Published Paper Details:

In latest years, with the recognition of social media, customers are an increasing number of eager to categorical their emotions and opinions in the structure of pics and text, which makes multimodal records with textual content and photos the con tent kind with the most growth. Most of the statistics posted via customers on social media has apparent sentimental aspects, and multimodal sentiment evaluation has end up an necessary lookup field. Previous research on multimodal sentiment evaluation have principally centered on extracting textual content and photograph aspects one by one and then combining them for sentiment classification. These research frequently bypass the interplay between textual content and images. Therefore, this challenge proposes a new multimodal sentiment evaluation model. The mannequin first eliminates noise interference in textual statistics and extracts greater necessary photo features. Then, in the feature-fusion section based totally on the interest mechanism, the textual content and pics study the interior points from every different via symmetry. Then the fusion elements are utilized to sentiment classification tasks. The experimental effects on two frequent multimodal sentiment datasets reveal the effectiveness of the proposed model.