The rapid evolution of new media has fundamentally reconfigured the logic of information dissemination, transforming public communication from unidirectional flows to interactive, multi-platform ecosystems. In this digital environment, platforms such as news outlets, online forums, and social media not only co-construct public discourse but also shape agenda-setting processes and emotional dynamics (Andersen et al., 2024). These mechanisms extend beyond technical mediation, influencing public perception, policy legitimacy, and urban image-making across sociocultural and geographic contexts (Ausat, 2023). However, how heterogeneous platforms interact in shaping communication during major events remains underexplored.
This study draws on agenda-setting theory, which explains how media not only spotlight issues but also determine their perceived importance (Djerf-Pierre et al., 2024), alongside emotional communication theory, which highlights the role of affect in shaping cognition and engagement (Fu et al., 2024). These theoretical frameworks jointly enable a critical inquiry into how digital media construct urban narratives and mediate public sentiment across regions.
In this study, the term “Maritime Silk Road” is used primarily as a historical and geographical reference to the Fujian coastal cities of Fuzhou, Xiamen, and Quanzhou, which have long served as important maritime and cultural links between mainland China and Taiwan. Rather than examining perceptions of state policy, the study focuses on how these cities are represented, discussed, and affectively engaged within Taiwanese online discourse.
Focusing on Fuzhou, Xiamen, and Quanzhou—three coastal cities with deep historical, cultural, and media ties to Taiwan—this study examines how Taiwanese Internet users engage in discourse about these urban spaces. Structural links such as the Mini Three Links and symbolic continuities associated with the maritime exchange tradition, coupled with linguistic and digital proximity, foster high levels of engagement and cultural resonance.
To address gaps in the literature that often treat platforms in isolation, this study employs a large-scale Taiwan-based dataset and a pre-test/post-test design to examine how platform type, content format, and event stimuli influence online voice and sentiment. By incorporating cross-platform behavioral metrics and temporal triggers into a unified analytical framework, the study contributes to agenda-setting theory and offers practical insights for civic communication and regional branding.
Building on this foundation, the study advances classical agenda-setting theory by proposing the concept of cross-platform agenda cycling, which explains how issue salience circulates across news media, social platforms, and forums through iterative feedback. In the event windows examined in this study, institutional news media often supplied the initial visibility of urban issues, social platforms intensified affective circulation, and forums sustained longer-term discussion. This pattern describes the dominant empirical trajectory observed in the present dataset, while allowing for other entry points through which social media posts or forum discussions may also trigger subsequent news coverage and wider cross-platform circulation. Informed by research on algorithmic agenda-setting (Einarsson et al., 2024), cross-platform diffusion and platformization (Hase et al., 2022; B. Kim et al., 2024), and affective publics (Papacharissi, 2015), this framing extends agenda-setting from a linear hierarchy of influence to a networked process shaped by institutional visibility, algorithmic mediation, participatory circulation, and regional context.
Conceptually, this study integrates four core constructs—platform logic, affective communication, agenda-setting dynamics, and online voice—into a unified analytical framework. Platform logic refers to the distinct communicative affordances and algorithmic structures of news media, social media, and online forums, which shape how content is produced, circulated, and engaged with. Within these platform environments, affective communication operates as a process through which emotional expressions, evaluative language, and collective sentiment influence user attention and participation. Agenda-setting is thus reconceptualized not as a linear transmission of issue salience from institutional media to audiences, but as a dynamic, cross-platform process in which salience emerges through iterative interactions between institutional, algorithmic, and participatory actors. Online voice, in this framework, represents the observable outcome of these processes, capturing both the visibility and the participatory intensity of public discourse. By linking platform-specific logics to affective dynamics and agenda formation, we provide a cohesive theoretical foundation for examining how online voice is generated and negotiated across heterogeneous digital environments.
Literature Review and Research Hypotheses
The Multidimensional Dynamics of Online Topic Diffusion
In the digital era, topic diffusion emerges through the interdependent dynamics of information volume, user engagement, and emotional resonance. These elements collectively determine the intensity and trajectory of online discourse. Agenda-setting theory provides a foundational framework for understanding how communicative intensity translates into issue salience by structuring media exposure and shaping public attention (Buturoiu et al., 2023; Djerf-Pierre et al., 2024; McCombs & Shaw, 1972). However, in algorithmically mediated environments, salience formation is no longer driven solely by institutional gatekeeping but emerges from the interplay between human agency and algorithmic curation (Wang et al., 2023).
Empirical evidence shows that a higher volume of online voice enhances content visibility and amplifies public awareness (Newman et al., 2023). Interactive behaviors—such as reposting, commenting, and liking—extend visibility, particularly when content is designed to encourage participation (Guo et al., 2024). However, visibility alone does not guarantee sustained momentum; content type and contextual relevance are critical in determining whether engagement leads to diffusion over time (Sangiorgio et al., 2025).
This dynamic aligns with the notion that diffusion is not merely a quantitative escalation but a recursive process shaped by platform feedback loops that privilege emotionally resonant and interaction-rich content (Brady, 2024; Brady et al., 2021; Rathje et al., 2021).
Beyond structural affordances, affective dynamics play an equally crucial role in shaping topic diffusion within digital ecosystems. Emotion functions as the engine of amplification rather than a peripheral cue. Content that evokes strong affect—such as anger, joy, or awe—circulates faster and travels farther, synchronizing user cognition and behavior through processes of emotional contagion (Andersen et al., 2024; Hatfield et al., 1994). Recent studies suggest that emotional intensity interacts with algorithmic prioritization, creating asymmetric visibility across platforms (Humprecht et al., 2024).
To capture these dynamics, this study integrates platform logic theory (van Dijck et al., 2018), which conceptualizes digital platforms as socio-technical systems in which platform-specific visibility patterns, datafication, and user participation jointly structure communication flows. This lens emphasizes how platform-specific affordances mediate the circulation and amplification of topics within Taiwan’s online ecosystems. Complementarily, communication infrastructure theory (Y. C. Kim & Ball-Rokeach, 2006) explains how communication ecologies connect institutional, community, and individual actors through locally embedded media networks.
Building on these theoretical foundations, the study proposes the concept of cross-platform agenda cycling—a recursive process in which institutional media initiate topics, social media amplify emotional engagement, and forums sustain reflective discussion across time. This model extends classical agenda-setting theory by incorporating feedback dynamics across heterogeneous platforms, where algorithmic mediation and participatory agency jointly shape issue salience.
In sum, online topic diffusion is not a linear cascade but a multilayered, affect-driven process in which engagement and emotion co-produce issue salience through iterative feedback between platforms. Rather than treating agenda-setting as a unidirectional media effect, this perspective reframes it as a networked process of agenda circulation—an evolving system embedded within cross-platform infrastructures and sustained by both institutional and user-driven feedback loops.
Types of New Media Platforms and Online Voice
The rapid diversification of new media platforms has fundamentally altered patterns of information dissemination, public engagement, and media consumption (Espeland, 2024). These shifts directly and indirectly influence the dynamics of online voice and reshape agenda-setting processes (Boni et al., 2025).
While digital media are often treated as a single ecosystem, each platform operates through a distinct communicative logic that mediates visibility and participation—a configuration of technological affordances, user cultures, and governance regimes that together mediate how voice is produced and amplified (Hegarty, 2024; van Dijck et al., 2018). In cross-strait communication, these logics acquire additional significance: the same message may circulate freely on Taiwan-based social media but remain constrained or reframed on Mainland-linked platforms, producing asymmetric visibility and sentiment.
Platform architectures thus create differentiated capacities for amplification—not only because of their interface design, but also through embedded political and cultural filters that shape what becomes speakable or shareable. For example, online forums, as participatory hubs of brand communities, promote sustained interaction and enhance both electronic word-of-mouth (eWOM) and issue visibility (Martínez-López et al., 2021). High interactivity within forums fosters collective discourse and information sharing.
By contrast, social media platforms operate within economies of attention that reward speed, virality, and emotional intensity. Twitter (now X) accelerates the rise of trending topics through real-time brevity, whereas Facebook and Instagram sustain engagement through visual narration and algorithmic personalization (Dhanesh et al., 2022). Messaging apps such as Line, while more closed, facilitate micro-circulation of trust-based information—an especially salient form of mediated intimacy in Taiwanese digital culture (Shahbaznezhad et al., 2022).
These contrasts matter particularly for discourse about Fuzhou, Xiamen, and Quanzhou, where cross-strait historical proximity coexists with informational distance. Forums may attract diasporic nostalgia and community dialogue; social media amplify emotional and visual cues tied to cultural identity; news portals frame the same topics within policy or economic narratives. Thus, each platform not only hosts communication but performs a distinct interpretive function in constructing online voice.
Cross-platform diffusion further amplifies reach. Topics initially emerging on Twitter often cascade across Facebook and Instagram, or PTT, a major Taiwanese bulletin-board-style online forum, generating recursive loops of exposure and reinterpretation (Mueller-Herbst et al., 2020; Muhammed & Mathew, 2022). The interplay of immediacy and persistence across these spaces defines the rhythm of issue salience.
In sum, platform architecture, political economy, and cultural context jointly determine how online voice takes shape. The convergence of technological affordances, user behavior, and cross-strait communicative asymmetries defines the evolving ecology of visibility.
Hypothesis 1 (H1): Online voice concerning Fuzhou, Xiamen, and Quanzhou differs significantly across news media, forums, and social media.
Content Types and Online Voice
Content format plays a pivotal role in shaping the effectiveness of digital communication and influencing online voice. Different modalities—text, image, video, or hybrid—engage audiences through distinct cognitive and emotional pathways, shaping agenda-setting outcomes (Moran et al., 2020).
Across contemporary media ecosystems, content forms are inseparable from the technological and cultural environments through which they circulate: each format carries an affordance grammar—bundles of cues and sensory intensities— that couple with platform algorithms and user expectations to shape visibility and engagement (Huszár et al., 2022; Scharlach & Toma, 2023). Within cross-strait communication, these grammars become politically and emotionally inflected: textual reports may stress credibility and authority, while visual or video-based narratives invite affective identification and cultural intimacy.
Text-based formats, such as long-form reports or detailed commentary, remain central to digital discourse, particularly in contexts where credibility and depth are essential. Hasan et al. (2024) found that such content enhances perceived trust and engagement. Similarly, Hossain and Kibria (2024) note that rating and commenting behaviors bolster brand loyalty, while Anastasiei et al. (2023) demonstrate how emotional tone and word-of-mouth (WOM) significantly influence user reactions.
The rise of visual-first platforms like Instagram and TikTok has amplified the role of imagery and video. These formats attract attention rapidly and benefit from algorithmic amplification via likes, shares, and comments (B. Yang et al., 2023). Videos, in particular, evoke stronger emotional responses through multimodal stimulation (S. Yang et al., 2022).
Recent studies indicate that short-form videos in East Asian cultural spheres often embed emotional nationalism and soft-power storytelling, thereby transforming engagement metrics into ideological resonance (Fung & Hu, 2022; Zhang & Ma, 2023). Such dynamics complicate the notion that modality simply determines attention; rather, it co-constructs meaning within geopolitical narratives.
Hybrid content—integrating text, visuals, and video—has become standard in digital storytelling and crisis communication. Its multimodal nature supports clarity, emotional resonance, and virality, especially in time-sensitive contexts (Algiriyage et al., 2022). Hybridization also reflects the logic of convergence culture (Peng et al., 2023), where cross-platform circulation blurs boundaries between news and affect, professionalism and participation. In the Maritime Silk Road cities, hybrid formats often serve as cultural bridges, blending informational framing with heritage symbolism to attract both domestic and overseas audiences.
Beyond modality, content typology also shapes engagement. Hard news, soft news, and commentary each generate varying degrees of “buzz,” influencing second-level agenda-setting and editorial agendas (J. Kim et al., 2024).
Across contemporary media spaces, distinctions among these categories blur progressively (Perreault & Hackett, 2025). Elements of soft news—humor, personalization, visual storytelling—permeate hard-news genres, while commentary increasingly borrows from meme culture and short-form video aesthetics (Visperas, 2025). The permeability among genres underscores that “content type” is not a fixed attribute but a negotiated cultural form shaped by platform norms and audience practices (Bhatia et al., 2024).
In short, content type is a key determinant of digital communication strategy, but its influence derives from how it intersects with platform logic and sociocultural context. Aligning content affordances with participatory and emotional expectations enhances both visibility and discursive influence.
Hypothesis 2 (H2): Online voice and public engagement concerning Fuzhou, Xiamen, and Quanzhou differ significantly across news, discussion, and commentary content.
Posting Frequency and Online Voice
Posting frequency—the rate at which content is published—plays a critical role in shaping online voice. It influences content visibility, modulates engagement cycles, and affects the longevity of issue relevance in digital discourse (L. Yuan et al., 2023). Rather than functioning as a purely causal variable, posting frequency interacts dynamically with audience attention and algorithmic amplification, forming a mutually reinforcing rhythm within digital ecosystems (Milli et al., 2025).
Empirical research confirms that frequent posting generally enhances online voice, though its effects vary across platforms. On algorithm-driven feeds, high-frequency updates sustain visibility and trigger cumulative engagement (Rüther et al., 2023). Repetitive exposure increases interaction likelihood, reinforcing topic salience (Unnava & Aravindakshan, 2021). In saturated digital spaces, frequent content helps cut through ambient noise and maintain user attention (Lehrer et al., 2021).
However, the effect is not uniform. On news websites, multiple daily updates can prolong audience interest (Hasan et al., 2024), while on social platforms, excessive posting may lead to diminishing returns. Limniou et al. (2022) suggest that moderate posting paired with visuals is more effective on Facebook and Instagram. In contrast, Twitter’s real-time structure is more conducive to high-frequency posting (Vicente, 2023).
Frequency and engagement also reinforce each other. Increased posting often boosts user response, which is amplified by platform algorithms, creating a feedback loop (Voinea et al., 2024). Posting frequency and engagement function as part of an algorithmic feedback loop in which heightened user attention stimulates further content production, and vice versa (Milli et al., 2025; Voinea et al., 2024). This cyclical interaction blurs the direction of causality, reflecting a co-evolutionary relationship between audience activity and media supply. During major events, well-timed content bursts can extend a topic’s digital lifespan (Appel et al., 2020).
In sum, while frequent posting can elevate online voice, its effectiveness depends on platform logic and user expectations. Strategic calibration is essential to optimize impact without inducing audience fatigue.
Hypothesis 3 (H3): Higher posting frequency is associated with greater online voice concerning Fuzhou, Xiamen, and Quanzhou.
Time Periods, Online Voice, and Sentiment Trends
Temporal context is critical to understanding how online voice evolves, particularly within pre-test/post-test research designs. Comparing time periods helps reveal how major events and media interventions alter both the volume and tone of public discourse (Liu et al., 2024). For instance, Liu et al. (2024) documented a sharp rise in COVID-19 discussions during outbreak peaks, showing how event timing can intensify public attention and online discussion. In the present study, temporal comparison provides an empirical basis for tracing how issue salience and sentiment vary around major events.
Beyond volume, time influences the speed and trajectory of information flow. Malecki et al. (2021) observed that online voice tends to peak early in crises, then declines as public attention dissipates. Such temporal rhythms provide a basis for comparing measured engagement indicators across event windows, including discourse volume, interaction rates, and sentiment trajectories.
In the cross-strait media environment, these temporal rhythms intersect with regionally symbolic events concentrated in Fujian Province, particularly in the Maritime Silk Road cities of Fuzhou, Xiamen, and Quanzhou. These cities frequently host policy forums, cultural exchanges, and tourism campaigns that serve as communicative nodes linking Mainland China and Taiwan (K.-M. Chang, 2024; Tsui, 2025). Their prominence during such periods reflects how historical and cultural proximity translates into recurring media visibility.
Sentiment dynamics also shift across time. Initial stages of crises often generate emotional extremes, while later phases reflect stabilization or emotional neutrality (Gu et al., 2022). These patterns are both shaped by and reflected in online voice trends (Sun et al., 2023). Platform differences matter: Twitter (now X) often exhibits early emotional surges, whereas Facebook supports more sustained and moderated discourse (Islam et al., 2024).
Taken together, these temporal dynamics underscore the importance of situating online voice within time-sensitive and regionally contextual frameworks, where the cross-strait salience of Fujian’s coastal cities provides both the temporal triggers and cultural anchors of digital discourse.
Hypothesis 4 (H4): Across different time periods (pre-test vs. post-test), external events and media interventions significantly shape online voice and sentiment trends related to Fuzhou, Xiamen, and Quanzhou, with region-specific variations.
Platform Interaction and the Dynamics of Online Voice
Digital platforms—news media, social media, and online forums—exhibit distinct dissemination logics that, when intersecting with temporal and regional contexts, jointly shape online voice and agenda-setting dynamics.
News media typically initiate information flows through institutional authority but offer limited interactivity (Obasi, 2024). Social media, by contrast, amplify discourse through user-generated content and algorithmic acceleration, especially during crises (Andersen et al., 2024). Forums contribute fewer messages but support sustained, in-depth discussions within niche communities (Pärli & Fischer, 2020).
These dynamics are temporally layered. News platforms often generate initial post-event spikes, while social media and forums prolong engagement via real-time updates and ongoing discussion (Damayanti et al., 2024). Platform-specific rhythms contribute differently to the lifecycle of issue salience.
Regional preferences further modulate dissemination. In Fuzhou, institutional media remain dominant, whereas in Xiamen and Quanzhou, social platforms are more influential—reflecting variation in digital literacy, infrastructural development, and communicative norms (Apostol & Hernández-Rodríguez, 2024; Nguyen, 2024).
Importantly, platforms do not operate in isolation. Cross-platform synergy enhances both the reach and resonance of messages, especially during critical events (Fischer et al., 2024). In moments of heightened public attention, major events function as temporal anchors that synchronize media agendas and audience engagement across platforms (Einarsson et al., 2024). Such events often trigger bursts of multi-platform activity—news outlets initiate coverage, social media accelerate emotional amplification, and forums sustain interpretive discussion—forming a cascading cycle of agenda interaction (Hase et al., 2022; Lepird et al., 2024; Zeib & Shahzad, 2025). Defining these events through measurable indicators such as cross-platform exposure or discussion volume enables a quasi-experimental observation of how digital salience evolves in real time. Yet the dynamic interplay between platform type, temporal staging, and regional variation remains underexplored.
Research Question 1 (RQ1): How do interactions between platform type, time, and regional context shape patterns of online voice dissemination?
Material and methods
Research Design
This study employs a quantitative pre-test/post-test design to assess how platform type, content format, posting frequency, and event timing influence online discourse and sentiment surrounding the three historically connected coastal cities of Fuzhou, Xiamen, and Quanzhou. It uses event-based observation windows to track behavioral changes across key cross-strait communication episodes.
These three cities were selected because they constitute historically interconnected coastal nodes in Fujian and sustain deep cultural, demographic, and communicative ties with Taiwan. More than eighty percent of Taiwan’s population descends from migrants originating in Fujian’s coastal regions—particularly Fuzhou, Xiamen, and Quanzhou—whose linguistic, religious, and social traditions, such as the Mazu belief, continue to shape cross-strait cultural affinity. These long-standing genealogical and cultural linkages have established a distinctive foundation for sustained media visibility and public resonance between these cities and Taiwanese society (S.-H. Chen, 1965; Lin, 2002). In contemporary contexts, these cities remain recurrent reference points in Taiwanese news coverage and online discussions due to their high levels of cross-strait public visibility and issue relevance. For instance, the Jinjiang–Kinmen water-supply project has attracted extensive media attention and online discussion in Taiwan, not primarily because of its policy attributes, but because it directly intersects with everyday concerns, regional proximity, and shared social imaginaries across the Taiwan Strait (M. H. Chen, 2024). Moreover, Xiamen’s Maritime Silk Road Forum and Quanzhou’s World Heritage promotion activities have repeatedly drawn Taiwanese media coverage and participant engagement, rendering these cities empirically observable arenas of cross-platform discourse rather than abstract policy symbols (S. Yuan, 2025). Together, these historical, cultural, and communicative linkages make Fuzhou, Xiamen, and Quanzhou analytically meaningful cases for examining how urban heritage, media salience, and digital discourse converge within Taiwan’s online public sphere.
Data Sources and Collection
Data were collected from 2,962 digital sources across Taiwan. Here, “digital sources” refer to identifiable data collection units—such as news outlets, forum boards, public pages, channels, or discussion spaces—rather than distinct platforms. These sources span multiple platform categories including news websites (e.g., Yahoo! News, Line Today, SET News, ETtoday), discussion forums (e.g., PTT, Dcard, Mobile01), and social media platforms (e.g., Facebook, Instagram, and Line). The dataset focuses on Taiwanese public discourse surrounding Fujian’s coastal cities, as captured through domestic digital platforms with the highest user penetration in Taiwan (Taiwan Network Information Center, 2023). The analysis reflects Taiwanese audience behavior and media habits (W.-C. Chang, 2017).
Both official and user-generated content—including news articles, forum threads, comments, posts, and multimedia materials (text, images, and videos)—were analyzed. Data acquisition combined direct API access and large-scale web scraping, using 217 curated keywords in Mandarin and Taiwanese, conducted with technical support from a Taiwan-based third-party analytics firm specializing in digital discourse monitoring. The firm is described generically to maintain commercial neutrality and avoid implying endorsement. It provided technical support for data crawling and preliminary classification, while keyword construction, analytical design, statistical modeling, and interpretation were conducted independently by the authors.
Data Cleaning and Validation
To ensure data integrity, duplicates, spam, and irrelevant commercial or non-Chinese content were removed. Posts were algorithmically categorized by platform and content type using automated classifiers. Data validation relied on a rule-based and iterative inspection strategy rather than probabilistic sampling. Keyword lists, platform classifications, and content labels were repeatedly examined and refined through manual inspection of high-salience posts during event peaks.
Event Window Design
The baseline period (September–November 2018) aligns with Taiwan’s local elections. To enhance analytic transparency, “major events” were operationally defined using observable indicators of cross-platform visibility and public salience. Events were included when they attracted substantial attention across both institutional and user-generated media—typically reflected in extensive news coverage, reposting volume, or discussion volume exceeding 1,000 items within the relevant event-month observation window, an approximately four-week period. This criterion ensured that selected events represented moments of demonstrable public relevance within Taiwan’s digital sphere. Under this definition, the Straits Forum (2019 and 2020 sessions), Typhoon Lekima (2019), and the Kinmen–Jinjiang water-supply project announcement (2020) each qualified as major cross-strait communication episodes, given their sustained public visibility and cross-platform engagement within Taiwan’s digital sphere (K.-M. Chang, 2024; M. H. Chen, 2024; Wu et al., 2021). The post-test phase (April 2019–April 2021) covers major events, including the Straits Forums, Typhoon Lekima, and infrastructure and transportation policy announcements. These events were identified through monthly trend analysis and high-volume topic patterns within the broader post-test observation period. The relevant event-month window was used as a temporal anchor for comparing discourse patterns and interpreting changes in online voice and sentiment.
To mitigate potential reverse causality, the time-window design also enables temporal ordering between posting frequency and online voice, allowing the analysis to approximate causal direction rather than simultaneity.
Sentiment Analysis
Sentiment polarity (positive, neutral, negative) was assessed using Jieba and SnowNLP, with customized dictionaries reflecting Taiwan-specific lexicon and sarcasm. Parameters such as token frequency and context window were optimized for contextual accuracy. The sentiment dictionary incorporated region-specific expressions and sarcasm markers derived from Taiwan’s online forums and commentaries to reduce misclassification in ironic or political contexts. To assess contextual adequacy, automated sentiment outputs were iteratively checked against manually reviewed high-impact posts during key event periods, with particular attention to sarcasm and region-specific expressions.
Analytical Methods
Quantitative analyses were conducted using SPSS 26.0 and R 4.3.0. In this study, mentions refer to the total count of messages or posts in which the names of the target cities—Fuzhou, Xiamen, and Quanzhou—appeared within the observation period, indicating the raw volume of discourse. Online voice was operationalized as a composite indicator combining the number of original posts with cumulative user interactions (comments, shares, replies, and reactions), reflecting both discursive visibility and participatory intensity (Baqir et al., 2025; Trunfio & Rossi, 2021).
Descriptive statistics examined the distribution of discourse volume across cities and platforms, and multiple regression was used to assess the effects of platform type, content format and posting frequency. Time-series analysis traced temporal fluctuations within defined event windows, and sentiment trajectories were compared across pre- and post-event stages.
Scope and Limitations
This study examines online discourse related to Fujian’s coastal cities within Taiwan’s digital environment, emphasizing cross-strait communication dynamics. Several methodological constraints should be acknowledged, including API access restrictions, evolving platform algorithms, and linguistic ambiguities such as sarcasm or code-switching. While triangulation and qualitative inspection were employed to enhance data reliability, future research could broaden platform coverage, refine linguistic models, and incorporate ethnographic or interview-based approaches to deepen interpretive insight.
Results
The double-coding validation achieved over 80% agreement between human and automated classifications, confirming the reliability of sentiment analysis prior to hypothesis testing. To test Hypothesis 1, this study examined whether platform type—namely news media, forums, and social media—significantly influenced the online voice related to Fuzhou, Xiamen, and Quanzhou.
Descriptive statistics reveal that forums dominated the digital discourse, accounting for 97,691 mentions—almost twice that of news outlets and more than seven times that of social media. Forums also showed the highest average reply rate (16.60), underscoring their role in fostering sustained and interactive discussions. While news media contributed approximately 30% of total mentions, they exhibited the lowest interaction rate (0.23), reflecting their more top-down, broadcast-oriented function. Social media platforms, although contributing only 8% of the total volume, demonstrated a relatively high average interaction (6.10), indicating their value in stimulating participatory engagement rather than volume-based visibility (see Table 1).
A one-way ANOVA confirmed a significant effect of platform type on online voice (F = 1.43 × 10³¹, p < .001, η² =.47), with nearly half the variance explained by platform differences (see Table 2). Although the F-statistics appear extremely large due to the high sample size and low within-group variance, this does not indicate model misspecification. In large-scale digital datasets, even small between-group differences can yield large F-values. Therefore, effect size (η²) rather than F magnitude was used as the primary basis for interpretation, following conventional reporting standards for big-data communication research (Cohen, 1988; Field, 2024). Tukey HSD post hoc comparisons indicated that forums generated significantly more online voice than both news and social media (p < .001). Additionally, news media significantly outperformed social media in volume (p < .001) (see Table 3).
These findings offer strong empirical support for H1. Among the three platform types, forums emerged as the most influential in amplifying online voice, owing to their high interactivity and sustained engagement. In contrast, news media served a more unidirectional, authoritative role, while social media facilitated moderate yet participatory discourse. Together, these platforms formed a differentiated yet complementary ecosystem of digital communication.
Hypothesis 2 proposed that online voice and public engagement would differ across three operationally defined content categories: news, discussion, and commentary. News refers to information-oriented reports produced by news outlets or institutional sources. Discussion refers to user-generated posts or threads that invite exchange, replies, or shared information. Commentary refers to evaluative or opinion-oriented content that interprets events, expresses judgments, or takes a stance.
Descriptive statistics indicate that discussion-based content had the greatest impact, generating 74,019 mentions and the highest average engagement (mean replies = 11.24), suggesting its strong capacity to sustain interactive dialogue. News articles, defined as information-oriented reports produced by news outlets or institutional sources, represented 41% of total mentions (62,473) but showed minimal engagement (0.25 replies), underscoring their primarily one-way communication function. Commentary content, defined as evaluative or opinion-oriented material that interprets events, expresses judgments, or takes a stance, was coded by communicative function rather than by platform source; it could therefore appear in news outlets, institutional sources, forums, or social media. This category was the least frequent (15,462 mentions) but achieved a relatively high average engagement rate (6.68). This indicates that commentary content can provoke strong audience reactions despite its more limited reach (see Table 4).
When the three target cities were compared as subjects of online discussion, content patterns differed across Fuzhou, Xiamen, and Quanzhou. News content was most prevalent in discourse about Fuzhou, discussion-based content dominated discourse about Xiamen, and commentary content achieved higher engagement in discourse about Quanzhou. These patterns indicate city-specific differences in how the three urban subjects were represented and engaged with on Taiwan-based digital platforms.
A one-way ANOVA revealed a highly significant effect of content type on online voice (F = 1.97 × 10³¹, p < .001, η² =.53), with more than half of the variance explained by content differences (see Table 5). As in other analyses, the extremely large F-value reflects the very large sample size and minimal within-group variance rather than any model irregularity. Effect size (η²) was therefore emphasized as the primary indicator of substantive influence. Post hoc comparisons via Tukey HSD confirmed that discussion posts significantly outperformed both news and commentaries in driving online voice (p < .001). News content also yielded significantly more mentions than commentary (p < .001) (see Table 6).
These findings strongly support H2. Discussion-oriented posts not only amplified message diffusion but also maximized user interaction. Meanwhile, news and commentary formats fulfilled more specialized roles—either broadcasting at scale or fostering targeted emotional engagement. Together, these content types underscore the layered nature of digital discourse.
Hypothesis 3 posited that high-frequency posting would significantly influence online voice across platforms. The results reveal distinct platform-specific patterns.
News outlets increased posting frequency by 27.44%, resulting in a moderate rise in mentions (from 48,355 to 52,779) but only a marginal increase in engagement (from 0.25 to 0.29 average replies), suggesting enhanced visibility with limited interaction. Forums experienced a 25.64% rise in frequency, translating to a 21.21% growth in mentions and a notable boost in engagement (from 16.60 to 18.15 average replies), demonstrating their responsiveness to posting volume. Social media saw the largest increase in posting frequency (43.14%) and a modest engagement gain (from 6.10 to 7.30), but total online voice remained relatively limited—highlighting strong interactivity without extensive diffusion (see Table 7).
At the regional level, Fuzhou exhibited the highest news output but limited engagement, suggesting a consumption-oriented model. Xiamen showed growth in social media activity, while Quanzhou’s strongest gains occurred in forum-based discourse, reflecting localized digital behaviors.
Two-way ANOVA results confirmed that both platform type (η² =.49) and posting frequency (η² =.42) had statistically significant effects on online voice. A significant interaction effect was also observed (p = .004), indicating that the influence of posting frequency varied by platform (see Table 8). Consistent with other results, the magnitude of the F-statistics stems from the dataset’s scale and low within-group variability. Interpretation thus centers on η² values, which provide a more stable measure of effect strength. Tukey HSD post hoc tests revealed that forums significantly outperformed both news and social media under high-frequency conditions (p < .001), while social media exhibited greater engagement than news but still lagged behind forums in reach (see Table 9).
Overall, the findings support H3. Increased posting frequency was associated with higher online voice across platforms, but the size and form of this association varied by platform affordances. Forums were most responsive to increased posting frequency, showing higher mentions and average replies than news media and social media. News platforms produced relatively high mention volume but low average replies, whereas social media showed higher average replies than news media while remaining lower in total online voice. These dynamics were further shaped by city-specific discourse patterns, underscoring the need for context-sensitive posting strategies.
Hypothesis 4 examined how time-based external events influence online voice and public sentiment. From pretest to posttest, discourse volume increased across platforms—news by 9.1%, forums by 21.2%, and social media by 22.7%. Despite the rise in volume, engagement patterns remained stable: news media continued to function as low-interaction broadcast channels, forums maintained high engagement, and social media saw a slight increase in interaction rate (from 6.10 to 7.30) without a corresponding gain in volume, indicating limited capacity for sustained dialogue.
Regionally, Xiamen recorded the largest increase in online voice (+29.1%), followed by Quanzhou (+27.9%) and Fuzhou (+27.4%). Changes in average interaction rates varied across the three city-related discourse categories: Fuzhou increased slightly from 0.29 to 0.31, while Xiamen declined from 0.54 to 0.51 and Quanzhou declined from 0.38 to 0.35 (see Table 10).
In terms of sentiment, all three cities showed an increase in neutral tone (by 2.1% to 2.5%), while both positive and negative sentiments declined slightly, indicating a trend toward more rationalized discourse. Forums saw the sharpest rise in negativity (+5.0%), possibly reflecting intensified debate. News platforms shifted toward neutrality (+2.3%), and social media remained the most emotionally reactive (see Table 11). Among cities, Fuzhou moved most clearly toward neutrality, Xiamen recorded the largest drop in positivity (–2.5%), and Quanzhou sustained higher levels of negativity (see Table 12).
Multiple regression analysis confirmed that the posttest period significantly predicted increases in online voice (B = 9,473.62, p < .001, η² = 0.38), with forums generating the highest volume. Negative sentiment also emerged as a significant predictor of increased discourse (p = .018, η² = 0.27) (see Table 13). Paired-sample t-tests supported these findings: online voice significantly increased (p = .021), while sentiment shifted from positive to neutral (p = .027), indicating a marked transition in emotional tone from pre-test (M = 9,867.00) to post-test (M = 12,671.67), t(2) = 6.78, p = .021, η² = 0.95, 95% CI [972.42, 4893.25] (see Table 14).
In sum, the findings support H4. Time-sensitive external stimuli not only amplify discourse volume but also shift its emotional composition. Forums remain the most sensitive to these changes, intensifying opinionated discussions. News platforms exhibit a stabilizing neutral trend, while social media sustains reactive sentiment with emotional fluctuations. Regional patterns reinforce the value of localized strategies for public engagement during media events.
Research Question 1 explored how platform interaction shapes the dissemination of online voice, measured as a composite of original posts and cumulative user interactions, while accounting for temporal and city-related variation. Post-test results showed that news media generated the largest share of total online voice (41.95%) but maintained the lowest engagement rate (.25), underscoring its role as a unidirectional broadcasting channel. In contrast, social media exhibited a higher engagement rate (.55), and forums demonstrated the strongest interaction (.81), reinforcing their position as participatory and deliberative arenas (see Table 15).
Two-way ANOVA results confirmed that both platform type and time significantly influenced online voice (p < .001), with a notable interaction effect (p < .001), suggesting platform impact varied across time periods. Social media and forums recorded faster growth in the post-test phase (see Table 16).
OLS regression analysis further supported the influence of time on online discourse (B = 16,810, p < .001), indicating that external developments heightened public engagement. While news media retained the highest absolute voice, both social media (B = –8,635, p < .001) and forums (B = –16,640, p < .001) proved more effective in facilitating user interaction. Regionally, Fuzhou demonstrated significantly higher discourse volume than Xiamen (B = 7,785.61, p < .001), implying greater reliance on institutional media channels. In contrast, Xiamen exhibited more platform-based participatory behavior (see Table 17).
Interpretive Note on Contextual Confounders
While the statistical results are robust, the extended pre-/post-test period (late 2018–early 2021) may have introduced unmeasured contextual influences, such as algorithmic shifts, policy changes, or evolving digital habits. Although event-based segmentation helped reduce potential confounding, residual biases may persist. Therefore, findings should be interpreted as indicative rather than strictly causal, particularly within the dynamic environment of cross-platform communication.
Discussion
This study examines how platform type, content format, posting frequency, and temporal dynamics shape online voice and sentiment across platforms and regions. The findings are interpreted in relation to the communicative role of historically connected Fujian cities in Taiwanese online discourse, with particular attention to cultural proximity, media visibility, and cross-platform dynamics. In this sense, the analysis speaks primarily to the mediated construction of urban discourse rather than to direct evaluations of public policy.
Findings suggest that news media, forums, and social platforms play distinct yet complementary roles. News platforms remain central to institutional visibility—high in visibility but low in interaction—reflecting a predominantly one-way model (Al-Quran, 2022). Forums foster peer-based dialogue and sustained discussion, especially on shared concerns (Christian, 2024). Social media amplify reach via algorithmic spread, though often within short-lived cycles due to rapid content turnover (Jerez-Villota et al., 2025).
These findings extend McCombs and Shaw’s (1972) agenda-setting theory into a multi-platform environment where salience formation is no longer hierarchical but recursive. Rather than a single agenda imposed by institutional media, issue visibility now emerges through iterative exchanges between top-down and participatory communication channels (B. Kim et al., 2024; Su & Xiao, 2024). This challenges the classical assumption that mainstream media act as sole agenda-setters, suggesting instead that users collectively negotiate prominence through dialogic and algorithmically mediated feedback (Einarsson et al., 2024; Yoo, 2024).
Moreover, the strong role of forums complicates the notion of them as purely “deliberative” spaces; their sustained engagement and occasional conflict point to what Papacharissi (2015) describes as “affective publics,” where emotional tension and civic debate coexist.
This differentiation reveals a sequential flow: news initiates, social media amplify, and forums sustain discourse. Public discussion often migrates from forums to social media, sometimes prompting renewed media coverage. While this dynamic deviates from classical agenda-setting, it highlights cross-platform influence in a nonlinear, feedback-driven digital ecosystem. Importantly, these platform-specific patterns should not be interpreted as universal features of all urban topics. Rather, they emerge in relation to cities characterized by dense historical, cultural, and communicative linkages to Taiwanese society.
Second, content type significantly shapes both the scale and quality of online discourse across platforms and regions. Aligning with prior studies, discussion-oriented content yielded the highest engagement and widest dissemination, reinforcing its role as a driver of participatory communication. While Shahbaznezhad et al. (2022) stress its dialogic nature, this study further shows its adaptability across varying platform logics and regional user behaviors.
These findings extend Shugars and Ha (2025) dialogic communication framework by demonstrating that dialogue in digital spaces is not merely interpersonal but structurally mediated through platform affordances. In the Taiwanese online sphere, dialogic participation often intertwines civic deliberation with affective expression, forming a mode of hybrid civic engagement that spans institutional and participatory spaces. Rather than assuming that participatory dialogue is inherently rational or consensus-oriented, such engagement often operates through emotional resonance and situational alignment across multiple platforms (Humprecht et al., 2024; J. Kim et al., 2024; Leppert et al., 2022).
Moreover, the adaptability of discussion-oriented content across regions reveals the contextual negotiation of communication norms—where informal, peer-driven exchanges can achieve greater legitimacy and reach than institutional narratives.
In contrast, news content—though prevalent—remains anchored in a one-way communication model, consistent with institutional media norms (Blassnig & Wirz, 2025). Commentary-based content, while less far-reaching, achieved higher engagement-per-post, highlighting its capacity to trigger emotional or polarizing exchanges (Kitchens et al., 2020). Crucially, the findings suggest that interactive content, rather than purely informational forms, plays a formative role in shaping downstream discourse, indicating a shift toward more decentralized, user-driven engagement.
Third, increased posting frequency significantly boosts online voice, but its impact varies by platform and regional behavior. Forums show strong amplification in both volume and interaction under high-frequency conditions, reinforcing their role as discursive hubs. This supports the view that asynchronous, text-based environments foster layered, iterative dialogue over time (Unnava & Aravindakshan, 2021). The results further suggest that message visibility is contingent on temporal intensity and platform algorithms that privilege recency and engagement metrics (Einarsson et al., 2024; B. Kim et al., 2024). However, the divergent effects across platforms indicate that frequency alone cannot sustain attention without contextual or affective relevance. Social media’s rapid turnover exemplifies what Crary (2022) terms “24/7 temporality,” where visibility is fleeting and engagement is optimized for recency rather than depth. In contrast, forums’ slower, accumulative rhythm reflects a counter-temporality of sustained deliberation, allowing discursive continuity beyond algorithmic cycles.
Meanwhile, social media sees short-term engagement spikes that quickly level off, echoing Lehrer et al.'s (2021) argument that these platforms favor trend-driven, time-sensitive content over sustained framing. News media, despite output growth, yield only modest interaction gains, consistent with their broadcast-oriented logic. City-specific patterns further illustrate contextual mediation. In discourse about Fuzhou, higher news volume was accompanied by limited interaction. Discourse about Xiamen showed stronger responses to frequent social media posts, while discourse about Quanzhou remained forum-centric, favoring cumulative and user-led discussions. These regional divergences also reveal how platform temporality interacts with cultural media practices. In regions like Quanzhou, where community-based online participation is historically rooted, iterative posting cultivates trust and collective memory, unlike the ephemeral virality typical of social media. Thus, posting frequency functions less as a technical parameter and more as a socio-temporal negotiation between algorithmic rhythms and audience habits.
Fourth, the results indicate that major events and the public attention surrounding them were associated with changes in online voice and sentiment across platforms and city-related discourse categories (Sun et al., 2023). News media showed modest volume growth but remained low in average interaction, a pattern consistent with research on conditional engagement with traditional news (Espeland, 2024). By comparison, forums and social media recorded higher average interaction rates, suggesting that participatory platforms carried a larger share of post-event interaction. This pattern points to a shift toward sustained, user-driven discourse, particularly through continued replies and forum discussions after the initial event peak. These results illustrate how platform algorithms and audience affectivity interact in governing visibility and attention flows across digital media (Einarsson et al., 2024; Gillespie, 2018). During high-salience moments, emotional content—particularly negative affect—tends to receive algorithmic reinforcement, amplifying both visibility and contagion (Bakir & McStay, 2017; Humprecht et al., 2024). As emotional intensity stabilizes, however, platforms often recalibrate ranking systems toward neutral or informational content, leading to a process of affective normalization (Wahl-Jorgensen, 2025). This cyclical pattern transforms crises from immediate emotional reactions into more reflective public discourse, illustrating how algorithms modulate collective emotion over time.
The observed post-event neutrality, therefore, does not signal emotional disengagement but an adaptive recalibration of public discourse across platforms. Forums’ rising negativity underscores their dual role as deliberative and contentious arenas—spaces where affective critique and civic reasoning coexist (Papacharissi, 2015).
Collectively, these dynamics reveal that emotional trajectories in digital discourse are not spontaneous but structurally mediated through temporal algorithms and regional communication ecologies. By tracing how neutrality and negativity evolve across time and context, this study extends crisis communication research into a cross-platform, temporally adaptive model of emotion circulation.
Finally, the analysis shows that platform interactions—conditioned by time and region—exert distinct influences on online voice dynamics. News media remain key sources of verified information but consistently show low interaction, reaffirming their role in agenda-setting without fostering dialogue (Blassnig & Wirz, 2025). These findings extend agenda-setting theory into a cross-platform ecology, where influence is negotiated through iterative feedback rather than linear transmission (Su & Xiao, 2024). Interactions between institutional media and participatory channels form associative issue networks in which issue salience emerges from content circulating recursively across platforms, with algorithmic systems modulating visibility along the way (Einarsson et al., 2024; Ginossar et al., 2022; B. Kim et al., 2024; Wang et al., 2023). Social media’s algorithmic acceleration transforms traditional top-down flows into decentralized, time-sensitive feedback loops, while forums provide the deliberative depth necessary for narrative stabilization.
Social media, driven by algorithmic curation, trigger short-term engagement spikes but rarely sustain thematic continuity (Metzler & Garcia, 2024). Forums, though contributing less to overall volume, generate the highest interaction rates, highlighting their role in prolonged, grassroots deliberation (Jerez-Villota et al., 2025).
Temporal variation is particularly salient: post-event periods consistently show spikes in volume and engagement, suggesting media salience and sociopolitical stimuli as external triggers (Sun et al., 2023). These effects are regionally uneven—Fuzhou leans on traditional flows, while Xiamen and Quanzhou favor participatory dynamics via social platforms and forums.
Fuzhou, Xiamen, and Quanzhou are meaningful cases for examining cross-strait digital perception because their historical and symbolic links to Taiwan make them recognizable urban references within maritime exchange narratives. The findings also carry indirect implications for understanding how the Maritime Silk Road may acquire meaning in digital communication. Within the Taiwan-based dataset analyzed here, discourse about these three cities showed recurring visibility across news media, social platforms, and forums. This observed visibility suggests that historically resonant nodal cities can function as mediated entry points through which broader policy-related meanings are selectively recognized, amplified, or contested. In this regard, the Maritime Silk Road is articulated through urban narratives, affective cues, and platform-specific patterns of engagement associated with Fuzhou, Xiamen, and Quanzhou.
Collectively, these patterns illustrate what this study conceptualizes as a cross-platform deliberative cycle—a dynamic process in which issue visibility develops through interactions among institutional media, platform environments, and participatory spaces. Drawing on communication-ecology theory, this formulation positions online voice within an interdependent system of reciprocal influence shaped by temporal rhythms, platform-specific visibility patterns, and city-related discourse differences. This cross-platform deliberative cycle contributes a conceptual lens for understanding how deliberation and visibility co-evolve across news media, social platforms, and forums.
Taken together, the findings demonstrate a coherent theoretical linkage between platform logic, affective communication, agenda-setting dynamics, and online voice. Differences in platform affordances and platform-specific visibility patterns shape the emotional tone and interaction patterns of discourse, which in turn influence how issue salience is collectively constructed across platforms. Rather than operating in isolation, affective communication and agenda-setting function as interdependent processes through which visibility, participation, and emotional resonance are negotiated. Online voice thus emerges not as a static measure, but as a dynamic outcome of cross-platform interactions conditioned by platform-specific logics and temporal rhythms. By empirically tracing these connections, the study moves beyond treating online voice, affect, and agenda-setting as separate analytical components and instead demonstrates how they operate as an integrated communication ecology within Taiwan’s digital public sphere.
In sum, the configuration of online voice reflects the intersection of platform logic, content strategy, temporal rhythm, and regional media culture. Rather than operating as discrete variables, these dimensions interact dynamically to constitute a cross-platform communication ecology in which discourse circulates through recursive feedback among institutional media, algorithms, and user communities. Understanding this interplay reframes agenda-setting in the digital era—not as a top-down transfer of salience, but as a distributed process of negotiation shaped by contextual factors, affective dynamics, and technological affordances.
Conclusion
This study investigates how platform-specific strategies—spanning content formats, posting frequency, and temporal responsiveness—shape online voice and sentiment across three Maritime Silk Road cities: Fuzhou, Xiamen, and Quanzhou. Our findings reveal that news media, forums, and social platforms serve differentiated yet interdependent roles. In the event windows analyzed here, news media generated substantial institutional visibility but retained low interaction, a pattern consistent with their relatively unidirectional model of online news engagement (Blassnig & Wirz, 2025). Forums foster participatory deliberation and sustained issue framing (Christian, 2024), while social media drive rapid but fleeting attention cycles shaped by algorithmic volatility (Metzler & Garcia, 2024).
Platform affordances interact with posting frequency in nuanced ways: higher posting enhances forum engagement but yields diminishing returns for news media, underscoring the limits of top-down communication (Shahbaznezhad et al., 2022). Commentary-based content spurs niche engagement, while discussion-driven formats achieve broader resonance. This shift suggests a reconfiguration of agenda-setting, driven more by user participation than institutional control (Uth et al., 2023).
Beyond empirical observations, this study advances the theoretical scope of agenda-setting by conceptualizing the cross-platform agenda cycling model—a dynamic process in which issue salience emerges through iterative feedback between institutional media, social platforms, and participatory forums. Unlike classical models that assume a linear media–audience relationship, this framework highlights multidirectional agenda negotiation driven by algorithmic amplification, user interaction, and regional media cultures. This study thereby extends classical agenda-setting theory into a networked, recursive model of agenda formation that captures how digital platforms collectively initiate, amplify, and sustain public discourse across time. By situating agenda-setting within a hybrid communication ecology, the study extends existing theory toward a more relational and temporal understanding of digital discourse formation.
Temporal analysis shows that events like policy announcements and cross-strait cultural exchanges trigger spikes in volume and sentiment shifts. Forums exhibit sustained negative sentiment (Achuthan et al., 2025), social media show emotional volatility (Knutson et al., 2024), and news media trend toward neutral tones—highlighting how affective responses co-evolve with platform logics. Although the temporal comparison reduces simultaneity bias, the causal direction between posting frequency and online voice remains open to longitudinal verification.
Limitations include residual sociopolitical confounding, the exclusion of short-video platforms such as TikTok and Douyin, and a lack of demographic granularity. During the study period (2018–2021), short-video applications had not yet become major venues for news or cross-strait public discourse in Taiwan, where engagement was still concentrated on Line, Facebook, and YouTube. Consequently, the omission of these platforms constrains the analysis of emerging audiovisual forms of agenda-setting. Even so, this study advances agenda-setting theory by revealing hybrid influence structures within digital media ecosystems. It also offers practical insight for civic communicators and policymakers aiming to optimize message delivery, audience targeting, and emotional tone. Future research should employ cross-platform tracking and algorithmic impact modeling across geopolitical contexts.
Funding
This study was supported by the 2025 Fujian Social Science Foundation Project “Research on Cultural Communication and New Media Marketing in Fujian along the Maritime Silk Road” (Project No. FJ2025T009).