Video Sentiment Analysis: A Fusion Approach using CNN and LSTM Techniques

For the past few decades, textual sentiment analysis leveraging text mining techniques has been responsible for the majority of dedication done on sentiment analysis. However, there is still much to learn about the topic of audio sentiment analysis. In the proposed study, we use speaker differentiated voice transcripts to perform sentiment analysis to identify the feelings of the various speakers. To identify effective algorithms for this task, we examined various methods for sentiment analysis and speaker discrimination. We started out by reviewing the approaches that various academics have proposed for sentiment analysis across all data modalities. Then, we made an effort to put some of those tactics into action. We have focused on audio sentiment analysis in this work. New methods are being employed to analyze audio data as a result of the ongoing research into audio sentiment analysis. Here, we try to utilize machine learning to divide audio into several emotions. In this work, audio sentiment was explored in addition to multimode analysis. The categorization of emotions utilizing a range of data formats, such as text, audio, and video, is a component of multimodal sentiment analysis. Our method also yielded reasonable outcomes when merging the individual modality with attention networks for audio sentiment analysis. The aim was to build a system that can identify six various emotions in an audio recording, including anger, pleasure, disgust, sadness, fear, and surprise, when data is fed into it.