METHOD OF RECOMMENDING TV PROGRAM USING TEXT INFORMATION

摘要

PURPOSE: A method of recommending a TV program using text information is provided to analyze preference programs of a viewer based on titles and outlines of programs that the viewer watched. CONSTITUTION: Keywords are extracted from a program instance and the frequency is updated for each keyword(SA1). The frequency is inputted to calculate keyword preference in a mutual information mode to weaken preference of keywords having excessive frequency. In the case that a keyword has low frequency, it is considered as an accident and a mutual information value is weakened. When the calculated keyword preference is inputted to calculate preference of a program, the program preference is normalized in a text length and calculated. The calculation result is provided to a recommendation engine as an input when a final program is recommended.