is difficult to be used directly.So it is very important to extract the information efficiently and accurately.This paper firstly introduces information extraction.Also it gives an introduction to the problems during extraction.By introducing some popular extraction methods,it compared their advantages and disadvantages.Considering current methods can't find the data area accurate,this paper makes full use of page features and combining two heuristic rule to make decision.When split the data records,a new algorithm was proposed.The algorithm makes full use of the label information,and gives weight to all the nodes.So,the similarity between trees can be measured by the weights of the nodes at the same level.This algorithm makes full use of the features of list page,avoiding plenty of calculation of the similarity between trees.Eventually,sequence alignment was introduced to complement the missing property.It needs to generate a template record by doing iterative multiple sequence alignment,and then compare the template record with all other records so as to adding the missing property of each record.By doing so,we finish the extraction.The experiments show that our method is highly automatic and can effectively reduce the impact of noise.The extraction system have a high level at recall and precision.