SENTIMENT-BASED COLOR SCHEMES: USING AI TO PREDICT VISUAL APPEAL

Abstract [English]

Sentiment analysis Sentiment analysis and computational color modeling (SACM) represents a radical new method of visual attractiveness prediction in digital media, advertising, design, and user-centric visual systems. The paper presents the idea of using artificial intelligence to trace the emotional signs identified in the text, image, and user interactions onto the color palette that maximizes aesthetic appeal and psychological effect. Based on the progress in the area of sentiment detection, emotion lexicons, and deep learning, the study suggests a robust model, which would associate the affective indicators with multidimensional color features, including hue, saturation, and brightness. The information used to train convolutional neural networks and transformer based models to extract color features, emotional valence, contextual embeddings and preference indicators is a multimodal data set of annotated images, user feedback, and text datasets (sentiometric) rich in sentiment. The approach to methodology focuses on holistic strategies of the feature engineering and model fusion, which incorporates the visual descriptors together with linguistic sentiment vectors. Experimental analysis incorporates data preprocessing routines, hyperparameter optimization and cross-architectural comparative benchmarking. Findings indicate that hybrid CNNTransformer pipelines are more accurate in their ability to predict aesthetically pleasing color schemes and have high correspondence to human emotional perception. The quality of recommended palettes is further confirmed by user studies, which show some form of uniform enhancement in the perceived harmony, emotional relevance, and design usability.