Sentiment Analysis of Scenic Users’ Comment Using FastText and LSTM Model

Abstract

With the unprecedented growth of the tourism industry and the Internet, word-of-mouth about tourist attractions has become an invaluable reference factor. Currently, attractions receive lots of redundant data from visitor reviews, which is time consuming and tedious to read and analyze. This paper attempts to find a correlation between user reviews and the attractiveness of attractions. Therefore, by finding a suitable model to classify users’ comments in terms of sentiment and analyzing the results, the scenic spot can better understand the users’ evaluation and feedback, so that it can adjust and improve the scenic spot’s service in a timely manner. In this paper, we compare the accuracy of sentiment classification between the fastText model and Long-Short Term Memory (LSTM) for the same training set under the same conditions. Simulation results demonstrate that the LSTM is accurate and precise in capturing the scenic comments made by the users. Therefore, we use the LSTM to analyze the data for the emotional tendency and to make suggestions for the improvement of scenic spots.