This paper presents a multimedia data analysis system for evaluating "emotive effect" of advertisements in online video contents. This system evaluates effect of advertisements to viewers by referencing relevance between advertisement strategy descriptions and color-emotion definitions developed in the course of color psychology research. This paper proposes a hybrid feature space model for document data describing advertisement strategy and video data for actual advertisement. This model includes an extraction method for semantic linear features of colors with the spatial and temporal information from the entire stream of each video. This model also includes an extraction method for advertisement strategy by analyzing documents describing organizations' future directions. By using this system, an advertiser company evaluates how the video advertisement is relevant to the advertiser company's strategy. We show several experiments to clarify the effectiveness of this method.