Hotel review analysis based on LDA model and KeyBert model

Abstract

The popularity of Internet applications and the continuous advancement of technology have made it easier for people to choose hotels. More and more people not only tend to book hotels through travel websites or software but also leave online reviews during their stay or after leaving the hotel. Almost all online bookers will carefully refer to reviews from previous customers of different hotels before making an order and then making a choice. In the past few years, many language models in the fields of machine learning and artificial intelligence have been widely used to analyze various texts. For reviews in areas such as hotels, most studies focus on sentiment analysis, that is, analyzing whether the sentiment in the reviews is positive or negative. Most previous studies used the LDA model to divide the topic words in the reviews or the KeyBert model to extract keywords from the text and compare and analyze them with other models. Some of these studies also used the hotel dataset compiled by Tan Songbo for analysis. Based on the theories of previous studies, this study uses the LDA model to extract and analyze different topic words of hotel reviews based on the hotel review data set compiled by Professor Tan Songbo to find out the advantages and disadvantages of the hotel. This study also uses the KeyBert model to extract the number of occurrences of specific keywords estimate the number of repeat customers by analyzing the keywords in the positive reviews, and estimate the number of non-target customer groups based on the keywords in the negative reviews. The research results are of great significance to the management and marketing decisions of the hotel, and can also provide better help to the hotel’s booker.