Page 163 - Kỷ yếu hội thảo khoa học lần thứ 12 - Công nghệ thông tin và Ứng dụng trong các lĩnh vực (CITA 2023)

Linh Bui Khanh, Anh Nguyen Quynh, Hai Tran Van, Ha Nguyen Thi Thu 147

that analyzing customer reviews in text form can help uncover keywords that represent
the restaurant experience and identify underlying patterns in the text. Text analysis of
hotel experiences created by customers will be the real emotions that customers have
experienced in the hotel. Customer-centricity brings high efficiency when managing
relationships and taking care of customers. Analyze customer psychology with many
objective dimensions, not limited based on pre-designed questions. Data is available on
the Internet, saving costs easy for new data collection plans.
There are some methods of predicting, and classifying the sentiment of customers
using machine learning tools [1, 12, 16, 19, 22, 23, 25, 26, 27]. LSTM is a classification
method that has high accuracy and ensures the context of sentences different from
machine learning methods with other discrete inputs [13, 14, 15, 17, 18, 20, 21, 24].
In this paper, we used an LSTM model to predict customers' sentiment toward 4-5
star hotel services in Vietnam. First, we used the Harvey Web tool to collect online
review data from TripAdvisor about hotels representing six major cities in Vietnam:
Hanoi, Ho Chi Minh City, Quy Nhon, Nha Trang, Hue, and DaNang. Next, we
preprocess the labeling of these data and finally use an LSTM model with 02 classes to
predict the customer's sentiment after using hotel services in Vietnam.
The rest of the paper is structured as follows: Section 2 presents relevant research in
the field of customer sentiment analysis, section 3 introduces the LSTM model, the
methodology will be presented in section 4, section 5 is the results, and finally is the
conclusion.

2 Related Work

Research on data mining that is generated by customers on the Internet is one of the
research directions on the hotel industry in recent years [10]. The studies focused on
collecting data from the online hotel booking platform into a large dataset and analyzing
the data to understand customer attitudes and emotions towards hotel services [11], they
used using natural language processing tools to extract frequent words that represented
important aspects of the hotel services. These words are categorized into different
topics that also can understand hotel-
, - ,

dissatisfaction of the guests at the 4-5 star hotel [7]. Tools for using this big data mining
include various languages such as Python, R, or Rapidminer [9]. Hongxiu Li et al [6]
also approach the data analysis of the hotel online review. They assessed hotel aspects
for customer satisfaction by extracting 412,784 consumer-generated reviews from
TripAdvisor of 05 major Chinese cities including Sanya, Beijing, Guangzhou, Shanghai
Hai, and Hangzhou. A regression model ranks overall through variables: cleanliness,
price, room, service, location, etc. This study focuses on analyzing customer
satisfaction for some attributes of the hotel, they can classify domestic customers and
foreign customers. Hotel quality is also classified into different levels to show that in
each type of hotel, customers often have a greater demand for which standard of the
hotel.

ISBN: 978-604-80-8083-9 CITA 2023