Sentiment Analysis on Hotel Ratings Using Dynamic Convolution Neural Network

Abstract

Currently, the role of information technology is very important in everyday life because heavy workloads can become easier, communication time can be made shorter and data processing can be faster and more accurate. Hotel ranking sentiment analysis can provide important information for hotel owners and managers to improve the quality of service and guest experience. It can also be used by prospective guests to make the right booking decisions. Sentiment analysis can identify positive or negative feelings from guest reviews. There are 694,213 data reviews about hotels using English which are used as training data. The data was preprocessed and 76,905 vocabularies were obtained by utilizing Word2Vec. The training data was carried out at the encoding stage. The DCNN model is given a K-Max-Polling value of 2. The model is trained for 20 epochs. The model that has been formed is tested with 173,554 data and obtained an accuracy rate of 95%.