Sentiment Analysis in social media: Handling Noisy Data and Detecting Sarcasm Using a Deep Learning Approach

Social media is a platform that is highly developed with the features offered so that the use of deep learning will in the process have an artificial neural network architecture in completing tasks in certain cases. Sentiment analysis will do positive, negative and neutral assistance. The problem in this study is that the data contains in the form of slang sentences, slang and writing errors contained in social media data so that literary dictionaries and manual annotations will be applied later. can handle noisy data and can do sarcasm detection. The results obtained in this study by approaching the deep learning model with the LSTM algorithm in sentiment analysis yielded an accuracy of 0.92, a precision value of 0.83, a recall value of 0.76 and an F1 value of 0.78 while the model from RNN in detecting sarcasm has a value 0.94, precision value 0.77, recall value 0.80 and F1 value 0.85.