Malaysian Chinese folk beliefs on Facebook based on LDA topic modelling

The study of Malaysian Chinese folk beliefs has a rich history, evolving from historical documentation in the early 20th century to fieldwork methods in later decades and now incorporating digital humanities approaches. Early research focused on the origins and attributes of deities, utilizing historical records and newspapers like *Le Bao*. Later, fieldwork became prominent, focusing on temple evolution, distribution, and religious practices. More recently, scholars have combined historical and field research methods to study dissemination networks and immigrant interactions. This study addresses the gap in research on Malaysian Chinese folk beliefs on social media, particularly Facebook, which has become a significant platform for the dissemination of these beliefs in Malaysia. The researchers utilize Latent Dirichlet Allocation (LDA), a natural language processing (NLP) technique, to analyze the thematic content of Facebook posts and investigate the roles of various actors in shaping these online discussions. The study aims to understand how social media influences the construction and practice of Malaysian Chinese folk beliefs.

Existing research on Malaysian Chinese folk beliefs has evolved significantly. Early studies (1930s-1950s) relied heavily on historical documents to trace deity origins and attributes, focusing on publications like the *Journal of the South Seas Society* and *Sin Chew Daily*. The 1960s and 70s saw a shift towards fieldwork methods, studying temple evolution and religious practices in Kuala Lumpur. The 1980s and 90s expanded this with case studies of specific beliefs like Nine Emperor Gods and I-Kuan Tao, alongside collecting cultural relics. The 21st century saw a move towards combining historical and field research to understand dissemination and immigrant interactions. The rise of digital humanities has led to interdisciplinary projects using GIS for data collection, representing the latest advancements in methodology. However, research focusing explicitly on Malaysian Chinese folk beliefs on social media platforms is lacking, which the current study addresses by using a large dataset and NLP methods.

This study employed a mixed-methods approach combining quantitative and qualitative analysis to investigate Malaysian Chinese folk beliefs on Facebook. First, keywords were collected from a website listing Malaysian Chinese temples (angkongkeng.com) and their patron deities (191 in total). Then, two researchers collected the first 25 Facebook text posts for each deity, resulting in an initial dataset of 4358 posts. Following data cleaning procedures (removing duplicates, invalid data, HTML tags, etc.) and Chinese text segmentation using the Jieba library, the final dataset comprised 4012 posts. Stop words were removed using the Baidu Chinese Stop Word List 2012. A custom dictionary was created to handle specific terms and synonyms related to folk beliefs. TF-IDF (Term Frequency-Inverse Document Frequency) was used for document representation. LDA topic modeling (using Gensim) was implemented to identify latent themes. The optimal number of topics was determined using coherence and perplexity metrics, settling on four topics. The study interpreted the resulting topics by examining keyword distributions and assigning titles to the topics. Finally, the researchers conducted a qualitative analysis of the identified themes and their relationships, discussing the roles of various actors involved in constructing and disseminating these beliefs online. Multidimensional scaling (MDS) was used to visualize the relationships between the four topics.

LDA topic modeling revealed four dominant themes in the Facebook data: 1. **Practitioners Worship (40.90% of tokens):** This theme focused on individual practices, including prayers for good fortune, safety, and blessings, along with rituals like worship, offerings, and temple visits. Key terms included 'whole family', 'safety and peace', 'bless', 'deity', 'grateful', and 'worship'. 2. **Temple Activities (31.59% of tokens):** This theme centered on temple-related events and activities, such as rituals, ceremonies, celebrations, and registration for temple events. Key terms included 'pray for good fortune', 'safety and peace', 'Tai Sui', 'rituals', and 'temple activities'. 3. **Deity Legends (17.38% of tokens):** This theme encompassed narratives and legends surrounding deities, highlighting Taoist elements, supernatural powers, and the origins and manifestations of various gods and goddesses. Key terms included 'Taoism', 'folk', 'belief', 'legends', and names of specific deities. 4. **Merchandise about Deity Statues (10.12% of tokens):** This theme focused on the commercial aspect of folk beliefs, particularly the sale and purchase of deity statues and related religious items. Key terms included 'deity statues', 'Buddhist amulets', 'camphor wood', and 'business'. Analysis of keyword overlap revealed a strong relationship between 'Practitioners Worship' and 'Temple Activities', with significant overlap in terms such as 'safety and peace', 'pray for demands', and 'merits and virtues'. These three concepts formed a central interconnected semantic nexus within the dataset. The study also identified four main constructors shaping online discourse: practitioners, temple organizations, media organizations, and merchants, each contributing to different aspects of the themes identified. The research highlighted the utilitarian nature of these beliefs, the regional diversity of deities worshipped, the multiple social functions served by temples, and the transnational flow of deity statues and practitioners.

The findings demonstrate that social media, specifically Facebook, plays a crucial role in constructing and disseminating Malaysian Chinese folk beliefs. The identified themes and the interactions between the four types of constructors illuminate the complex interplay between religious practice, social interaction, and commercial activities within the digital realm. The utilitarian nature of the beliefs, as evidenced by the emphasis on material and spiritual benefits, aligns with existing sociological understandings of religious practice in similar contexts. The strong presence of Taoist elements in the 'Deity Legends' theme points to the enduring influence of this religious tradition on Malaysian Chinese culture. The commercialization of religious items, as observed in the 'Merchandise about Deity Statues' theme, reflects the broader phenomenon of the beliefs economy and its adaptation to digital platforms. The study's methodology, employing LDA topic modeling and qualitative analysis, offers a valuable approach for understanding complex cultural phenomena within the digital sphere.

This study provides valuable insights into the dynamics of Malaysian Chinese folk beliefs in the digital age. The identification of four key themes, the roles of various actors in shaping online discourse, and the interconnectedness of key concepts contribute significantly to our understanding of this aspect of religious practice. The study highlights the adaptability and resilience of traditional beliefs in the face of modern technologies. Future research could explore the use of sentiment analysis to understand emotional responses toward these beliefs online and expand the data collection to include visual and video content from platforms like TikTok.

The study's limitations include the manual collection of data, which could have introduced bias. The focus on text posts from Facebook might not fully capture the nuances of beliefs expressed through images and videos on other platforms like TikTok, predominantly used by younger generations. The study did not use sentiment analysis to gauge emotional expression towards the beliefs. Future studies should address these limitations by employing more comprehensive data collection and analytical methods.