A Cross-Platform Public Opinion Dynamic Assessment Method and System Based on Real-Time Semantic Awareness

Abstract

This invention discloses a cross-platform dynamic evaluation method and system for public opinion based on real-time semantic perception, belonging to the field of online public opinion data processing. The method includes the following steps: collecting multi-platform, multi-modal, heterogeneous data; standardizing the data; semantically aggregating the standardized cross-platform multi-modal data corresponding to each propagation event based on an event semantic center vector model; iteratively updating the event semantic center vector through online exponential sliding updates to address the dynamic evolution of propagation events, continuously supplementing the event semantic center vector with newly added standardized data related to the propagation event in real time; quantitatively evaluating the comprehensive performance of each propagation event on different platforms based on a platform performance index model, and constructing a cross-platform content ranking based on the comprehensive performance using a multi-attribute normalized decision function; and intuitively presenting the cross-platform online public opinion analysis results using a three-level responsive visualization view.

Description

Technical Field

[0001] This invention relates to the field of online public opinion data processing, and in particular to a cross-platform dynamic evaluation method and system for public opinion based on real-time semantic awareness. Background Technology

[0002] Currently, significant progress has been made in the monitoring and analysis of public opinion on social media. Existing solutions generally cover single or multiple data sources and possess basic sentiment analysis, hotspot detection, and early warning functions. However, when addressing the practical needs of cross-platform, comprehensive public opinion analysis, existing technological systems still have the following key limitations: At the data integration level, the depth of data integration is insufficient, exhibiting aggregation rather than fusion. Most systems can collect and aggregate data from multiple platforms in parallel, but fail to achieve deep correlation centered on events or topics. Data from different platforms is isolated, resulting in fragmented analysis results. Users struggle to gain insights into the overall popularity of a specific event across all platforms, differences in sentiment, and the distribution of core dissemination content; the analysis remains merely a simple listing of data from various platforms.

[0003] At the analytical dimension level, the analytical perspective is limited to a single point and lacks an effective horizontal comparison mechanism. Existing tools are designed with an emphasis on in-depth analysis of a single platform or a single dimension, lacking cross-platform horizontal comparison analysis modules and visualization methods. For example, the system cannot intuitively present the quantitative relationship between the depth and breadth of discussion of the same topic on different platforms, nor can it automatically identify and display the key differences in content that triggers topics on different platforms, thus failing to accurately grasp the cross-platform resonance patterns and key nodes.

[0004] In terms of interactive experience, the system primarily relies on static reports, lacking real-time, exploratory, and panoramic visualization. Existing systems mainly output periodic reports or fixed data dashboards with weak interactivity. In particular, they lack interactive visualization interfaces that can guide users from a macro-level overview of the overall online situation to a micro-level cross-platform comparison, and allow for one-click tracing of original information sources. This makes it difficult for decision-makers to conduct efficient and independent exploratory analysis when faced with sudden and complex cross-platform public opinion events, resulting in a lengthy and disjointed path from discovering phenomena to tracing their root causes.

[0005] In summary, existing technologies have significant shortcomings in achieving semantic-level fusion, systematic comparison, and immersive interactive analysis of cross-platform social media data. Therefore, there is an urgent need for a new system that can break down data barriers across multiple platforms, construct an event-centric fusion analysis model, and provide users with a one-stop, traceable, and comprehensive cross-platform insight through an innovative visual interactive framework. Summary of the Invention

[0006] This invention provides a method and system for cross-platform public opinion dynamic assessment based on real-time semantic awareness. It aims to overcome the fragmentation and lack of quantitative assessment in existing technologies for cross-platform dissemination analysis, and to provide an engineered dissemination effectiveness evaluation system. This system upgrades from simple data aggregation to in-depth quantitative assessment of dissemination effectiveness through the fusion analysis and visualization interaction of dissemination data from multiple platforms. Specific objectives are as follows: Constructing a measurable cross-dissemination data foundation; building an automated data pipeline to periodically collect data from multiple platforms; achieving automatic clustering and fusion of cross-platform content based on semantic analysis; forming a unified data view centered on dissemination events, providing an accurate and consistent data foundation for quantitative evaluation; providing a panoramic and comparable effectiveness evaluation tool; designing a three-level visualization interaction system; transforming multi-dimensional effectiveness indicators into intuitively comparable and interactively explorable charts, making dissemination effect assessment clear; achieving closed-loop verification capabilities from evaluation to source tracing; integrating one-click drill-down and preview functions; forming a complete analysis loop; and improving the depth of evaluation and the credibility of decision support.

[0007] On the one hand, a method for dynamic evaluation of cross-platform public opinion based on real-time semantic awareness is provided. This method includes: collecting multi-platform multimodal heterogeneous data, standardizing the data, semantically aggregating the standardized cross-platform multimodal data corresponding to each propagation event based on the event semantic center vector model, and establishing a unique mathematical representation for each propagation event; iteratively updating the event semantic center vector through online exponential sliding updates for the dynamic evolution of propagation events, continuously supplementing the event semantic center vector with newly added standardized data related to the propagation event in real time, so that the mathematical representation of the event is synchronized with the latest data; quantitatively evaluating the comprehensive performance of each propagation event on different platforms based on the platform performance index model, and constructing a cross-platform content ranking list based on the comprehensive performance through a multi-attribute normalized decision function; and intuitively presenting the cross-platform network public opinion analysis results based on a three-level responsive visualization view.

[0008] On the other hand, a cross-platform public opinion dynamic assessment system based on real-time semantic awareness is provided, including: a data semantic fusion module, an event semantic update module, a popular content ranking module, and a visualization module. The data semantic fusion module collects multi-modal heterogeneous data from multiple platforms, standardizes the data, and semantically aggregates the standardized cross-platform multimodal data corresponding to each propagation event based on an event semantic center vector model, establishing a unique mathematical representation for each propagation event. The event semantic update module iteratively updates the event semantic center vector through online exponential sliding updates, continuously supplementing the event semantic center vector with newly added standardized data related to the propagation event in real time, ensuring that the mathematical representation of the event is synchronized with the latest data. The popular content ranking module quantitatively evaluates the comprehensive performance of each propagation event on different platforms based on a platform performance index model, and constructs a cross-platform content ranking list based on the comprehensive performance through a multi-attribute normalized decision function. The visualization module intuitively presents the cross-platform online public opinion analysis results based on a three-level responsive visualization view.

[0009] Beneficial effects The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following: 1. By collecting heterogeneous data from multiple platforms and modalities and performing cleaning, deduplication, and standardization, and by using the event semantic center vector model to achieve semantic aggregation of cross-platform data, a unique mathematical representation of each propagation event is constructed. This breaks down cross-platform data silos, solves the defects of traditional data aggregation rather than fusion, and achieves a leap from physical stacking of multi-source data to deep semantic fusion, providing a unified and high-fidelity data view for upper-level quantitative analysis.

[0010] 2. By using an online exponential sliding update mechanism and combining iterative updates of the event semantic center vector with a time decay factor, new standardized data is added in real time, thereby ensuring that the mathematical representation of the event is synchronized with the latest data, adapting to the dynamic evolution of the event, and thus achieving a minute-level response capability to sudden public opinion events, accurately capturing the real-time evolution trajectory of the event across the entire network.

[0011] 3. By integrating multi-dimensional indicators such as voice contribution, discussion depth, emotional bias, and cross-platform leadership through the platform effectiveness index model, the comprehensive effectiveness of an event on different platforms is quantitatively evaluated. Then, by combining multi-attribute normalized decision functions, a cross-platform content ranking is constructed, thereby providing the ability to make synchronous quantitative comparisons of events and platforms from two dimensions. This breaks the limitations of the single-point perspective of traditional analysis and achieves an upgrade from isolated platform analysis to global network evolution insight, deepening the understanding of complex propagation dynamics.

[0012] 4. By constructing a three-level responsive visualization view of overview, comparison, and details, and adopting a linked layout of left list and right dashboard, a guided interactive link is created, thereby creating an immersive exploration and analysis environment. This allows users to seamlessly drill down from the macro situation to the micro content entity, eliminating the cost of switching between multiple interfaces and cross-verifying information. In this way, a complete closed loop of perception, analysis, and verification is achieved within a single interface, greatly improving the smoothness and rigor of public opinion analysis.

[0013] 5. By integrating modules such as data semantic fusion, event semantic update, popular content sorting and visualization, a complete technical architecture of data fusion, quantitative model and interactive presentation is formed, outputting in-depth reports that combine phenomenon description and root cause diagnosis, thereby promoting the transformation of decision support from extensive intervention to precise targeting, avoiding resource waste, and realizing the evolution of the decision-making process from passive response to proactive prediction and optimal efficiency allocation, thus improving the accuracy and effectiveness of public opinion guidance. Attached Figure Description

[0014] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0015] Figure 1 A flowchart of a cross-platform public opinion dynamic assessment method based on real-time semantic awareness provided in an embodiment of this application; Figure 2 A technical roadmap for a cross-platform public opinion dynamic assessment method based on real-time semantic awareness, provided in the embodiments of this application; Figure 3 This is a schematic diagram of the structure of the cross-platform public opinion dynamic assessment method system based on real-time semantic awareness provided in the embodiments of this application. Detailed Implementation

[0016] Embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While some embodiments of the present disclosure are shown in the drawings, it should be understood that embodiments of the present disclosure may be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of the present disclosure.

[0017] It should be understood that the accompanying drawings and embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of protection of this disclosure. In the description of the embodiments of this disclosure, the terms "including" and similar expressions should be understood as open-ended, i.e., including but not limited to. The term "based on" should be understood as at least partially based on. The term "an embodiment" or "this embodiment" should be understood as at least one embodiment. The terms "first," "second," etc., may refer to different or the same objects.

[0018] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0019] like Figure 1 The diagram shows a flowchart of a cross-platform public opinion dynamic assessment method based on real-time semantic awareness provided in this application embodiment. The method includes the following steps: collecting multimodal heterogeneous data from publicly available data sources such as Weibo, Douyin, Bilibili, Zhihu, Xiaohongshu, Tieba, and Kuaishou, as well as news websites. This multimodal heterogeneous data includes text data, image data, audio data, video data, metadata, and network relationship data, all of which are strongly correlated with each dissemination event. The data is standardized. Based on the event semantic center vector model, the standardized cross-platform multimodal data corresponding to each dissemination event is semantically aggregated. A unique mathematical representation is established for each dissemination event, which is the event semantic center vector, condensing the core semantic information of the corresponding cross-platform related data. The standardization process includes cleaning invalid noise data, removing duplicate cross-platform data, and transforming... To unify the standard data model within the system and ensure compatibility of heterogeneous data in associating with the same propagation event, the system iteratively updates the semantic center vector of the event through online exponential sliding updates, continuously supplementing the event semantic center vector with newly added standardized data related to the propagation event in real time, thus synchronizing the mathematical representation of the event with the latest data. Based on the platform effectiveness index model, the system quantitatively evaluates the comprehensive effectiveness performance of each propagation event across different platforms, and constructs a cross-platform content ranking list based on the comprehensive effectiveness performance using a multi-attribute normalized decision function. Finally, a three-level responsive visualization view intuitively presents the cross-platform online public opinion analysis results. This three-level responsive visualization view is driven by an intelligent engine consisting of a unified event representation model, a platform effectiveness index model, and a fused content ranking model, enabling the automatic extraction of measurable and interpretable propagation effectiveness knowledge from multi-source heterogeneous data.

[0020] like Figure 2 The diagram shows the technical roadmap for a cross-platform public opinion dynamic assessment method based on real-time semantic awareness provided in this application. This technical roadmap achieves dynamic perception and accurate assessment of cross-platform public opinion through a closed-loop process of multi-source heterogeneous data collection, event semantic fusion, cross-platform performance evaluation, and visual interactive analysis. The steps are described in detail below: 1. Multi-source heterogeneous data acquisition and standardization Multi-source heterogeneous data acquisition and standardization encompasses two aspects: multi-platform data source aggregation and data cleaning and standardization. It provides basic data support for the entire process. Multi-platform data source aggregation targets various data sources such as social media, news media, audiovisual media, and professional media, integrating heterogeneous raw data across platforms to form multimodal data including text, images, and videos. Data cleaning and standardization perform deduplication, noise reduction, and format unification on the multimodal data to eliminate the heterogeneity of cross-platform data and output a standardized dataset with consistent structure and controllable quality, providing reliable input for subsequent semantic computing.

[0021] 2. Event Semantic Fusion Event semantic fusion comprises three aspects: platform-specific semantic encoding, weighted aggregation of semantic vectors, and dynamic semantic updates. This enables unified semantic representation and dynamic updates of cross-platform public opinion events. Platform-specific semantic encoding splits standardized multimodal data by source platform, with each platform's content set semantically mapped using a dedicated encoder to generate corresponding platform semantic vectors. Weighted aggregation of semantic vectors from each platform enters a unified semantic vector space, combining content weights to form a central event semantic vector that represents the global semantics of the event. Dynamic semantic updates utilize an online sliding update mechanism and a time-sensitive update strategy to iterate the event semantic central vector in real time, generating updated semantic central vectors and ultimately outputting a dynamic semantic representation to support real-time perception of public opinion. Simultaneously, it supports feedback from external strategy adjustments, enabling dynamic optimization of the semantic computation logic.

[0022] 3. Cross-platform performance evaluation Cross-platform performance evaluation comprises three aspects: performance index calculation, key indicator extraction and normalization, and closed-loop optimization feedback. It quantifies the dissemination effectiveness of public opinion events across various platforms. The performance index calculation, based on dynamic semantic representation, uses a platform performance index model to calculate the comprehensive dissemination effectiveness of each platform, reflecting the event's influence on the corresponding platform. Key indicator extraction and normalization extracts key dimensions from the comprehensive effectiveness, eliminates differences in indicator dimensions through multi-attribute normalization, and then ranks the indicators to obtain a content ranking list. In the closed-loop optimization feedback, the evaluation results of the content ranking list are used to output optimization suggestions and are also fed back to the event semantic fusion module, driving the dynamic adjustment of semantic computing strategies and forming a closed-loop iteration of performance evaluation and semantic perception.

[0023] 4. Three-level responsive visualization view The three-tiered responsive visualization view enables multi-dimensional interactive analysis of public opinion trends through layered visualization. The first-level view provides an overall overview of the entire network, offering a comprehensive search and overview based on the overall network dissemination effectiveness. It presents the overall dissemination scale, trends, and cross-platform distribution characteristics of public opinion events from a macro perspective, supporting global search and filtering. The second-level view handles cross-platform comparisons of events, drilling down from the first-level view to show a comparison of the dissemination effectiveness of the same event on different platforms such as Weibo, Douyin, and Bilibili, intuitively presenting the differences in the dissemination of events on each platform. The third-level view handles in-depth comparisons of each platform, drilling down further from the second-level view to focus on hotspot insights on individual platforms, including micro-characteristics such as dissemination nodes, content evolution, and user interaction within that platform. This achieves three-level linkage analysis from a global, cross-platform, and single-platform perspective, supporting accurate public opinion judgment and decision-making.

[0024] In this embodiment, a dissemination event refers to a core public opinion issue with a unified semantic theme across platforms that triggers widespread discussion. Data serves as the concrete carrier of the dissemination event, used to characterize its content, trajectory, and impact. This invention achieves significant technological advancements in the quantitative evaluation and in-depth insight of cross-platform social media dissemination by constructing a complete engineering analysis architecture. Its technical effectiveness is specifically reflected in four core dimensions: data processing efficiency, analysis dimensions, interactive operation, and decision support, systematically improving the efficiency of the entire chain from data to decision. Addressing the shortcomings of existing technologies, such as data aggregation rather than fusion and slow response, this system achieves fundamental improvements. First, through a unified automated data pipeline constructed using multi-source data aggregation and fusion processing modules, it realizes minute-level and even second-level processing of multimodal heterogeneous data from collection, cleaning, to standardization, enabling minute-level response capabilities to sudden public opinion events. More importantly, through the event semantic center vector model in this module, the system achieves a fundamental leap from the physical stacking of data across multiple platforms to the semantic fusion of information centered on events, generating a unified, high-fidelity data view to support upper-level quantitative analysis and completely solving the problem of cross-platform data silos. Addressing the shortcomings of existing technologies in terms of singular analytical perspectives and lack of horizontal comparison, this system achieves a paradigm shift through a cross-platform effectiveness quantitative analysis module. The core platform effectiveness index model and multi-platform normalized fusion ranking model of this module provide, for the first time, synchronous and quantitative comparison capabilities across both events and platforms. Analysts can not only intuitively compare the differences in volume and sentiment across platforms, but also... These indicators accurately reveal the cross-platform public opinion migration paths and resonance patterns, upgrading the analytical perspective from what happened within isolated platforms to how events evolve across the entire network, greatly deepening the understanding of complex propagation dynamics. To overcome the shortcomings of traditional systems that primarily rely on static reports and lack exploratory interaction, this system constructs a guided three-level visual interactive link. The coherent design of overview, comparison, and details, combined with the linked layout of left-side lists and right-side dashboards, creates an immersive exploratory analysis environment. Users can seamlessly drill down from the macro-level situation to micro-level content entities, completing a complete perception, analysis, and verification loop within a single interface, from problem discovery and attribution to locating the source of information. This design fundamentally eliminates the costs of switching between multiple interfaces and cross-verifying information, elevating the smoothness and rigor of the analysis work to a new level. Based on the aforementioned multi-dimensional and traceable quantitative analysis capabilities, this system achieves a precise leap in decision support models from describing phenomena to driving action. The system outputs not only public opinion briefings but also in-depth diagnostic reports: it not only reveals what happened but also clearly points out which platform the problem occurred on and what its root cause is through quantitative indicators. This allows strategy formulation to shift from extensive intervention relying on experience to precise targeting based on panoramic data insights, for example, targeting high... The platform provides emotional support, targeting low-income individuals. The platform focuses on content creation. Ultimately, the system drives the decision-making process to evolve from a passive response and resource-sharing model to a proactive prediction and optimal allocation model. In summary, this invention is not a simple aggregation of existing functions, but rather achieves a qualitative breakthrough in four key indicators—data fusion degree, analytical systematicity, interactive fluency, and decision-making accuracy—through a complete and innovative data fusion and quantification model and interactive presentation technology architecture. It provides an engineering-ready systematic solution for the quantitative evaluation and in-depth insight of cross-platform social media communication.

[0025] Furthermore, the calculation method for the event semantic center vector model is as follows: ; In the formula, Let p be the semantic center vector of event E, where p is the platform ID, such as {Weibo, Douyin, Zhihu, etc.}, p = 1, 2, 3, ..., P, where P is the total number of platforms, and i is the content ID, i = 1, 2, 3, ... , This is the collection of relevant content about event E on platform p; It is a multimodal semantic encoding function that converts text, images, etc., into high-dimensional vectors; Content weight is determined by a combination of interaction data such as likes and shares, and the decay effect over time.

[0026] In this embodiment, the event semantic center vector model accurately constructs a unique mathematical representation of a propagation event through a specific calculation formula. It clarifies the data source and attribution using platform IDs and content IDs, converts multimodal data such as text and images into high-dimensional vectors using multimodal semantic encoding functions, and combines content weights determined by interaction data such as likes and shares with publication time decay to perform weighted aggregation calculations on relevant content of the event across various platforms. The final result is an event semantic center vector that condenses the core semantic information of cross-platform related data. This model offers significant and comprehensive technical benefits. First, it achieves a breakthrough innovation in data processing, completely changing the traditional physical stacking of multi-platform data. Through deep semantic integration, it breaks down cross-platform data silos, constructing a unified, high-fidelity data view. This enables previously heterogeneous multimodal data to be compatible with the same propagation event, laying a solid data foundation for subsequent cross-platform quantitative analysis. Second, it achieves a qualitative improvement in analytical efficiency. The event semantic center vector generated by this model can systematically integrate propagation data of different platforms and types, freeing analysts from isolated platform-specific perspectives. Instead of limiting the scope of the event, the system focuses on understanding the overall online dissemination trend from the core of the event. Combined with an online index sliding update mechanism, it enables real-time synchronization of the event's mathematical representation with the latest data, ensuring accurate capture of the dynamic evolution of the dissemination event. This provides core support for subsequent platform performance index evaluation, cross-platform content ranking, and three-level responsive visualization, thereby promoting a full-link efficiency improvement in online public opinion guidance and analysis, from data integration, analytical dimensions, interactive operations to decision support. It achieves a fundamental shift from traditional technical shortcomings such as aggregation rather than fusion, slow response, and single-point perspectives to deep data fusion, minute-level response, and a global analytical perspective.

[0027] Furthermore, the calculation method for the online index sliding update is as follows: ; In the formula, Let be the semantic center vector of event E at time t. Let E be the semantic center vector at time t-1. The time decay factor, This represents the set of relevant content added at time t. This model merges fragmented content into a single, comparable, and dynamically updatable event semantic center vector, ensuring that all subsequent cross-platform analyses are based on the same semantic benchmark.

[0028] In this embodiment, the calculation method of online exponential sliding update is based on the dynamic iteration of the event semantic center vector. Its formula assigns weights to the semantic center vector of event E at time t-1 through a time decay factor, while also combining... The weights of the newly added relevant content at time t are incorporated into the weighted aggregation result. The aggregation calculation of the newly added content is also based on content weights and multimodal semantic encoding functions. By transforming and weighting the newly added multimodal data from each platform, and then dividing by the total weights, the semantic contribution value of the newly added data is obtained, ultimately updating the semantic center vector at time t. This method has extremely significant technical effects and multidimensional value: On the one hand, it perfectly adapts to the dynamic evolution of dissemination events, breaking the limitations of traditional static data processing in not being able to absorb new information in a timely manner. It can continuously supplement the event semantic center vector with standardized data added across platforms in real time, ensuring that the mathematical representation of the event is highly synchronized with the latest data, allowing public opinion analysis to always be based on a complete and fresh data foundation, providing key technical support for minute-level response to sudden public opinion events; on the other hand, it reasonably balances the weights of historical data and newly added data through a time decay factor, preserving the continuity of the core semantics of the event and avoiding the impact of fragmented new content. This approach addresses semantic shifts while highlighting the impact of the latest information, enabling smooth iteration and precise optimization of the semantic center vector. It effectively integrates fragmented, dynamically generated cross-platform content into a single, comparable, and dynamically updatable unified semantic representation. This ensures that all subsequent cross-platform analyses are based on a consistent semantic benchmark, providing stable and reliable data input for upper-level analysis modules such as platform performance index evaluation and cross-platform content ranking. It further promotes the improvement of the entire chain of efficiency from data fusion to quantitative analysis to decision support, helping analysts accurately capture the evolution trajectory of events across the entire network and providing a scientific basis for the dynamic adjustment of public opinion guidance strategies.

[0029] Furthermore, the calculation method for the platform efficiency index model is as follows: ; In the formula, To assess the overall performance of event E on the p platform, .

[0030] In this embodiment, the calculation method of the platform effectiveness index model takes the comprehensive effectiveness performance of event E on platform p as the core objective. It achieves quantitative evaluation through the collaborative operation of multi-dimensional indicators. Its formula integrates four key elements: the proportion of platform voice contribution, the relative value of discussion depth, the sentiment bias correction term, and the relative value of cross-platform leadership. Among them, the square root of the product of voice contribution and discussion depth constitutes the basic effectiveness dimension. The sentiment bias correction term adjusts the deviation between the platform's average sentiment value and the average sentiment value of the entire network through the sentiment bias sensitivity coefficient. The cross-platform leadership is calculated based on the cross-platform explicit citation chain. The platform's leadership role is quantified by statistically analyzing the citation of content by other platforms through an indicator function and combining the total number of citations. Finally, a quantitative index that comprehensively reflects the platform's overall performance in the dissemination of the event is formed. The model's technical effects are groundbreaking and systematic: First, it completely revolutionizes the limitations of traditional, singular perspectives in public opinion analysis. For the first time, it constructs a simultaneous quantitative comparison system across two dimensions: events and platforms. This breaks the previous limitation of only being able to observe public opinion dynamics within a single platform in isolation. Analysts can intuitively grasp the differences between platforms in terms of volume contribution, discussion depth, sentiment, and cross-platform leadership, accurately revealing the cross-platform public opinion migration path and resonance patterns. It upgrades the analytical perspective from what happened within a platform to how events evolve across the entire network, greatly deepening the understanding of complex communication dynamics. Second, it achieves standardization and refinement of platform effectiveness evaluation. Through the scientific integration and normalization of multi-dimensional indicators, it eliminates evaluation biases caused by differences in data formats and interaction rules across different platforms. This ensures that cross-platform effectiveness comparisons have a unified benchmark and comparability. Furthermore, the detailed design of each indicator accurately captures the core value of a platform in public opinion dissemination. The volume contribution reflects the breadth of the platform's dissemination, the depth of discussion reflects the penetration of the content's influence, the sentiment bias correction item highlights the guiding role of the platform's public opinion atmosphere, and the cross-platform leadership demonstrates the platform's core position in the overall online dissemination, providing a clear target basis for subsequent precise policy implementation. Thirdly, it provides a solid quantitative foundation for upper-level analysis and decision support. Its output platform efficiency index is not only the core basis for constructing cross-platform content rankings, but also the core data support for the multi-dimensional analysis dashboard in the three-level responsive visualization view. Through synergy with other modules, it enables public opinion analysis to leap from describing phenomena to driving action, helping strategists to conduct emotional guidance for platforms with high sentiment bias and large sentiment deviation, and to cultivate content for platforms with low discussion depth and insufficient discussion depth. It promotes the evolution of the decision-making process from a passive response and resource allocation model to a proactive prediction and optimal efficiency allocation model, systematically improving the accuracy and effectiveness of public opinion guidance.

[0031] Furthermore, the cross-platform leadership coefficient quantifies the leading role of platform p in the propagation of events on other platforms, calculated through cross-platform explicit reference chains: ; In the formula, for , This refers to the i-th piece of content on platform p that belongs to event E; It is an indicator function that has a value of 1 when content i is explicitly referenced by content q on platform q, and a value of 0 otherwise; It represents the total number of times content i has been cited.

[0032] In this embodiment, the method for calculating the cross-platform leadership coefficient takes the set of relevant content belonging to event E on platform p as the core analysis object. It uses an indicator function to count the number of times each piece of content i is explicitly referenced by other platforms (the value is 1 if it is referenced, and 0 otherwise). Then, combined with the total number of references received by content i, the method uses double summation and ratio calculation to accurately quantify the leading role of platform p in the propagation of the event on other platforms. The technical effects of this calculation method are groundbreaking and of great practical value: On the one hand, it achieves, for the first time, a quantifiable and traceable assessment of cross-platform leadership, breaking the limitations of traditional public opinion analysis in accurately defining the transmission path of platform dissemination influence. By capturing the core correlation clue of explicit cross-platform citation chains, it transforms the abstract platform leadership into a concrete and calculable quantitative indicator, enabling analysts to clearly identify the platforms that play a core leading role in the dissemination of events, as well as the migration path of public opinion across different platforms. On the other hand, its calculation logic is closely integrated with the actual citation situation of the content. It considers both the breadth of a single piece of content being cited by multiple platforms and normalizes the results through the total number of citations, avoiding the bias caused by some highly cited content excessively dominating the results. This ensures the objectivity and accuracy of the coefficient. As a key component of the platform effectiveness index model, this coefficient provides core support for the dual-dimensional quantitative comparison of events and platforms, helping analysts to understand the evolution of events across the entire network from a global perspective, deeply understand the complex dissemination dynamics, and thus provide precise targeting basis for the formulation of public opinion guidance strategies. For example, for platforms with high leadership coefficients, the focus can be on strengthening the output of positive information to leverage their radiating and driving effect. For platforms with low leadership coefficients but high potential, their dissemination influence can be enhanced through content optimization, thus promoting the evolution of public opinion guidance from passive response to proactive prediction and optimal allocation of effectiveness.

[0033] Furthermore, It uses a unified sentiment score calculation, adapted to the content characteristics of different platforms, specifically: ; In the formula, This represents a quantitative value indicating the overall sentiment tendency implied by content i. A value closer to 1 indicates a more positive sentiment, while a value closer to 0 indicates a more negative sentiment. It is calculated using a platform-adaptive sentiment analysis model. ; It is a sentiment analysis model optimized for the content characteristics of platform P, such as short videos. It is a multimodal model that comprehensively analyzes video titles, comments, and audio / subtitle information.

[0034] In this embodiment, the average sentiment value is calculated based on the set of relevant content for event E on platform p. The sentiment of each piece of content i is quantified using a platform-adaptive sentiment analysis model, resulting in a sentiment tendency value ranging from [0,1] (closer to 1 is more positive, closer to 0 is more negative). The average sentiment value of platform p is then calculated by combining the content weights determined by interaction data and publication time decay, and the weighted sum is compared with the ratio of the total weights. Furthermore, the sentiment analysis model is adapted to a multimodal model to account for the content characteristics of different platforms such as short videos, integrating information such as video titles, comments, and audio / subtitles for sentiment calculation. This calculation method boasts a high degree of accuracy, adaptability, and support. Firstly, it achieves standardization and objectivity in cross-platform sentiment assessment, breaking down barriers to sentiment analysis caused by differences in content formats across platforms. Through a unified sentiment scoring logic and platform-adaptive analytical models, it ensures the comparability of sentiment data across platforms, avoiding biases caused by neglecting platform characteristics in traditional sentiment analysis. This provides a reliable basis for subsequent calculations of the overall average sentiment score and comparisons of sentiment differences between platforms. Secondly, its content-weighted calculation logic highlights the sentiment weight of highly interactive and recently released content with greater dissemination influence, while also taking into account the overall... The sentiment distribution of content allows the average sentiment value to accurately reflect the mainstream sentiment trend of events within the platform, precisely capturing the sentiment trajectory of public opinion. Simultaneously, this average sentiment value serves as a core input indicator for the platform's effectiveness index model. By calculating the deviation from the average sentiment value across the entire network, it becomes a key basis for measuring the platform's sentiment deviation. This helps analysts clearly identify platforms with extreme sentiment or those deviating from the mainstream sentiment across the network, providing precise quantitative support for targeted sentiment guidance strategies such as emotional management and content optimization. It promotes a shift in decision-making from experience-driven, extensive intervention to data-driven, precise targeting, enhancing the depth of cross-platform public opinion analysis and the scientific rigor of decision-making.

[0035] Furthermore, the sorting method for cross-platform content ranking lists is as follows: ; In the formula, This indicates how closely content i resembles the ideal optimal content. = It is the attribute vector of content i on platform p. To normalize the popularity, it is calculated using weighted averages of native platform interaction metrics, such as likes, shares, and virtual coins. (This represents the intensity of emotion; the closer it is to 1, the more extreme the emotion and the greater its impact on public opinion.) The propagation speed, i.e., the heat gained per unit time, reflects the diffusion efficiency. The platform's weight is dynamically adjusted based on its historical performance in specific events. It is the ideal solution, consisting of the maximum values ​​of each attribute column, i.e., (max(NormHeat), max(EmoIntensity), max(SpreadSpeed), max( )), It is a negative ideal solution, consisting of the minimum values ​​of each attribute column, i.e., (min(NormHeat), min(EmoIntensity), min(SpreadSpeed), min( )), Let W represent the weighted Euclidean distance, where W is the diagonal weight matrix diag. This is used to adjust the importance of each attribute in the overall ranking, satisfying ∑β=1.

[0036] The specific method for obtaining the attribute vector is as follows: ; In the formula, k is the interaction indicator number, k=1,2,3,... , This refers to the set of interactive metrics covered by platform p (e.g., number of reposts, comments, likes, shares, favorites, bullet comments, coins, etc.). This represents the raw value of content i on the interaction metric k. For the weighting coefficients corresponding to platform p and interaction metric k, satisfying These weights reflect the value differences of different interactive behaviors within the platform's ecosystem; ; ; In the formula, To increase the popularity of content on the platform. The number of hours since content i was published on platform p; ; In the formula, Number historical events =1,2,3,... , A collection of historical events.

[0037] In this embodiment, the ranking method of the cross-platform content ranking is based on a multi-attribute normalized decision function. It constructs an attribute vector of content i on platform p (including normalized popularity calculated by weighting the platform's native interaction indicators, emotional intensity derived from the quantitative value of sentiment tendency, dissemination speed reflecting diffusion efficiency, and platform weight dynamically adjusted based on the platform's historical performance). It also sets a positive ideal solution composed of the maximum value of each attribute column and a negative ideal solution composed of the minimum value. The closeness between the content and the positive and negative ideal solutions is calculated by using weighted Euclidean distance, and the ranking of the content is finally determined by the closeness. This ranking method boasts comprehensiveness, accuracy, and practicality. Firstly, it achieves a systematic integration of cross-platform content evaluation dimensions, breaking away from the limitations of traditional ranking methods that rely solely on a single popularity indicator. By considering four core attributes—popularity, emotional intensity, dissemination speed, and platform weight—it comprehensively captures the content's overall performance in terms of dissemination breadth, influence depth, diffusion efficiency, and platform value, avoiding the bias caused by single-dimensional evaluation and ensuring the ranking results better reflect the content's actual impact on public opinion. Secondly, it achieves comparability and normalization of cross-platform content. Through adaptive processing and standardized calculation of interaction indicators and content characteristics across different platforms, it eliminates evaluation barriers caused by differences in data formats and ecosystem rules between platforms. This allows content from different platforms such as Weibo, Douyin, and Zhihu to be fairly compared under the same evaluation system, providing analysts with a globally unified benchmark for judging content value. Simultaneously, its flexible weight adjustment mechanism and dynamically updated attribute data can adapt to the characteristics of different dissemination events and the needs of public opinion analysis, ensuring the relevance and timeliness of the ranking results. This ranking list serves as a three-level responsive visualization... Figure 2 The core component of the interface, together with the multi-dimensional analysis dashboard on the right, forms a linked interactive system with the left-side list and right-side dashboard. This allows users to quickly drill down from popular content to detailed data and key information, greatly improving the smoothness and convenience of public opinion analysis. It also provides direct support for accurately locating the core content of public opinion, tracing the source of dissemination, and formulating targeted guidance strategies. This promotes the upgrading of public opinion analysis from scattered content browsing to systematic core information mining, helping decision-makers to efficiently grasp key nodes in public opinion and improve the accuracy and timeliness of public opinion guidance.

[0038] Furthermore, the specific acquisition method of attribute vectors revolves around the core dissemination characteristics of content i on platform p. A comprehensive attribute representation is constructed through multi-dimensional and precise calculation: Normalized popularity is obtained by weighting and summing the set of interactive indicators such as the number of reposts and comments covered by platform p, combined with the weight coefficients of each indicator within the platform ecosystem. This not only reflects the value differences of different interactive behaviors but also achieves the standardization of popularity data; Emotional intensity is based on the quantitative value of the emotional tendency of content i, intuitively reflecting the degree of emotional extremes and its potential impact on public opinion; Dissemination speed is based on the popularity of content i on platform p and the number of hours since its publication. The ratio of popularity to time increment is used to accurately capture the dissemination efficiency of the content; Platform weight is calculated based on the comprehensive performance of platform p in various historical events in the historical event set ℋ, combined with the maximum historical performance of the entire platform, and dynamically adapts the value weight of the platform in specific events. The technical effects of this acquisition method are comprehensive and far-reaching. First, it achieves standardization and precise quantification of cross-platform content attributes. Addressing the differences in interaction rules and content formats across different platforms, it eliminates evaluation biases caused by data heterogeneity through adaptive calculation logic. This makes core attributes such as normalized popularity and dissemination speed comparable across different platforms, laying a solid foundation for fair ranking of cross-platform content leaderboards. Second, its multi-dimensional attribute construction logic comprehensively characterizes the dissemination characteristics of content from four core dimensions: popularity, sentiment, dissemination efficiency, and platform value. This avoids the one-sidedness of single-attribute evaluation, enabling attribute vectors to... It accurately and comprehensively reflects the actual impact of content on public opinion; at the same time, the calculation of each attribute is closely integrated with the platform's ecological characteristics and historical data, highlighting the dynamic value of real-time dissemination data, and providing objective basis for platform weights through historical performance analysis, ensuring the timeliness and reliability of attribute vectors. This provides high-quality, high-fidelity input data for the subsequent calculation of multi-attribute normalized decision functions, thereby supporting the accurate ranking of cross-platform content rankings, helping analysts quickly locate core public opinion content, providing precise targeted support for the formulation of public opinion guidance strategies, and improving the depth of cross-platform public opinion analysis and the scientific nature of decision-making.

[0039] Furthermore, the three-tiered responsive visualization view includes a first-level view, a second-level view, and a third-level view. The first-level view is a large screen providing an overall overview of the network situation, presenting online index updates. Based on real-time data aggregated and fused from multimodal data, it generates core summary indicators through a basic statistical model, revealing the focus and anomalies of the macro-level communication situation. It is equipped with a global search box and a time selector, allowing users to switch to the second-level view after entering a topic or selecting a time range. The second-level view is a sub-screen for cross-platform event comparison and analysis, driven by a quantitative model from the cross-platform effectiveness quantitative analysis module, presenting the output results of the platform effectiveness index model. The cross-platform content ranking uses a left-list, right-dashboard layout. The left side is a cross-platform popular content ranking dynamically generated by a multi-platform normalized fusion ranking model (with content source and interaction data labeled), while the right side is a multi-dimensional analysis dashboard corresponding to the output results of the platform efficiency index model. Users can click on the items on the left to view detailed data and key information excerpts on the right, forming an interactive loop. The third-level view is a single-platform in-depth analysis interface, accessed through the platform entry in the second-level view. It reuses the PEI model framework to focus on displaying detailed trend information for a single platform and can be used as a daily monitoring dashboard to switch analysis perspectives.

[0040] In this embodiment, the three-level responsive visualization view constructs a complete and smooth interactive link for public opinion analysis through a hierarchical and interconnected design: a first-level overall network situation overview screen, a second-level cross-platform event comparison and analysis sub-screen, and a third-level single-platform in-depth analysis interface. Its technical effects are comprehensive and groundbreaking: First, it achieves seamless connection and in-depth extension of the public opinion analysis perspective. The first-level view, based on real-time data after multimodal data aggregation and fusion and online index sliding update results, generates core summary indicators through a basic statistical model, accurately revealing the focus and anomalies of the macro-level dissemination trend, providing users with a global overview of public opinion. Combined with a global search box and time selector, it supports quick location of target topics and time ranges and switching to the second-level view, solving the problem of unintuitive macro-level situation perception in traditional analysis; the second-level view quantifies cross-platform effectiveness. Driven by the analysis module, the platform performance index model output results and cross-platform popular content integrated rankings are presented in a linked layout of left-hand list and right-hand dashboard. The left side clearly marks the content source and interaction data, while the right side displays a multi-dimensional analysis dashboard. Users can click on the left-hand items to obtain detailed data and key information extracts on the right side, forming a complete interactive loop. This breaks the limitations of previous cross-platform data comparisons that were scattered and information fragmented, allowing analysts to intuitively grasp the differences in comprehensive performance and the distribution of popular content across platforms. The third-level view is accessed through the platform entry point of the second-level view. It reuses the PEI model framework to focus on displaying detailed trend information for a single platform. It can be used as a daily monitoring dashboard to flexibly switch analysis perspectives, filling the gap in traditional analysis that lacks in-depth insights into a single platform. It achieves comprehensive coverage from macro to micro, and from cross-platform comparisons to in-depth analysis of a single platform. Meanwhile, this view system is driven by an intelligent engine consisting of a unified event representation model, a platform performance index model, and a fusion content ranking model. It can automatically extract measurable and interpretable knowledge of dissemination effectiveness from multi-source heterogeneous data, completely changing the status quo of traditional systems that are mainly static reports and lack exploratory interaction. It creates an immersive exploratory analysis environment, allowing users to complete the entire closed loop from problem discovery, comparison and attribution to source location within a single interface. It fundamentally eliminates the cost of switching between multiple interfaces and cross-validating information, greatly improving the smoothness, rigor, and efficiency of analysis work. It provides an intuitive and efficient presentation carrier for the leap of decision support models from describing phenomena to driving action, helping users accurately grasp public opinion dynamics and scientifically formulate guidance strategies.

[0041] like Figure 3The diagram shown is a structural schematic of the cross-platform public opinion dynamic assessment method system based on real-time semantic awareness provided in this application embodiment. It includes: a data semantic fusion module, an event semantic update module, a popular content ranking module, and a visualization module. The data semantic fusion module collects multi-platform, multi-modal, heterogeneous data, standardizes the data, and semantically aggregates the standardized cross-platform multi-modal data corresponding to each propagation event based on the event semantic center vector model. A unique mathematical representation is established for each propagation event, which is the event semantic center vector, condensing the core semantic information of the corresponding cross-platform related data. The event semantic update module updates the event semantics online through exponential sliding updates, taking into account the dynamic evolution of propagation events. The central vector is iteratively updated, continuously supplementing the event semantic central vector with newly added standardized data related to the dissemination event in real time, so that the mathematical representation of the event is synchronized with the latest data; the popular content ranking module quantifies and evaluates the comprehensive performance of each dissemination event on different platforms based on the platform effectiveness index model, and constructs a cross-platform content ranking list based on the comprehensive performance through a multi-attribute normalized decision function; the visualization module intuitively presents the cross-platform network public opinion analysis results based on a three-level responsive visualization view. The three-level responsive visualization view is driven by an intelligent engine consisting of an event unified representation model, a platform effectiveness index model, and a fused content ranking model, realizing the automatic extraction of measurable and interpretable dissemination effectiveness knowledge from multi-source heterogeneous data.

[0042] Through the above description of the implementation methods, those skilled in the art can clearly understand that, for the sake of convenience and brevity, only the division of the above functional modules is used as an example. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the above functions can be divided into different functional modules to complete all or part of the functions described above.

[0043] In the embodiments provided in this application, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the embodiments described above are merely illustrative; for instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces, or indirect coupling or communication connection between devices or units, and may be electrical, mechanical, or other forms.

[0044] The units described as separate components may or may not be physically separate. A component shown as a unit can be one or more physical units, located in one place or distributed in multiple different locations. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0045] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0046] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a readable storage medium. Based on this understanding, the technical solutions of the embodiments of this application, or the parts that contribute to the solution, or all or part of the technical solution, can be embodied in the form of a software product. This software product is stored in a storage medium and includes several instructions to cause a device (which may be a microcontroller, chip, etc.) or processor to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0047] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A real-time semantic perception based cross-platform public opinion dynamic evaluation method, characterized in that, The method comprises the following steps: Collecting multi-platform and multi-modal heterogeneous data, performing standardization processing on the data, performing semantic aggregation on cross-platform multi-modal data corresponding to each propagation event after standardization processing based on an event semantic center vector model, establishing a unique mathematical representation for each propagation event, and the mathematical representation is an event semantic center vector which condenses core semantic information of cross-platform associated data of the corresponding propagation event; For dynamic evolution of the propagation event, the event semantic center vector is iteratively updated through online exponential sliding update, and the newly added standardized data related to the propagation event is continuously supplemented into the event semantic center vector in real time, so that the mathematical representation of the event is synchronized with the latest data; Based on the platform performance index model, the comprehensive performance of each propagation event on different platforms is quantitatively evaluated, and based on the comprehensive performance, a cross-platform content ranking list is constructed through a multi-attribute normalization decision function; Based on a three-level responsive visualization view, the cross-platform network public opinion analysis result is intuitively presented, the three-level responsive visualization view is driven by an intelligent engine composed of an event unified representation model, a platform performance index model and a fusion content sorting model, and measurable and interpretable propagation performance knowledge is automatically extracted from multi-source heterogeneous data.

2. The method of claim 1, wherein the method further comprises: The calculation method of the event semantic center vector model is: ； wherein, is the semantic center vector of event E, p is the platform number, p = 1, 2, 3,..., P, P is the total number of platforms, i is the content number, i = 1, 2, 3,...,, , is the relevant content set of event E on platform p; is a multi-modal semantic encoding function; is the content weight.

3. The method of claim 1, wherein the method further comprises: The calculation method of the online exponential sliding update is: ； In the formula, is the semantic center vector of the event E at time t, is the semantic center vector of the event E at time t-1, is a time decay factor, is the set of related content newly added at time t.

4. The method of claim 1, wherein the method further comprises: The calculation method of the platform performance index model is: ； In the formula, is the comprehensive performance of event E on the p platform, 。 5. The method of claim 4, wherein the method further comprises: The cross-platform leading force system quantifies the leading role of platform p in event propagation to other platforms through cross-platform explicit citation chain calculation: ； where is , denotes the i-th piece of content on platform p belonging to event E; is an indicator function that takes the value 1 when content i is explicitly referenced by content on platform q, and 0 otherwise; is the total number of references received by content i.

6. The method of claim 4, wherein the method further comprises: The is calculated by using a unified emotional score calculation method and adapting to different platform content characteristics, specifically: ； In the formula, represents the quantitative value of the overall emotional tendency contained in the content i, which is calculated by a platform adaptive sentiment analysis model, and specifically is: ; is a sentiment analysis model optimized for the content characteristics of the platform p.

7. The method of claim 1, wherein the method further comprises: The sorting method of the cross-platform content ranking list is: ； wherein, represents the closeness of content i to the ideal optimal content, is the attribute vector of content i on platform p, is the normalized heat, is the emotional intensity, is the spread speed, is the platform weight, is the positive ideal solution, which is composed of the maximum value of each attribute column, is the negative ideal solution, which is composed of the minimum value of each attribute column, represents the weighted Euclidean distance, wherein W is a diagonal weight matrix.​ 8. The method of claim 7, wherein the method further comprises: The specific acquisition method of the attribute vector is: ； where k is the interaction index number, k = 1, 2, 3,..., , is the set of interaction indexes covered by platform p, represents the original value of content i on interaction index k, is the weight coefficient corresponding to platform p and interaction index k; ； ； wherein, is the heat of content i on platform p, is the number of hours since content i was published on platform p; ； In the formula, is a historical event number, = 1, 2, 3,..., , is a historical event set.

9. The method of claim 1, wherein the method further comprises: The three-level responsive visualization view comprises a first-level view, a second-level view and a third-level view, wherein the first-level view is a whole network situation overview large screen, presents the online exponential sliding update output result, and supports switching to the second-level view after user input topic or selection of time range; The second-level view is an event cross-platform comparative analysis sub-screen, which is driven by a cross-platform performance quantitative analysis module quantization model, presents platform performance index model output results and cross-platform content ranking list, and the third-level view is a single platform in-depth analysis interface, which is entered through a platform portal of the second-level view and focuses on a single platform to display detailed trend information.

10. A real-time semantic-aware cross-platform public opinion dynamic evaluation system, applied to the real-time semantic-aware cross-platform public opinion dynamic evaluation method of any one of claims 1 and 9, characterized in that, It comprises: a data semantic fusion module, an event semantic update module, a popular content sorting module and a visualization presentation module; The data semantic fusion module collects multi-platform and multi-modal heterogeneous data, performs standardization processing on the data, performs semantic aggregation on cross-platform multi-modal data corresponding to each propagation event after standardization processing based on an event semantic center vector model, establishes a unique mathematical representation for each propagation event, and the mathematical representation is an event semantic center vector which condenses core semantic information of cross-platform associated data of the corresponding propagation event; The event semantic update module iteratively updates the event semantic center vector through online exponential sliding update for dynamic evolution of the propagation event, and continuously supplements the newly added standardized data related to the propagation event into the event semantic center vector in real time, so that the mathematical representation of the event is synchronized with the newest data. The popular content ranking module quantifies and evaluates the comprehensive effectiveness of each dissemination event on different platforms based on the platform effectiveness index model, and constructs a cross-platform content ranking list based on the comprehensive effectiveness through a multi-attribute normalized decision function. The visualization module presents the cross-platform online public opinion analysis results intuitively based on a three-level responsive visualization view. The three-level responsive visualization view is driven by an intelligent engine consisting of an event unified representation model, a platform efficiency index model, and a fusion content ranking model, which enables the automatic extraction of measurable and interpretable dissemination efficiency knowledge from multi-source heterogeneous data.

Images