Abstract: Automatic identification of deep Web entries is the basis of deep Web data integration. Owing to the subjec- tivity of form design,deep Web entries lack unified standard and it is difficult to judge whether the form is a deep Web entry by the definite rules. Based on the statistics, this paper first chose several form attributes as the defining features, which can distinguish searchable forms from non-searchable forms. Then, an entry identification algorithm was proposed by using neural network. Unlike previous approaches, neural network can be trained, which is very suitable for entry i- dentification of the deep Web. I}he experimental results show that our proposed algorithm can be an effective way in au- tomatic identification of the deep Web.

Key words: Deep Web cntrics,Ncural nctwork,Fcaturc cxtraction,Machinc learning