ASIF: attention-based sentiment inquiry framework for profound product recommendations

Online product recommendation has gained much popularity in recent years and has become the most demanding research area that can help consumers make better purchasing decisions. Recently, many machine learning techniques have been tested on various datasets for analyzing customer sentiments through online portals. Still, customers have difficulty finding profound products due to a lack of depth-level recommendations. The existing models for product recommendation may rely on either text or image reviews and ignore the multus-medium based reviews that lead to a poor recommendation. Furthermore, the recommendation system does not properly utilize product ranking. To effectively analyze the sentiment of online products, a novel framework ASIF is suggested in this manuscript. The key steps of the proposed framework are multi-modal data collection, normalization, text and image-based feature extraction, two-level feature fusion and extended transfer learning-based recommendation at binary level and multilevel. Five different datasets have been used for the evaluation of ASIF. From the experimental analysis and comparison to the baseline methods, it has been observed that the accuracy, precision, recall and F-Score of ASIF is far better, giving 95.95% and 94.95% on the standard dataset.