An Integrated Model for Sentiment Analysis of Multimodal Inputs based on Artificial Intelligence Approach

Abstract

The rapid growth of internet and smartphone usage has led to a significant increase in user-generated content on digital platforms such as Amazon, TripAdvisor and Yelp. Users increasingly express their opinions through multimodal reviews that include text, images and videos. Analyzing these reviews is essential for understanding consumer sentiment and supporting decision-making in marketing, customer service and product development. However, traditional sentiment analysis approaches mainly rely on unimodal inputs, which limits their ability to capture the contextual richness of real-world reviews. newlineTo overcome this limitation, this research proposes an integrated multimodal sentiment analysis framework capable of processing unimodal inputs, text, image and videos as well as their combinations, including text-image, text video, image video and fully multimodal reviews. The framework analyzes each modality using dedicated feature extraction and classification models and then fuses the outputs to produce a unified sentiment prediction. Sentiments are classified into positive, negative or neutral based on computed sentiment scores. newlineBy supporting flexible modality combinations, the proposed system better reflects real-world user behavior and provides a more comprehensive interpretation of consumer sentiment. Experimental evaluation on benchmark datasets and real-world validation samples demonstrates the effectiveness and robustness of the framework across different modalities. The proposed approach bridges the gap between unimodal sentiment analysis and practical multimodal review analysis, making it suitable for real-world applications such as online review monitoring and customer feedback analysis. newline newline