Multi-target dynamic composition intelligent guiding system and method for sudden news event

Abstract

The invention relates to the field of news event analysis and visualization, in particular to a multi-target dynamic composition intelligent guiding system and method for sudden news events, and the system comprises a news event monitoring analysis subsystem which is used for monitoring and analyzing existing news events, mining important news entities and obtaining event dynamic chains, and a target analysis subsystem which is used for receiving the information and sending the information to a target analysis subsystem; the method comprises the following steps of: performing clustering analysis on news keywords and entity objects, identifying related important news events and main bodies, receiving the events and main body information by a composition subsystem, performing composition modeling to obtain a news relation graph, receiving the news relation graph by a multi-target composition intelligent guide subsystem, constructing an event topology spectrum tensor and generating a dynamic characteristic matrix, and performing multi-target composition intelligent guide on the news relation graph. According to the method, a multi-target optimization guide vector field is generated, finally, the news relation graph is dynamically guided, comprehensive monitoring, mining and traceability of sudden news events are achieved by establishing a complete news event monitoring and analysis system, and the comprehensiveness and accuracy of event discovery are effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of news event analysis and visualization, and in particular to a multi-target dynamic composition intelligent guidance system and method for breaking news events, which are used to realize automatic monitoring, analysis, composition and intelligent guidance of breaking news events. Background Art

[0002] With the rapid development of the internet and social media, the amount of information surrounding breaking news events has exploded, and related news reports and social media discussions have formed a complex information network. Against this backdrop, how to quickly identify important news events from this massive amount of information, explore the connections between events, and construct intuitive visualizations and intelligently guide responses have become major challenges for news media, government agencies, and corporate decision-makers.

[0003] Existing technologies usually use keyword extraction, simple cluster analysis and static visualization methods to process news events. These methods have the following shortcomings: first, they lack the ability to capture the dynamic development process of events and find it difficult to reflect the complex correlations between events; second, the composition method is too simple to meet the needs of multi-objective optimization, resulting in poor visualization effects; finally, there is a lack of intelligent guidance mechanism, and the composition strategy cannot be dynamically adjusted according to the importance of the event, making it difficult to highlight key event nodes.

[0004] Several technologies have attempted to address these issues, such as semantic analysis-based event relationship extraction and graph-based event visualization. However, these technologies still suffer from incomplete event discovery, limited relationship analysis, insufficiently intelligent image composition, and inaccurate user guidance. These technologies are unable to meet the practical needs of dynamic, multi-target image composition for breaking news events. Summary of the Invention

[0005] The purpose of the present invention is to provide a multi-target dynamic composition intelligent guidance system and method for breaking news events, aiming to solve the problems existing in the prior art and realize comprehensive monitoring and analysis, intelligent composition and precise guidance of breaking news events.

[0006] The present invention proposes a multi-target dynamic composition intelligent guidance system for breaking news events, which includes:

[0007] News event monitoring and analysis subsystem, used for:

[0008] Dynamically monitor and analyze existing news events to identify important news entities;

[0009] Get dynamic links of important news events;

[0010] The target analysis subsystem is connected to the news event monitoring and analysis subsystem and is used to:

[0011] Receiving important news entities and news event dynamic links sent by the news event monitoring and analysis subsystem;

[0012] Based on the important news entities and news event dynamic chains, automatically cluster analysis is performed on news keywords and entity objects in mined news events;

[0013] Obtain relevant important news events and subjects;

[0014] The imaging subsystem is connected to the target analysis subsystem and is used to:

[0015] Receive relevant important news events and subjects sent by the target analysis subsystem;

[0016] Conducting composition modeling for a plurality of the relevant important news events and subjects;

[0017] Get the news relationship diagram;

[0018] The multi-target composition intelligent guidance subsystem is connected to the composition subsystem and is used to:

[0019] receiving the news relationship graph sent by the composition subsystem;

[0020] Construct event topology spectrum tensor;

[0021] generating a dynamic feature matrix based on the event topology spectrum tensor;

[0022] generating a multi-objective optimization guidance vector field based on the dynamic feature matrix;

[0023] The news relationship graph is dynamically guided according to the multi-objective optimization guidance vector field.

[0024] Preferably, the news event monitoring and analysis subsystem specifically includes:

[0025] News event discovery subsystem, used for periodic discovery and pre-processing of news events, and mining of news events;

[0026] A news event extraction subsystem, connected to the news event discovery subsystem, is used to extract elements contained in news events, including the time of occurrence of the news event, geographical information, theme, keywords, object, subject, event chain, and event subject evolution path;

[0027] The news event tracing and dynamic analysis subsystem is connected to the news event extraction subsystem and is used to mine events with mutual correlation and analyze the dynamic relationship between various elements and sub-events in the same event;

[0028] The news event fusion subsystem is connected to the news event tracing and dynamic analysis subsystem and is used to fuse multiple events in the same news event to obtain a dynamic chain of news events.

[0029] Preferably, the monitoring and analysis process of the news event monitoring and analysis subsystem specifically includes:

[0030] Use Internet crawler modules to monitor the content of mainstream news sources and news blogs in real time;

[0031] Filter news through the news event discovery module to obtain news event templates;

[0032] Fill the results obtained by the Internet crawler module into the event extraction module template, modify the template, mine news events, and obtain news titles, news time, and news content information;

[0033] The news events obtained by the mining are traced and analyzed to obtain the dynamic evolution path of the subject and the dynamic evolution of the theme keywords, attributes, and important nodes in the event over time, and a news event relationship model is established.

[0034] Preferably, the target analysis subsystem performs cluster analysis based on the following multiple dimensions:

[0035] Topic distribution dimension;

[0036] The dimensions of communication sources and communication processes;

[0037] Theme and communication process dimensions;

[0038] The dimension of themes and hot figures in communication;

[0039] Thematic and semantic network dimensions;

[0040] The target analysis subsystem integrates the cluster analysis results of the multiple dimensions to obtain important news events and subjects.

[0041] Preferably, the composition subsystem uses a composition module to compose multiple important events, obtain a news relationship graph G(v,e), including key nodes v and connecting edges e, and provide node and edge descriptions for the modeling graph.

[0042] Preferably, the event topology spectrum tensor construction process in the multi-objective composition intelligent guidance subsystem is calculated according to the following formula:

[0043]

[0044] in: is the (i, j, k)th element of the event topology spectrum tensor, i∈[1,n], j∈[1,m], k∈[1,l] are the node index, edge index and time index respectively, w ij ∈[0,1] is the node-edge association weight coefficient, R(v i ,e j ):V×E→[0,1] is the node v i With edge e j The relationship strength function, t c ∈R + is the current time point, t ref ∈R + is the reference time point, τ∈R + is the time attenuation coefficient, is a nonlinear heat kernel function, n∈N + is the number of event nodes, m∈N + is the number of relationship edges, l∈N + is the number of time slices, V is the node set, and E is the edge set.

[0045] Preferably, the dynamic feature matrix generation process in the multi-objective composition intelligent guidance subsystem is calculated according to the following formula:

[0046]

[0047] Where: F(t)∈R n×n is the dynamic feature matrix at time t, F(t+Δt)∈R n×n is the dynamic characteristic matrix at time t+Δt, α∈[0,1] is the time continuity coefficient, β∈[0,1] is the curl influence coefficient, γ∈[0,1] is the spectral decomposition weight coefficient, is a three-dimensional vector field derived from the tensor, is the curl of the vector field, λ k ∈R + is the kth principal eigenvalue, W k ∈R n×n is the kth feature weight matrix, is the matrix representation of the kth tensor slice at time t, ⊙ is the Hadamard product (element-to-element product), det(C k (t))∈R is the determinant of the kth covariance matrix at time t, K∈N + is the number of main modes considered, r = [x,y,z] T ∈R 3 is the space coordinate vector.

[0048] Preferably, the multi-objective optimization guidance vector field generation process in the multi-objective composition intelligent guidance subsystem is calculated according to the following formula:

[0049] V(x,y,z,t)=G potential (x,y,z)+G spectral (x,y,z,t)+G attention (x,y,z),

[0050]

[0051] Where V(x,y,z,t):R 3 ×R + →R 3 is the three-dimensional time-varying guidance vector field, G potential :R 3 →R 3 is the potential field guide component, G spectral :R 3 ×R + →R 3 is the spectral domain guided component, G attention :R 3 →R 3 is the attention field component, U(x,y,z):R 3 →R is the scalar potential function, is the gradient vector of the potential function, μ1, μ2, μ3∈[0,1] are the weight coefficients of the three components and μ1+μ2+μ3=1, is the Laplace inverse transform operator with respect to the time variable t, H F (s):C→R 3 is the vectorized Laplace transform of the dynamic feature matrix F(t), s∈C is the Laplace complex frequency variable, ω0∈R + is the system characteristic frequency, Ω(s):C→R is the frequency domain modulation function, and Re{·} is the complex real part operation a i ∈R 3 is the direction vector of the i-th attention center, r = [x, y, z] T ∈R 3 is the current spatial position vector, r i ∈R 3 is the i-th attention center position vector, σ i ∈R + is the Gaussian diffusion parameter of the i-th attention region, N∈N + The number of attention centers is determined by the number of important event nodes and usually does not exceed 5.

[0052] Preferably, the news event monitoring and analysis subsystem is composed of an Internet crawler module, a news event discovery module, a news event extraction module, and an event tracing and dynamic analysis module;

[0053] The Internet crawler module simulates a browser by proxy IP through a crawler framework and dynamically loads news sites;

[0054] The news event discovery module filters the news to obtain a news event template, fills the extracted information into the template, modifies the template, mines news events, and obtains news titles, news time, and news content information;

[0055] The news event extraction module extracts event elements from news events and reassembles the event elements into complete independent events;

[0056] The event tracing and dynamic analysis module performs tracing analysis on the acquired news events, obtains the dynamic evolution path of the subject and the dynamic evolution of the theme keywords, attributes, and important nodes in the event over time, and establishes a news event relationship model.

[0057] The method for intelligently guiding multi-target dynamic composition of breaking news events includes the following steps:

[0058] Dynamically monitor and analyze existing news events, dig out important news entities, and obtain the dynamic chain of important news events;

[0059] Automatically cluster news keywords and entities in mined news events to obtain graphs of different dimensions, including topic distribution, dissemination sources and dissemination processes, topics and dissemination processes, topics and hot figures in dissemination, and topics and semantic networks.

[0060] Conduct composition modeling for multiple important events to obtain a news relationship graph;

[0061] Based on the news relationship graph, construct an event topology spectrum tensor;

[0062] generating a dynamic feature matrix based on the event topology spectrum tensor;

[0063] Based on the dynamic feature matrix, a multi-objective optimization guidance vector field is generated;

[0064] The news relationship graph is dynamically guided according to the multi-objective optimization guidance vector field.

[0065] The present invention has the following beneficial effects:

[0066] 1. By establishing a complete news event monitoring and analysis system, we can achieve comprehensive monitoring, mining and tracing of breaking news events, effectively identify important news entities and event dynamic chains, and improve the comprehensiveness and accuracy of event discovery.

[0067] 2. Through multi-dimensional cluster analysis, we conduct in-depth analysis of news events from multiple perspectives, including topic distribution, communication sources, communication processes, hot figures, and semantic networks, to explore the deep correlations between events and improve the depth and value of the analysis results.

[0068] 3. An innovative event topology spectrum dynamic guidance algorithm is proposed, which includes three progressive stages: event topology spectrum tensor construction, dynamic feature matrix generation, and multi-objective optimization guidance vector field generation. It realizes the intelligent composition and dynamic guidance of the event relationship network, greatly improving the visualization effect and user experience.

[0069] 4. Use multi-objective optimization methods to dynamically layout and visually render event nodes, effectively highlighting important event nodes and achieving collaborative optimization of multiple objectives such as information transmission efficiency, visual guidance effect, and attention distribution. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] Figure 1 This is the overall architecture diagram of the multi-target dynamic composition intelligent guidance system for breaking news events of the present invention.

[0071] Figure 2 It is a structural diagram of the news event monitoring and analysis subsystem of the present invention.

[0072] Figure 3 It is a structural diagram of the target analysis subsystem of the present invention.

[0073] Figure 4 It is a structural diagram of the composition subsystem of the present invention.

[0074] Figure 5 It is a structural diagram of the multi-objective composition intelligent guidance subsystem of the present invention.

[0075] Figure 6 It is a schematic diagram of the construction of the event topology spectrum tensor of the present invention.

[0076] Figure 7 It is a schematic diagram of the generation of the dynamic feature matrix of the present invention.

[0077] Figure 8 It is a schematic diagram of the generation of the multi-objective optimization guided vector field of the present invention.

[0078] Figure 9 It is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0079] Please refer to the attached Figure 1-9 , the technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the present invention.

[0080] like Figure 1As shown, the multi-target dynamic composition intelligent guidance system for breaking news events provided by the present invention mainly includes a news event monitoring and analysis subsystem 1, a target analysis subsystem 2, a composition subsystem 3 and a multi-target composition intelligent guidance subsystem 4.

[0081] News event monitoring and analysis subsystem 1 is responsible for dynamically monitoring and analyzing existing news events, mining important news entities, and obtaining dynamic chains of important news events. Target analysis subsystem 2 is connected to news event monitoring and analysis subsystem 1, receiving important news entities and news event dynamic chains sent by news event monitoring and analysis subsystem 1. It automatically performs cluster analysis on news keywords and entity objects in mined news events to obtain relevant important news events and subjects. Composition subsystem 3 is connected to target analysis subsystem 2, receiving relevant important news events and subjects sent by target analysis subsystem 2, and performing composition modeling on multiple relevant important news events and subjects to obtain a news relationship graph. Multi-objective composition intelligent guidance subsystem 4 is connected to composition subsystem 3, receiving the news relationship graph sent by composition subsystem 3, constructing an event topology spectrum tensor, generating a dynamic feature matrix based on the event topology spectrum tensor, generating a multi-objective optimization guidance vector field based on the dynamic feature matrix, and dynamically guiding the news relationship graph based on the multi-objective optimization guidance vector field.

[0082] like Figure 2 As shown, the news event monitoring and analysis subsystem 1 includes a news event discovery subsystem 11, a news event extraction subsystem 12, a news event tracing and dynamic analysis subsystem 13, and a news event fusion subsystem 14.

[0083] The news event discovery subsystem 11 is used to periodically discover and pre-process news events and mine news events. Preferably, the news event discovery subsystem 11 uses an Internet crawler module to monitor the content of mainstream news sources and news blogs in real time, and simulates a browser in the form of a proxy IP through a crawler framework to dynamically load news sites. In one embodiment of the present invention, the crawler framework uses an open source framework such as Scrapy or Seleni um, configures a proxy IP pool to achieve rotating access, and avoids being blocked by news sites. In addition, to improve monitoring efficiency, the system can set the access frequency of different news sources based on historical data analysis, for example, crawling hot news websites every 5 minutes and crawling small news blogs every 30 minutes.

[0084] The news event extraction subsystem 12 is connected to the news event discovery subsystem 11 and is used to extract the elements contained in the news event. These elements include the time of occurrence of the news event, geographical information, theme, keywords, object, subject, event chain, and the evolution path of the event subject. In one embodiment of the present invention, the news event extraction subsystem 12 uses a named entity recognition (NER) algorithm combined with spatiotemporal information extraction technology to identify entities such as time expressions, place names, person names, and organizational structures in the text with an accuracy rate of more than 92%. At the same time, text themes and keywords are extracted through topic models (such as LDA, BERT, etc.), and complete events are constructed through event template matching methods.

[0085] The news event tracing and dynamic analysis subsystem 13 is connected to the news event extraction subsystem 12, and is used to dig out events with mutual correlation and analyze the dynamic relationship between each element and sub-event in the same event. Preferably, the news event tracing and dynamic analysis subsystem 13 adopts timeline analysis and event graph technology to track the development process of events and identify key turning points and influencing factors. In an embodiment of the present invention, the system constructs an event causal network and uses a PageRank algorithm variant to calculate the importance of event nodes. The importance threshold is set to 0.75 (based on the value after standardization of 0-1) to screen out event nodes with significant influence.

[0086] The news event fusion subsystem 14 is connected to the news event tracing and dynamic analysis subsystem 13 and is used to fuse multiple events within the same news event to obtain the dynamic chain of the news event. In one embodiment of the present invention, the news event fusion subsystem 14 constructs an inter-event evolution matrix based on the event relationship model and decomposes this matrix to obtain the development dynamics of the event. This method effectively addresses the issue of event fragmentation, organically integrating related events reported separately to form a complete event development context.

[0087] The monitoring and analysis process of the news event monitoring and analysis subsystem 1 specifically includes: first, using the Internet crawler module to monitor the content of mainstream news sources and news blogs in real time; second, filtering the news through the news event discovery module to obtain the news event template; then, filling the results obtained by the Internet crawler module into the event extraction module template, modifying the template, mining news events, and obtaining news titles, news time, and news content information; finally, conducting traceability analysis on the mined news events to obtain the dynamic evolution path of the subject and the dynamic evolution of the theme keywords, attributes, and important nodes in the event over time, and establishing a news event relationship model.

[0088] like Figure 3As shown, the target analysis subsystem 2 is used to automatically perform cluster analysis on news keywords and entity objects in mined news events to obtain important news events and subjects. The target analysis subsystem 2 performs cluster analysis based on the following multiple dimensions: topic distribution dimension, communication source and communication process dimension, topic and communication process dimension, topic and communication hot figure dimension, and topic and semantic network dimension.

[0089] In one embodiment of the present invention, the target analysis subsystem 2 adopts a multi-dimensional cluster analysis method, combined with algorithms such as K-means, hierarchical clustering and density clustering, to perform multi-angle analysis on event features. Preferably, the system uses a topic model (such as LDA) to extract topic distribution features in the topic distribution dimension, and the number of clusters is determined by the silhouette coefficient, which is usually selected between 3 and 7; in the dimension of propagation source and propagation process, the system combines information entropy and propagation speed characteristics to construct a propagation feature vector, and uses the DBSCAN algorithm for density clustering, with the neighborhood parameter ε set to 0.15 (empirical value, based on a large amount of historical data analysis); in the dimension of topic and hot figures, the system calculates the person-topic association strength, constructs a bipartite graph structure, and divides the community through the spectral clustering algorithm.

[0090] Furthermore, within the topic and semantic network dimensions, Target Analysis Subsystem 2 uses graph embedding techniques (such as Node2Vec) to obtain node vector representations, which are then grouped using a hierarchical clustering algorithm. The clustering results from each dimension are then combined using an improved voting ensemble method, where the weights of each dimension are dynamically adjusted based on historical accuracy, to form a final list of important news events and topics.

[0091] like Figure 4 As shown, the composition subsystem 3 uses the composition module to compose multiple important events, obtains the news relationship graph G(v,e), including key nodes v and connecting edges e, and provides node and edge descriptions for the modeling graph.

[0092] In one embodiment of the present invention, the graphing subsystem 3 uses a force-directed layout algorithm combined with a hierarchical layout strategy to achieve preliminary visualization of news event relationships. Preferably, the system uses an improved Fruchterman-Reingold algorithm to process node layout, introducing a node importance weight factor to adjust the repulsion calculation formula, bringing important nodes closer to the center of the graph. The node importance weight is calculated based on a comprehensive calculation of PageRank score, event timeliness, and attention, with weight coefficients of 0.5, 0.3, and 0.2, respectively (based on the optimal ratio verified by experiments).

[0093] At the same time, the composition subsystem 3 also optimizes edge weights. Edge weights are calculated based on event association strength, temporal correlation, and topic similarity. Min-Max normalization is used to map weights to the interval [0.1, 1.0] to ensure that even weak associations are appropriately represented. Furthermore, the system uses edge bundling technology to optimize the visual effects of complex relationship networks, reduce visual clutter, and improve graph readability.

[0094] like Figure 5 As shown, the multi-objective composition intelligent guidance subsystem 4 is the core innovation of this invention, enabling intelligent guidance of the news relationship graph. This subsystem first receives the news relationship graph sent by the composition subsystem 3 and then dynamically guides the news relationship graph using the event topology spectrum dynamic guidance algorithm. This algorithm includes three core steps: constructing the event topology spectrum tensor, generating a dynamic feature matrix, and generating a multi-objective optimization guidance vector field.

[0095] like Figure 6 As shown, the multi-target composition intelligent guidance subsystem 4 first constructs the event topology spectrum tensor, which is calculated according to the following formula:

[0096]

[0097] in: is the (i, j, k)th element of the event topology spectrum tensor, i∈[1,n], j∈[1,m], k∈[1,l] are the node index, edge index and time index respectively, w ij ∈[0,1] is the node-edge association weight coefficient, R(v i ,e j ):V×E→[0,1] is the node v i With edge e j The relationship strength function, t c ∈R + is the current time point, t ref ∈R + is the reference time point, τ∈R + is the time attenuation coefficient, is a nonlinear heat kernel function, n∈N + is the number of event nodes, m∈N + is the number of relationship edges, l∈N + is the number of time slices, V is the node set, and E is the edge set.

[0098] The complete definition of the heat kernel function:

[0099]

[0100] d topo (v i,e j ):V×E→R + For node v i With edge e j The topological distance function between c ):R + →R + is the time-varying diffusion parameter function.

[0101] The iterative algorithms need to meet the following convergence conditions:

[0102] Convergence conditions for dynamic feature matrix update:

[0103] ||F(t k+1 )-F(t k )|| F <∈ F ,

[0104] where ||·|| F is the Frobenius norm, ∈ F ∈R + Is the convergence threshold. Stability conditions of the vector field:

[0105]

[0106] in, is a bounded composition space, M V ∈R + is the upper bound of the vector field magnitude.

[0107] Boundary conditions and initial conditions:

[0108] Tensor initialization conditions:

[0109]

[0110] Matrix initialization conditions:

[0111]

[0112] Among them, SVD truncated (·) is the truncated singular value decomposition, and k0 is the initial time slice index.

[0113] This method transforms the news relationship graph into a three-dimensional topological spectral tensor, effectively integrating spatial relationships and temporal evolution. In practice, the system dynamically adjusts parameters based on the event type. For example, for breaking disaster news, the time decay coefficient τ is set to a small value (approximately 12 hours) to highlight the latest information; for long-term political and economic events, τ is set to a large value (up to 120 hours or longer) to preserve the historical context.

[0114] like Figure 7As shown, based on the event topology spectrum tensor, the multi-target composition intelligent guidance subsystem 4 further generates a dynamic feature matrix, which is calculated according to the following formula:

[0115]

[0116] Where: F(t)∈R n×n is the dynamic feature matrix at time t, F(t+Δt)∈R n×n is the dynamic characteristic matrix at time t+Δt, α∈[0,1] is the time continuity coefficient, β∈[0,1] is the curl influence coefficient, γ∈[0,1] is the spectral decomposition weight coefficient, is a three-dimensional vector field derived from the tensor, is the curl of the vector field, λ k ∈R + is the kth principal eigenvalue, W k ∈R n×n is the kth feature weight matrix, is the matrix representation of the kth tensor slice at time t, ⊙ is the Hadamard product (element-to-element product), det(C k (t))∈R is the determinant of the kth covariance matrix at time t, K∈N + is the number of main modes considered, r = [x,y,z] T ∈R 3 is the space coordinate vector.

[0117] Vector field V τ Definition of (r,t):

[0118]

[0119] where Φ ijk (r)∈R 3 is the spatial basis function vector corresponding to the (i, j, k)th tensor element, V x ,V y ,V z :R 3 ×R + →R are the component functions of the vector field in the x, y, and z directions respectively.

[0120] The specific definition of the curl operator is:

[0121]

[0122] This formula transforms the topological spectral tensor into a dynamic characteristic matrix and simulates the chaotic evolution of event relationships, capturing the nonlinear dynamic characteristics of event propagation. Preferably, the system dynamically adjusts parameters based on the speed of news event propagation. For example, for rapidly developing hot events, the β value is increased to 0.6 to strengthen the chaotic dynamics; for stable, long-term events, the γ value is increased to 0.5 to enhance the spectral decomposition effect.

[0123] like Figure 8 As shown, based on the dynamic feature matrix, the multi-objective composition intelligent guidance subsystem 4 finally generates a multi-objective optimization guidance vector field. This process is calculated according to the following formula:

[0124] V(x,y,z,t)=G potential (x,y,z)+G spectral (x,y,z,t)+G attention (x,y,z),

[0125]

[0126] Where V(x,y,z,t):R 3 ×R + →R 3 is the three-dimensional time-varying guidance vector field, G potential :R 3 →R 3 is the potential field guide component, G spectral :R 3 ×R + →R 3 is the spectral domain guided component, G attention :R 3 →R 3 is the attention field component, U(x,y,z):R 3 →R is the scalar potential function, is the gradient vector of the potential function, μ1, μ2, μ3∈[0,1] are the weight coefficients of the three components and μ1+μ2+μ3=1, is the Laplace inverse transform operator with respect to the time variable t, H F (s):C→R 3 is the vectorized Laplace transform of the dynamic feature matrix F(t), s∈C is the Laplace complex frequency variable, ω0∈R + is the system characteristic frequency, Ω(s):C→R is the frequency domain modulation function, and Re{·} is the complex real part operation a i ∈R 3 is the direction vector of the i-th attention center, r = [x, y, z] T ∈R 3 is the current spatial position vector, r i ∈R3 is the i-th attention center position vector, σ i ∈R + is the Gaussian diffusion parameter of the i-th attention region, N∈N + The number of attention centers is determined by the number of important event nodes and usually does not exceed 5.

[0127] For the dynamic feature matrix F(t)∈R n×n , first perform vectorization:

[0128]

[0129] Where vec(·) is the matrix vectorization operator, is the vectorized time function.

[0130] Laplace transform definition:

[0131]

[0132] Where s∈C, s is the Laplace complex frequency variable, the convergence domain is Re(s)>σ0 and σ0∈R is the growth exponent of the system.

[0133] This formula generates a guidance vector field that guides the placement and visual emphasis of event nodes, enabling intelligent guidance of multi-objective dynamic composition. In practice, the system dynamically adjusts parameters based on user interaction and event importance.

[0134] The present invention also provides a multi-target dynamic composition intelligent guidance method for breaking news events, such as Figure 9 As shown, the method includes the following steps:

[0135] 1) Dynamically monitor and analyze existing news events, dig out important news entities, and obtain the dynamic chain of important news events;

[0136] 2) Automatically cluster news keywords and entities in mined news events to obtain graphs of different dimensions, including topic distribution, dissemination sources and dissemination processes, topics and dissemination processes, topics and hot figures, and topics and semantic networks;

[0137] 3) Conduct composition modeling for multiple important events to obtain a news relationship graph;

[0138] 4) Based on the news relationship graph, construct the event topology spectrum tensor;

[0139] 5) Generate dynamic feature matrix based on event topology spectrum tensor;

[0140] 6) Generate a multi-objective optimization guidance vector field based on the dynamic feature matrix;

[0141] 7) Based on the multi-objective optimization guidance vector field, the news relationship graph is dynamically guided.

[0142] This approach enables full-process intelligent processing of breaking news events, from event monitoring, analysis, composition to intelligent guidance, forming a complete technical solution.

[0143] In a preferred embodiment of the present invention, the specific implementation method of step 1) is: real-time content monitoring of mainstream news sources and news blogs is performed through an Internet crawler; from the monitoring results, the news is first filtered using a news event discovery module to obtain a news event template; the results are filled into the event extraction module template, the template is modified, and news events are mined to obtain news titles, news time, and news content information; the mined news events are traced and analyzed to obtain the dynamic evolution of topic keywords, attributes, and important nodes in the events over time, and a news event relationship model is established.

[0144] In another preferred embodiment of the present invention, steps 4), 5), and 6) utilize the aforementioned event topology spectrum dynamics guidance algorithm to achieve intelligent guidance of the news relationship graph. This algorithm fully considers the temporal and spatial characteristics of events, thematic relevance, and dissemination dynamics, effectively improving the intuitiveness of the composition and guiding effectiveness.

[0145] The technical solution of the present invention is further described below through specific embodiments.

[0146] Example 1: Dynamic composition guidance for major natural disasters

[0147] In this embodiment, the system is used to handle dynamic composition guidance for news events related to a major earthquake disaster. First, the news event monitoring and analysis subsystem 1 uses an internet crawler module to capture relevant news reports, setting a high-frequency crawling frequency of 5 minutes per report, and obtaining a total of 1,200 relevant reports. The news event extraction subsystem 12 extracts key information such as time, location, casualties, and rescue operations to construct an event template. The news event tracing and dynamic analysis subsystem 13 uncovers a chain of related events, including earthquake early warning, post-earthquake rescue, and post-disaster reconstruction, and analyzes the evolutionary paths of the main entities at each stage.

[0148] Target Analysis Subsystem 2 performs multi-dimensional cluster analysis on the extracted entities. In terms of topic distribution, it identifies five clusters: "earthquake briefing," "rescue operations," "material deployment," and "post-disaster epidemic prevention." In terms of communication sources and processes, it distinguishes between official media, social networks, and rescue organizations. In terms of topics and key figures, it focuses on key figures such as rescue commanders and medical experts.

[0149] Based on the cluster analysis results, the composition subsystem 3 generates a preliminary news relationship graph G(v,e), consisting of 78 key nodes and 156 connecting edges. The multi-objective composition intelligent guidance subsystem 4 receives this relationship graph and first constructs the event topology spectrum tensor. Because this is a sudden disaster event, the time decay coefficient τ is set to 12 hours to emphasize the latest information. It then generates a dynamic feature matrix, setting β = 0.6 and γ = 0.3 to highlight the chaotic nature of the event's development. Finally, it generates a multi-objective optimization guidance vector field, setting three focus points: the epicenter, the main rescue command center, and the material distribution center.

[0150] Through this process, the system forms a clear dynamic picture of the disaster event, intuitively showing the progress of rescue operations, material distribution and changes in public attention, providing important decision-making support for disaster relief command.

[0151] Example 2: Dynamic composition guidance of long-term political and economic events

[0152] In this embodiment, the system processes news events related to an international trade negotiation. The news event monitoring and analysis subsystem 1 uses a low-frequency capture strategy (30 minutes / time) to track relevant news over a long period of time, acquiring a total of 3,500 reports. The news event extraction subsystem 12 extracts information such as negotiation participants, negotiation content, and position changes. The news event tracing and dynamic analysis subsystem 13 tracks the historical background of the negotiation, the progress of multiple rounds of negotiations, and key turning points.

[0153] Target analysis subsystem 2 identifies topic clusters such as "trade policy," "industry impact," and "market reaction" in the topic distribution dimension; distinguishes between different information sources such as official statements, expert analysis, and market reports in the communication source and communication process dimension; and constructs a complex semantic network that includes core negotiation terms, related industries, and market indicators in the topic and semantic network dimension.

[0154] The composition subsystem 3 generated a news relationship graph consisting of 120 nodes and 275 edges. When processing this relationship graph, the multi-objective composition intelligent guidance subsystem 4 set the time decay coefficient τ to 120 hours, as this is a long-term political and economic event, to preserve a longer period of historical information. When generating the dynamic feature matrix, β = 0.3 and γ = 0.5 were set to emphasize the role of spectral decomposition. The multi-objective optimization guidance vector field was set to four focal points, corresponding to the negotiating parties, the affected industries, and the financial market.

[0155] The resulting dynamic composition clearly demonstrates the complex network of position changes, market reactions, and industry impacts during the negotiation process, providing valuable decision-making reference for policy analysts and investors.

[0156] Example 3: Dynamic composition guidance for sudden social security incidents

[0157] In this embodiment, the system is used to handle a public safety emergency in a specific city. The news event monitoring and analysis subsystem 1 uses a high-frequency capture strategy of three minutes per event, paying particular attention to real-time information on social media, capturing a total of 850 pieces of relevant information. The news event extraction subsystem 12 quickly extracts key information such as the location, time, involved personnel, and emergency response.

[0158] In its multidimensional cluster analysis, Target Analysis Subsystem 2 focused specifically on the dimensions of the source and process of communication, distinguishing between official reports, eyewitness accounts, and media reports, and analyzing the paths of information dissemination. The system identified clusters of topics such as "incident description," "emergency response," and "public opinion reaction," capturing the changes in the words and actions of key actors.

[0159] The composition subsystem 3 generated a news relationship graph consisting of 65 nodes and 130 edges. When processing this relationship graph, the multi-objective composition intelligent guidance subsystem 4, taking into account the sensitivity and urgency of the incident, set the time decay coefficient τ to 8 hours and the dynamic feature matrix generation parameters β = 0.65 and γ = 0.25 to highlight the uncertainty and dynamism of the event's development. The system set two primary focus points, corresponding to the incident location and the command center.

[0160] Through this processing flow, the dynamic composition generated by the system reflects the progress of events, the spread of rumors and changes in public sentiment in real time, providing timely situational awareness and decision-making support for emergency management departments.

[0161] The above embodiments show that the multi-target dynamic composition intelligent guidance system and method for breaking news events provided by the present invention have good adaptability and practicality, and can provide effective dynamic composition and intelligent guidance for different types of news events to meet the needs of various application scenarios.

[0162] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be included within the scope of protection of the present invention.

Claims

1. A multi-target dynamic composition intelligent guidance system for breaking news events, characterized by: The system includes: News event monitoring and analysis subsystem, used for: Dynamically monitor and analyze existing news events to identify important news entities; Get dynamic links of important news events; The target analysis subsystem is connected to the news event monitoring and analysis subsystem and is used to: Receiving important news entities and news event dynamic links sent by the news event monitoring and analysis subsystem; Based on the important news entities and news event dynamic chains, automatically cluster analysis is performed on news keywords and entity objects in mined news events; Obtain relevant important news events and subjects; The imaging subsystem is connected to the target analysis subsystem and is used to: Receive relevant important news events and subjects sent by the target analysis subsystem; Conducting composition modeling for a plurality of the relevant important news events and subjects; Get the news relationship diagram; The multi-target composition intelligent guidance subsystem is connected to the composition subsystem and is used to: receiving the news relationship graph sent by the composition subsystem; Construct event topology spectrum tensor; generating a dynamic feature matrix based on the event topology spectrum tensor; generating a multi-objective optimization guidance vector field based on the dynamic feature matrix; The news relationship graph is dynamically guided according to the multi-objective optimization guidance vector field.

2. The system according to claim 1, wherein: The news event monitoring and analysis subsystem specifically includes: News event discovery subsystem, used for periodic discovery and pre-processing of news events, and mining of news events; A news event extraction subsystem, connected to the news event discovery subsystem, is used to extract elements contained in news events, including the time of occurrence of the news event, geographical information, theme, keywords, object, subject, event chain, and event subject evolution path; The news event tracing and dynamic analysis subsystem is connected to the news event extraction subsystem and is used to mine events with mutual correlation and analyze the dynamic relationship between various elements and sub-events in the same event; The news event fusion subsystem is connected to the news event tracing and dynamic analysis subsystem and is used to fuse multiple events in the same news event to obtain a dynamic chain of news events.

3. The system according to claim 2, characterized in that The monitoring and analysis process of the news event monitoring and analysis subsystem specifically includes: Use Internet crawler modules to monitor the content of mainstream news sources and news blogs in real time; Filter news through the news event discovery module to obtain news event templates; Fill the results obtained by the Internet crawler module into the event extraction module template, modify the template, mine news events, and obtain news titles, news time, and news content information; The news events obtained by the mining are traced and analyzed to obtain the dynamic evolution path of the subject and the dynamic evolution of the theme keywords, attributes, and important nodes in the event over time, and a news event relationship model is established.

4. The system according to claim 1, wherein: The target analysis subsystem performs cluster analysis based on the following dimensions: Topic distribution dimension; The dimensions of communication sources and communication processes; Theme and communication process dimensions; The dimension of themes and hot figures in communication; Thematic and semantic network dimensions; The target analysis subsystem integrates the cluster analysis results of the multiple dimensions to obtain important news events and subjects.

5. The system according to claim 1, wherein: The composition subsystem uses a composition module to compose multiple important events, obtains a news relationship graph G(v,e), including key nodes v and connecting edges e, and provides node and edge descriptions for the modeling graph.

6. The system according to claim 1, wherein: The event topology spectrum tensor construction process in the multi-objective composition intelligent guidance subsystem is calculated according to the following formula: in: is the (i, j, k)th element of the event topology spectrum tensor, i∈[1,n], j∈[1,m], k∈[1,l] are the node index, edge index and time index respectively, w ij ∈[0,1] is the node-edge association weight coefficient, R(v i ,e j ):V×E→[0,1] is the node v i With edge e j The relationship strength function, t c ∈R + is the current time point, t ref ∈R + is the reference time point, τ∈R + is the time attenuation coefficient, is a nonlinear heat kernel function, n∈N + is the number of event nodes, m∈N + is the number of relationship edges, l∈N + is the number of time slices, V is the node set, and E is the edge set.

7. The system according to claim 6, characterized in that The dynamic feature matrix generation process in the multi-objective composition intelligent guidance subsystem is calculated according to the following formula: Where: F(t)∈R n×n is the dynamic feature matrix at time t, F(t+Δt)∈R n×n is the dynamic characteristic matrix at time t+Δt, α∈[0,1] is the time continuity coefficient, β∈[0,1] is the curl influence coefficient, γ∈[0,1] is the spectral decomposition weight coefficient, is a three-dimensional vector field derived from the tensor, is the curl of the vector field, λ k ∈R + is the kth principal eigenvalue, W k ∈R n×n is the kth feature weight matrix, is the matrix representation of the kth tensor slice at time t, ⊙ is the Hadamard product (element-to-element product), det(C k (t))∈R is the determinant of the kth covariance matrix at time t, K∈N + is the number of main modes considered, r = [x,y,z] T ∈R 3 is the space coordinate vector.

8. The system according to claim 7, characterized in that The multi-objective optimization guidance vector field generation process in the multi-objective composition intelligent guidance subsystem is calculated according to the following formula: V(x,y,z,t)=G potential (x,y,z)+G spectral (x,y,z,t)+G attention (x,y,z), Where V(x,y,z,t):R 3 ×R + →R 3 is the three-dimensional time-varying guidance vector field, G potential :R 3 →R 3 is the potential field guide component, G spectral :R 3 ×R + →R 3 is the spectral domain guided component, G attention :R 3 →R 3 is the attention field component, U(x,y,z):R 3 →R is the scalar potential function, is the gradient vector of the potential function, μ1, μ2, μ3∈[0,1] are the weight coefficients of the three components and μ1+μ2+μ3=1, is the Laplace inverse transform operator with respect to the time variable t, H F (s):C→R 3 is the vectorized Laplace transform of the dynamic feature matrix F(t), s∈C is the Laplace complex frequency variable, ω0∈R + is the system characteristic frequency, Ω(s):C→R is the frequency domain modulation function, and Re{·} is the complex real part operation a i ∈R 3 is the direction vector of the i-th attention center, r = [x, y, z] T ∈R 3 is the current spatial position vector, r i ∈R 3 is the i-th attention center position vector, σ i ∈R + is the Gaussian diffusion parameter of the i-th attention region, N∈N + The number of attention centers is determined by the number of important event nodes and usually does not exceed 5.

9. The system according to claim 1, wherein: The news event monitoring and analysis subsystem consists of an Internet crawler module, a news event discovery module, a news event extraction module, and an event tracing and dynamic analysis module; The Internet crawler module simulates a browser by proxy IP through a crawler framework and dynamically loads news sites; The news event discovery module filters the news to obtain a news event template, fills the extracted information into the template, modifies the template, mines news events, and obtains news titles, news time, and news content information; The news event extraction module extracts event elements from news events and reassembles the event elements into complete independent events; The event tracing and dynamic analysis module performs tracing analysis on the acquired news events, obtains the dynamic evolution path of the subject and the dynamic evolution of the theme keywords, attributes, and important nodes in the event over time, and establishes a news event relationship model.

10. A multi-target dynamic composition intelligent guidance method for breaking news events, applied to the system according to any one of claims 1 to 9, characterized in that: The following steps are involved: Dynamically monitor and analyze existing news events, dig out important news entities, and obtain the dynamic chain of important news events; Automatically cluster news keywords and entities in mined news events to obtain graphs of different dimensions, including topic distribution, dissemination sources and dissemination processes, topics and dissemination processes, topics and hot figures in dissemination, and topics and semantic networks. Conduct composition modeling for multiple important events to obtain a news relationship graph; Based on the news relationship graph, construct an event topology spectrum tensor; generating a dynamic feature matrix based on the event topology spectrum tensor; generating a multi-objective optimization guidance vector field based on the dynamic feature matrix; The news relationship graph is dynamically guided according to the multi-objective optimization guidance vector field.

Images