Discovering and Recognizing User´s Intention Based on Pro-patterned Extendable Network in Web Active Service

Discovering and recognizing user´s intention is a vital technique in Web active server, which can select proper services for meeting client requirements. User intentions depend upon external modes and internal modes of requester, so it is difficult to distinguish user´s intention with statistical methods and data mining under fixed computing modules. A PENN (Pro-patterned Extendable Neural Network) is introduced to discover and recognize user´s intention based on two policies: template matching and attention focus changing mechanisms. The structure of PENN can be adjusted by modifying number of latent layer´s node and/or increasing the pattern to meet discovery and recognizing needs, And select an optimum pattern output as teacher to train the other pattern for sharing knowledge each other, improving efficiency and precision of the PENN network in discovering and recognizing. The experiment and date analysis are shown that the PENN has self-adaptive and retractile features in discovering and recognizing user´s intention.