Sentiment classification and visualization analysis of tourism comments: A canton tower example

Abstract

As calculated by the Ministry of Culture and Tourism, there were 308 million domestic tourist trips in China during the Spring Festival in 2023, witnessing a year-on-year increase of 23.1%. And the satisfaction of tourists with certain spots can be partly reflected in the comments and scores they made on social media. Therefore, this research was aimed at mining useful information from the comment and scores of Canton Tower. After collecting detailed comment information from the web, this research used the plot module of Python to make data visualization to observe the distribution of users location, comment time, and comment label as well as the word cloud of remarks. Then the research used the data set to train three different sentiment analysis models including Nave Beyas, SnowNLP, and Bert, then compared their accuracy in predicting. This research shows that over half of the comments came from Guangdong Province, most of the tourists were content with Canton Tower, and the number of comments has increased obviously since 2023. In addition, the research found that the model having the highest accuracy of sentiment analysis is the Bert model, about 90%.