Analisis Sentimen Dari Keywords Yang Dimasukan Pengguna Di Twitter Indonesia Untuk Penunjang Pembelajaran Strategi Komunikasi Di Program Studi Ilmu Komunikasi Universitas Kristen Petra Dengan Metode Cnn-Bidirectional Lstm

Abstract

To increase online media traffic, the first effort made by online media is to examine the trending phenomenon with the right marketing strategy. One of the methods that online media is used is a communication strategy that utilizes the sentiment analysis method. In reality, students of Communication Science Major at Petra Christian University are not optimally using sentiment analyst system because the sentiment analysis system for the Communication Studies Study Program (Netray) cannot be run by more than one student or is not used simultaneously and the price of the application is still not affordable if the students want to subscribe Netray. So a sentiment analysis system is needed to support the learning of the Communication Science Major at Petra Christian University. In previous related research, there was research that discussed the analysis of the #crowdfunding campaign on Twitter but there was not include sentiment analysis, there are only topological analysis, spatial analysis and others analysis. In addition, there are studies that use various deep learning methods of sentiment analysis, by researching CNN, DNN, RNN, Bi-Lstm, but none of them combine these methods. So it can be concluded that research will be made that analyzes sentiment analysis and combines deep learning methods. Sentiment analysis is the process of using text analytics to obtain various data sources from the internet and various social media platforms. Sentiment analysis can be utilized with artificial intelligence or with computing, because it is more efficient . Sentiment analysis can be complemented by methods from artificial intelligence systems, namely deep learning CNN-BILSTM. CNN-BILSTM is a combination of the two methods of CNN and bidirectional LSTM where CNN is the input layer and bidirectional LSTM is the layer that extracts features from the input. The dataset used in this application is retrieved from github by adopting the CC BY-NC (Common Creative Non Commercial) License. Data used in the deep learning model which contains a collection of Indonesian tweets containing neutral, positive, negative sentiments.From two testing this thesis using twitter as the online media. From the first test, 20 tweets were searched, the tweet contain "Shin tae yong” and yielded an accuracy of 30%. The second test was tested by 45 students of the Petra Christian University Communication Science Program at Petra Christian University Surabaya in the Q2.505 building where this application was tried and applied, after that the application was assessed with a satisfaction questionnaire which resulted in an average score of 4.01, so this application can meet the needs of the Petra Christian University Communication Science Program with the initial target of a satisfaction questionnaire of 3.75.