The Enrichment Of MVSA Twitter Data Via Caption-Generated Label Using Sentiment Analysis

Due to the widespread use of Twitter data and the diversity of scientific research, it is a goldmine of data. Besides the emergence of new areas, including visual sentiment analysis, it was necessary to develop datasets with distinct requirements that would meet the purpose of this area. To compensate for the lack of annotated data, data enrichment is commonly used to generate new instances by leveraging newly available information or relying on other types of knowledge. This study will enrich and augment a dataset of images and captions with a sentiment analysis-generated label. A method has been put forth to determine the ideal sentiment. This technique uses the caption feature to implement six different sentiment analysis techniques. Finalizing the classification of each dataset item based on the percent of positive, negative, and neutral polarities extracted for each caption has been determined. Apart from that, the classification will have the largest percentage of polarities. Results from experiments run on the Multi-View Sentiment Analysis (MVSA) Twitter dataset, which contains pairs of images and textual image captions. All images within the dataset will be labelled, introducing a new feature to this dataset. The polarity of the generated sentiment is identical to the polarity of the objective text perspective. Note that the results were exceptional in every way.