Scalable

As the number of people using and participating in social media grows, academics become interest in studying this new media, specifically comment analysis, in order to comprehend public opinion and user behavior. However, there are no studies that map the development of comment analysis domain, which would be valuable for future research. To address the issue, we examine prior publications using PRISMA approach, and offer suggestions for further research. An investigation was conducted to locate pertinent publications published in databases between 2010 and 2022. On the basis of our examination of 115 relevant articles, we found that, within the scope of methodology, prior researches employ two methods (sentiment and content analysis) and three tools (human, software, and mixed coders), and the majority of them concentrate on gathering data from western countries, covering numerous platforms and topics. Based on these findings, we recommend that future research in comment analysis should synthesize methods and instruments. In addition, examine areas that have not been fully explore in terms of platforms (e.g., Instagram and Tiktok), topic (e.g., local government), and regions (e.g., eastern countries) that would be valuable in order to enhance the body of knowledge in this domain.


INTRODUCTION
Social media has gained popularity as a source of information , and a place for expressing opinions on a number of topics . Furthermore, the easier and more adaptable the device for accessing social media, the stronger and more powerful the connection will be (Lingam & Aripin, 2017). This connection is exemplified by the fact that every citizen can leave comments on various social media platforms , and liberated to express their opinions and attitudes regarding both public and sensitive issues due to the nature of  Towner, 2012). This phenomenon drives scholars to investigate the content, behavior, and patterns of citizens' comments on social media using comment analysis method (Chumwatana, 2018). Additionally, the analysis of comments constitutes a potentially interesting data source to mine for obtaining implicit knowledge about users, post, categories, community interests  and to comprehend the generative deliberation potential of the emerging technology and its capacity to create a virtual public sphere (Ksiazek, 2015).
Since more than a decade ago, scholars have been interested in analyzing public opinion generated from comment sections on a variety of topics, including food topics ( To the best of our knowledge, no previous research has attempted to comprehensively summarize the existing body of knowledge in the field of comment analysis. As a result, there is a limited understanding of the strengths and limitations of the research that has been conducted in this field, including methodologies, regions, tools, research objects, and platforms used. To address this gap, we propose a systematic literature review (SLR) on comment analysis literatures in order to advance the current state of knowledge (Hamid et  To achieve the objectives of this study, a SLR was conducted using PRISMA approach. Our research material consists of 115 scholarly articles published between 2010 and 2022. Our research findings have academic implications, namely mapping how past studies have addressed comment analysis and providing recommendations for future research. More specifically this study aims to: 1) conduct a comprehensive literature review on comment analysis; 2) Identify the various methodologies, regions, tools, research objects, and platforms explored by prior studies; and 3) provide recommendations for future research.
This paper is organized as follows. This article, as indicated, presents the studies conducted regarding comment analysis. The subsequent section describes in detail our research methodology, and then how previous research on comment analysis was carried out. The final section summarizes the primary contributions of the study, mainly recommendations for future research.

MATERIALS AND METHODS
This systematic review was conducted in accordance with the PRISMA principles, which provide a structured approach for synthesizing existing evidence and providing transparent and thorough reports that can assist decision making based on evidence (Page et al., 2021). The Scopus database and Scopus-indexed publishers such as Taylor & Francis, Springer, Science Direct, and Emerald were utilized to obtain publications relevant to our study. In addition, a snowball search step was used to identify further publications that were published between the period 2010 until 2022, this process followed the pattern that was established by Macke & Genari (2019) and Tandon et al. (2020). Furthermore, we included in academic publications studies on comment analysis of social media platforms that were published in English (e.g. books, book chapters, and peer-reviewed articles) and exclude studies that were primarily concerned with developing analytical tools or analyzing the effect of comments on other factors.
We began our search with the predefined keywords 'social media', 'comment', and 'analysis', as well as 'social media OR 'comment' OR 'analysis' This preliminary search produced over 1,965 results. The authors limited the number of papers included in the review by restricting the search to the title, abstract, and keywords generated by an algorithm from the citations or bibliographies of the records. After deleting duplicates and articles that did not meet the requirements, our initial search revealed 340 references, which served as the basis for our research. After reviewing the titles and abstracts of these papers, the authors discarded 11 that did not retrieved, leaving 329 to be retrieved.
We eliminated 214 papers as unrelated or accepted in our inclusion area to our examination because they focused insufficiently on the study of comments analysis, for example, by examining the effect of other variables or focusing on the comment distribution pattern. Based on this reason, we reserve 115 for evaluation. By analyzing platforms, and then provide recommendations for future study based on this data and utilizing the remaining 115 references, we construct a mapping of approaches, methods, contexts, and. The procedure depicted in Figure 1.

Number of Paper's Citation
To evaluate the influence and interest of previous studies, using a reliable bibliometric tool, specifically Google Scholar (Harzing & Alakangas, 2016), we track how many times each paper has been cited. The ten articles that received most citations are listed in Table 1. As shown in the table below, the most influential paper is address incivility in news website comments , which received 845 citations, followed by paper that discuss similar topic (Santana, 2014) which has 530 citations. However, this paper analyzed the user-comment sections of newspapers that allow anonymity and those that do not; then, compares the two in terms of civility. And the third most cited paper is identify which aspects of a dining experience are most important from customers' point of view (Pantelidis, 2010), which has 497 citations.

Distribution of Regions
In the context of region distribution, we mapped two types of papers, namely first, papers that focus on comments in general, in the sense that they do not focus on specific regions, for example papers written by Siersdorfer et al. (2010) and Thelwall (2011). The second is the type of papers that limits the object of their research in a particular region, for instance papers written by Rowe (2015a) and Ziegele et al. (2014). To illustrate our collected data, we give figure 3 shown below:

Figure 3. Distribution of Regions
According to our data, we discovered publications that limit their research of social media comments to particular areas of the region (several studies do not specify the region examined

Typology of Comment Analysis
Research on the analysis of comments in social media, in our results, involves two types of research. The first form of research is analysis that classifies or codes comments from social media users, as seen in the research of R.

Difference of Analyzing Tools
In addition to the two categories of research regarding comment analysis, there are also a variety of methods used by researchers to analyze user comments. According to Teh et al. (2018), tools are used by researchers to identify meaning as well as patterns in comments. We found that researchers was divided into three type when using tools to analyze comment analysis, namely human-coder, software-coder and mix between them and shown in Appendix 1.
As implicitly explained by  and Tao & Jacobs (2019), human-coder refers to the manual analysis conducted by the researcher without software analysis. We discovered 80 papers of comments analysis either content analysis or sentiment analysis which used this technique. In contrast to the human-coder, the software-coder refers to the analysis carried out using the comment analysis software tool .    Figure 6. According to R. da Silva & Crilley (2016), social media platforms such as YouTube, Facebook, Twitter, and others permit users to comment and express their opinion, one possible reason that encourages researchers to investigate these platforms.

DISCUSSION
In this research, a literature review of social media comment analysis has been conducted. We showed the distribution of articles by year, most influential papers and authors, region, method, tools, and contexts. On the basis of these data, we conclude that the analysis of social media comments conducted by researchers has a broad reach, encompassing a variety of context and methodologies. In the following section, we will examine and provide recommendations based on the findings reported in the previous section.

Urgency for more research on comments in the future
Our data indicate a decrease in social media comment analysis papers, specifically between 2020 and 2022. According to Global Active User Figure  (

Other objects that need to be explored
Based on the number of citations displayed in Table  1, it is clear that the majority of citations pertain to studies that analyze comments in the context of news, more precisely in the study on incivility and civility in users comments. For example, Santana

Asian region as the highest source of social media users
We find that research concentrate on evaluating social media comments from certain region. The North America is the most researched region, the second is Western Europe and then followed by Southern Europe. One possible explanation for this tendency is that Western regions have the most social media users. We argue, based on these facts, that social media participation in Asia is also very high. This information should also serve as a guideline for future research, with the Asian region serving as the study's point of focus.

Differences in approach and new recommendations
As stated in the preceding section, research on social media comments is divided into two categories: content analysis and sentiment analysis. Content analysis is a research method that involves making the content of messages manifest by identifying as objectively as feasible their properties. While sentiment analysis focuses on automatically assessing whether a text has an opinion, recognizing whether the polarity or sentiment represented is favorable, negative, or neutral, and extracting an author's assessment of particular features of a topic.

Advantages and Disadvantages of Tools
There are three tools that researchers use to analyze social media comments. As previously stated, the three tools are the human coder, the software coder, and the mixed-coder. Researchers utilize softwarecoder because it eliminates bias (Siersdorfer et  Other researchers utilize these two tools concurrently. The simultaneous use of both of these technologies to analyze social media comments is referred to as mixed-coding. These researchers utilize these two tools because they are more efficient, provide a high level of reliability, and can compensate for each tool's limitations (

TikTok and Instagram as platforms that need to be research
In the context of the platform, we discovered that YouTube was the most examined platform, followed by Facebook and the News Website. Because these platforms enable user's engagement and interaction (R. da Silva & Crilley, 2013), and are widely used, popular, and influential on their users' behavior, they have been investigated intensively by researchers Al-Zaman (2021) and Liebig et al. (2017) Arancibia & Montecino, (2017). In besides these three platforms, we discovered that Instagram (1440 million users) and TikTok (1023 million users) also have a very significant user base (Figure, 2022b). Therefore, we argue that future researchers must also explore the two platforms.

CONCLUSIONS
The aim of this study is to identify existing studies on comment analysis and to suggest future research based on the existing research gaps. There are seven findings from our study, including, first, the number of publications on comment analysis is decreasing in 2022, according to the distribution of research years, we discover in the first place. We demonstrate the need for further publications on this subject since, as social media usage rises annually, more academics are able to access the data for analysis, making research potential becoming increasingly essential. Second, our results reveals that research on civility and incivility in the news realm is highly cited, reflecting the interest in this topic among academics. We urge further exploration of civility and incivility in the context of citizens interacting with their local government, an issue that has been largely forgotten but seems to be important given the increased use of social media by local governments. There is a considerable quantity of data available from the comments posted by residents on local government social media, which gives an opportunity for scholars to investigate this topic.
Third, researchers concentrated more on Western regions as a result of our findings. In addition, as we previously demonstrated, there are a lot of social media users in Asian regions as well, thus we draw the conclusion that future studies should include study on Eastern region.
Fourth, we demonstrate that sentiment analysis and content analysis are the two main approaches used in commentary analysis research. Additionally, we discussed the advantages and disadvantages of these two techniques in terms of comprehending user patterns and behavior while leaving comments on social media. Based on these conclusions, we subsequently offer suggestions for combining methods for researching user comments on social media, with the goal of creating a more comprehensive and in-depth approach for examining user comments on social media.
Fifth, we demonstrate the three tools that researchers employ to examine comments including humancoder, software-coder, and mixed-coder. We also examine the advantages and limitations of each tool used for analysis in previous section. Following the researchers who combined the two techniques, we also recommend that future researchers should mix the two tools. We do point out that the two tools combined must place a strong emphasis on the human coder as the primary assessor. Because human-coders is better able to discern subtle messages than software-coders. However, the assistance of a software-coder is also necessary, since it can make the analysis process run more efficiently.
Finally, prior studies has primarly focused analyzing comments on specific social media platforms, including YouTube, Facebook, and news websites. Therefore, we suggested that future research include examination of comments on Instagram and TikTok, as these platforms have a vast user base and deserve investigation. This will contribute to a more thorough comprehension of comments on social media platforms. [1] Abdalla, M., Ally, M., & Jabri-Markwell, R.