Multimodal Sentiment Analysis using Prediction-based Word Embeddings
T Praveen Kumar, B. Vishnu Vardhan · 2022 International Conference on Edge Computing and Applications (ICECAA) · 2022
In the digital era, information in large amounts may be discovered in the form of reviews and writings of customers. People create reviews using only text or emojis that express the customer’s thoughts or opinions. In most instances, user comments reflect their gut feelings about a product. A significant amount of value is extracted from these remarks by using sentiment analysis. The classification is one of the most pressing challenges in sentiment analysis which is our main agenda in this work. The accuracy is used as a performance evaluator and gives weightage to sentiments given in the form of emojis. A multimodal classification model is utilized here by considering both text and emojis for the final output. The comparati ve study is also done on the word embedding methods of the Word2Vec model. According to the findings of the experiments, the proposed system outperforms the sentiment analysis models, which considers only text for evaluation.