Analysis of Social Media Data Using Contextual Embedding Leverage Models with Convolutional Neural Networks
Arif Ridho Lubis, Yulia Fatmi, Deden Witarsyah · 2023
This research proposes a model approach that uses leverage contextual embedding with Convolutional Neural Networks (CNN) in the context of sentiment analysis. The use of contextual embedding leverage aims to get a more contextual representation of words by using BERT. The results obtained from this study are a combination of batch size and epoch parameters with parameter values including batch values of 36, epoch values of 30 get an accuracy of 0.84, batch size values of 64, epoch values of 60 get an accuracy of 0.89, batch size values of 128, epoch values of 80 get an accuracy of 0.92 and batch size value of 256, epoch value of 100 obtains an accuracy of 0.94. so that the combination of CNN parameters and the use of contextual embedding leverage affects accuracy. This research is expected to provide an understanding of semantics in the context of sentiment analysis.