Research on Key Technologies of Preschool Video Site Searching System

This paper designs a preschool education-based video source searching system. It mainly focuses on multimedia resources extraction in web pages and multimedia noise filter in web pages to provide multimedia resource research service for people in preschool education field. We study TF-IDF function implemented in text classification, integrate relevant knowledge of information theory and further find out potential relationship of characterized characteristic items between the distributions among class and distribution information inside class. We further put forward an improved TF-IDF function used in text classification and introduce each step from training set arrangement to classifier evaluation in detail including difficulties we met and their solutions. By concluding a series of characteristics in non-thematic related resources we design principle-based video class noise filtration algorithm.