Abstract
Tourism industry is one of the largest sectors and also one of the reasons to the development of hotels around the world. Most of them also use online booking system. Many online reviews are increasing every day because travellers are willing to share their experiences. This report presents the worked progress of sentiment analysis on hotel reviews using convolutional neural network. It is implemented to help users which is hotel’s staff that want to know more about their hotels from the reviews. The users face problems while reading the reviews because it consumes a lot of times and expressions of the sentences that are hard to understand whether it is positive or negative opinions since human language is complex. The traditional algorithm that been used a lot also reach a plateau of performance, with no room for improvement. So, the aim of this paper is to develop a prototype for hotel reviews using CNN Algorithm, to identify the requirements of CNN technique for text classification and also to evaluate the accuracy of CNN algorithm. The hotel reviews will be analysed using CNN text classification whether it is positive or negative sentiment. The experiment was doing by replacing the batch size, filter size, dropout and also the split of the data between training and testing numbers. It also used word2vec as word embedding. The result shows that CNN algorithm method can have high accuracy with more than 90% accuracy. The prototype also been implemented using CNN model to predict the sentiment on hotel review, hi conclusion, the CNN algorithm can be used as text classification as it gives a high accuracy.
Metadata
Download
55656.pdf
Download (141kB)