Page 165 - GHES-3-2

Global Health Economics and
Sustainability
Online health community reviews

Sentiment analysis is the process of examining textual 2. Methods
data to ascertain the emotional tone as positive, negative,
or neutral and to discern whether the sentiment expressed 2.1. Data source
is in agreement or opposition. With the exponential growth Good Doctor Online is one of China’s leading online medical
of unstructured text data on the Internet, driven by the platforms. As of May 2024, more than 240,000 doctors
rapid advancement of web technologies, sentiment analysis have registered on the platform. In this study, we selected
has become integral to the automated assessment of online three different types of diseases (diabetes, leukemia, and
reviews. This technology enables us to comprehend the depression) as examples, analyzing the reviews by users in
perspectives and sentiments of users when they share the Good Doctor Online community. These three diseases
their thoughts online. For instance, Ahmed et al. (2020) rank among the top three in the number of visits on the
developed a methodology for emotion clustering based Good Doctor Online medical platform and are regarded
on sentence context and introduced a weakly supervised as common diseases in chronic diseases, hematological
neural network model that integrates manual and diseases, and psychology, all of which have typical clinical
automated learning to construct a multilingual sentiment significance and research representativeness. The Python
dictionary. This dictionary significantly enhances the language was used to design a multithreaded crawler tool
precision of sentiment discrimination. Ji & Fangbi (2016) to crawl all patients’ reviews on doctors that are related to
proposed a deep neural network model for sentiment the treatment of the three diseases. A total of 85425 reviews
analysis of massive open online course reviews, utilizing it of 7423 doctors were collected.
to evaluate the emotional tenor of online course feedback.
Liu et al. (2020) gathered posts on various topics, such as 2.2. Data analysis
heart disease, hypertension, depression, and obsessive- This study utilized the Python library SnowNLP (SnowNLP,
compulsive disorder, from Baidu Post Bar. Through 2017) of Chinese natural language processing to realize
thematic modeling and sentiment analysis of these posts, the emotional analysis of the reviews. The stop word list
they identified the thematic and emotional disparities is supplemented by the Harbin Institute of Technology
in content posted online by individuals suffering from Stop Word List and Baidu Stop Word List to use a more
physical and mental health conditions. Ortigosa et al. comprehensive list to remove stop words. First, links,
(2014) presented a novel method for sentiment analysis on special symbols, and pictures were removed from the text.
Facebook, achieving an accuracy of 83.27% in analyzing the Then, the Jieba segmentation package was used to segment
emotions expressed in user-written content. Consequently, each sentence, fetch the emotion dictionary, identify the
sentiment analysis has been extensively applied to the emotion classification of each word in the sentence, and
examination of a diverse array of comments across various calculate the emotion score. The resulting data included all
contexts.
the reviews associated with the corresponding doctor. The
In the OHC, the comments and feedback of patients are sentiment module in SnowNLP of the Python class library
evaluated accordingly. Sentiment analysis based on OHC was used to analyze a single sentence, and the corresponding
reviews is a method to analyze the subjective reviews of sentiment value of each review was obtained. The size of this
users and obtain their emotional tendencies and attitudes. value represents the probability that the review was positive.
In an OHC, every review by the patients of doctors The resulting data included all the reviews associated with
includes an evaluation and the emotional attitudes toward the corresponding doctor. Similarly, the sentiment module
doctors. Subsequent patients usually browse the comments was used to analyze a single sentence, and we obtained the
of other patients on doctors when choosing a doctor and corresponding sentiment value of each review. The size of the
subsequently make decisions by referring to these reviews. value represents the probability that the review was positive.
Through sentiment analysis, the emotional orientation
of users can be identified from these reviews. Sentiment For the emotional analysis of each review, the top five
analysis on OHC user reviews is particularly important; high-frequency words in all reviews corresponding to each
however, there are only a few sentiment analysis studies on doctor were extracted, and the word frequency statistics
OHC user reviews. were calculated. If there were limited reviews, the keywords
were extracted in a particular order.
In this study, the emotional value of reviews was
analyzed using a method based on an emotional dictionary 3. Results
to better understand the emotional trends of patients and 3.1. Sentiment analysis
the main concerns regarding doctors during medical
treatment, as well as to provide a reference for improving The sentiment value output results range from 0 to 1. From
the service quality of OHC doctors. the overall analysis of the emotional results, we selected

Volume 3 Issue 2 (2025) 157 https://doi.org/10.36922/ghes.7052