Public opinion event fusion method, device and equipment based on event knowledge graph

Abstract

The present invention provides a method, device and computer equipment for integrating public opinion events based on event knowledge graph, the method comprising: event identification step S1: segmenting social media and news document data, extracting frequent word itemsets, and extracting summaries of important articles as events to identify events in real society; event element extraction step S2: extracting event key elements from important documents in the event; event fusion step S3: multiple events are connected based on common key elements to form an event graph, and the community mining algorithm of graph computing is used to introduce modularity to mine the events, and finally related events are integrated to form an event knowledge base. The method can discover public opinion events, identify key elements in the events, establish event graphs based on key elements in massive data, and perform association analysis, so as to integrate various aspects of the same event to form a unified event description.

Description

Technical Field

[0001] The present invention relates to the field of computer data processing technology, and in particular to a method, device and computer equipment for integrating public opinion events based on event knowledge graphs. Background Art

[0002] In recent years, with the rise of the Internet and social media and the digital transformation of traditional media, a large amount of information has been pouring into the Internet platform, including users' reports, live broadcasts, discussions, and forwarding of social events. People directly participate in the discussion of events, but are limited by subjective differences such as the completeness of individual understanding of information, knowledge background, cognitive ability, and personality traits, which leads to one-sidedness and randomness in the dissemination of events, which can easily cause the dissemination of malicious information and adverse effects, and form huge public opinion pressure. For public opinion events, it is necessary to fully understand the time, location, key figures, behaviors, and results of the event, as well as the development of the event as the event develops. With the accumulation and improvement of event information, for new events, it is also possible to learn from past events and summarize the rules to make reasonable predictions for new events, provide a basis for research and judgment, and discover potential public opinion risk factors.

[0003] At present, when faced with explosive public opinion events, users are often unable to quickly identify valid data in the face of a massive influx of information. They are also unable to determine whether the various clues currently available to them are one-sided or lack key information. For new events, there is no way to find the development context of similar events, or to draw on historical events to analyze and respond to risk points. Summary of the invention

[0004] In view of the above, the present invention provides a method, device, and computer equipment for integrating public opinion events based on event knowledge graphs to discover public opinion events and identify key elements in events, establish event graphs based on key elements in massive data, and perform correlation analysis, thereby integrating various aspects of the same event spread at different times, different development stages, and among different groups of people to form a unified event description, and comprehensively manage events to form a historical event library, providing important support for public opinion event handling decisions. This method has the advantages of accuracy, efficiency, and flexibility.

[0005] The technical solution of the present invention:

[0006] In a first aspect, the present invention provides a method for integrating public opinion events based on event knowledge graph, comprising:

[0007] Event identification step S1: Segment the social media and news document data, extract frequent item sets, and extract summaries of important articles as events to identify real-world events;

[0008] Event element extraction step S2: extracting key elements of the event from important documents in the event;

[0009] Event fusion step S3: Multiple events are connected based on common key elements to form an event graph. The community mining algorithm of graph computing is used to introduce modularity to mine events, and finally related events are fused to form an event knowledge base.

[0010] In a second aspect, the present invention also provides a public opinion event fusion device based on an event knowledge graph, which is used to implement the above-mentioned public opinion event fusion method based on an event knowledge graph.

[0011] In a third aspect, the present invention further provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the above method when executing the computer program.

[0012] Compared with the prior art, the public opinion event fusion method, device, and computer equipment based on event knowledge graph of the present invention have the following beneficial effects:

[0013] 1. The implementation process of the present invention is fully automated. It only requires annotating a very small amount of sample data at the beginning of the process to train the model, so as to achieve the purpose of extracting accurate key elements of events from actual data. Other links are based on unsupervised algorithms, and basically no human participation in the calculation process is required.

[0014] 2. The data calculated in the three stages of the present invention are from original document data to event key element data, and then to event data fusion. The amount of data calculated decreases exponentially, and the calculation performance is excellent.

[0015] 3. The present invention performs calculations based on a graph structure, can dynamically increase the association relationship of events, and can perform calculations based on incremental data without being affected by the data source.

[0016] 4. With the increase of data, the present invention can make the relationship between events more complete; adding the time dimension to event data can achieve the effect of event tracing and event context; the data source is constructed from various vertical field data, and it can generate a vertical field event library, such as: financial risk event library, listed company behavior event library, etc., which has a very broad application scenario.

[0017] 5. The present invention can automatically divide and identify the development period of ongoing events, historical events, etc. in real time, and the entire process does not require human intervention, which can reduce human subjective errors, achieve rapid response, and provide assistance in locating and finding the internal causes of event mutations.

[0018] The present invention is based on the analysis and mining of massive document data, and uses NLP technology to segment, aggregate, and generate summaries of document content, so as to quickly identify and extract events on the entire network or events in a specific field. In terms of event key element extraction, the use of UIE technology can support a small amount of sample data training to extract event elements in event-related documents, thereby improving the performance and accuracy of event element extraction. In terms of traditional event extraction, only event recognition can be achieved, and several expressions of the same event or different stages of the same event cannot be merged. The present invention can achieve unsupervised event fusion based on event graph technology and graph community mining algorithm, and calculate event weights, thereby having a comprehensive and accurate measurement index for the event. The system corresponding to the present invention continuously runs the above logic to build an event knowledge base, and only a small amount of manual participation is required to complete the entire process of event discovery, event element extraction, event fusion, and event management.

[0019] The preferred embodiments of the present invention and their beneficial effects will be further described in detail in conjunction with specific implementation methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the following specific embodiments, they are used to explain the present invention, but should not constitute a limitation of the present invention.

[0021] Figure 1 It is the overall flow chart of data processing of the present invention;

[0022] Figure 2 It is a flowchart for extracting key elements of an event;

[0023] Figure 3 It is a schematic diagram of event graph data. DETAILED DESCRIPTION

[0024] The specific implementation of the present invention is described in detail below in conjunction with the accompanying drawings. It should be understood that the specific implementation described here is only used to illustrate and explain the present invention, and is not used to limit the present invention.

[0025] The public opinion event fusion method based on event knowledge graph provided in the embodiment of the present application can be applied to computer devices such as terminals and servers. Among them, the terminal can be, but is not limited to, various personal computers, laptops, smart phones, tablet computers and portable wearable devices, and the portable wearable devices can be head-mounted devices, etc.; the server can be implemented as an independent server or a server cluster composed of multiple servers.

[0026] See also Figure 1 and Figure 2 The present invention provides a method for integrating public opinion events based on event knowledge graph, comprising the following steps:

[0027] Event identification step S1: Segment the social media and news document data, extract frequent item sets, and extract summaries of important articles as events to identify real-world events;

[0028] Event element extraction step S2: extracting key elements of the event from important documents in the event, where the key elements include at least one of time, place, person, organization, behavior, and result;

[0029] Event fusion step S3: Multiple events are connected based on common key elements to form an event graph; modularity is introduced to mine events using a community mining algorithm based on graph computing, and finally related events are fused to form an event knowledge base.

[0030] The event identification step S1 includes the following sub-steps:

[0031] S11: Segment all social media and news data in the form of documents to obtain document and word data;

[0032] S12: Use the FPtree algorithm to construct the FPTre structure for the word lists in the documents in the same time period, and collect frequent word groups and their corresponding document relationships;

[0033] S13: For each frequent item set, select the top N documents with high interaction counts from social media documents and the top N documents with high reading counts from news documents. Extract the summary information of the documents as the event name and event description information, and record the association between the event and the document.

[0034] S14: Based on the degree of textual overlap between the event name and summary information, events with the same content are merged.

[0035] After the above four steps, we get several events and their corresponding document lists.

[0036] The event element extraction step S2 includes the following sub-steps:

[0037] S21: Train an extractive UIE model for plain text and document scenarios, use uie-base as the base model, set the schema, and fine-tune the feature extraction model suitable for events.

[0038] S21 specifically includes the following sub-steps:

[0039] S211: Given a structured pattern, annotate the training data and parse the annotated data into a structured pattern instructor SSI and a structured extraction language SEL representation;

[0040] For example:

[0041] Text: Steve became CEO of Apple in 1997.

[0042] Parse Text into SSI: [spot] person [spot]organization [spot] time;

[0043] Parse Text into SEL: ((person: Steve) (organization: Apple) (time:1997));

[0044] SEL: Structured Extraction Language, which can effectively encode different IE structures into a unified representation and unify the output structure of different tasks.

[0045] SSI: Structured Pattern Instructor, a pattern-based prompt mechanism. For entity recognition, the prompt is spot prompt + entity category. For relation extraction and event extraction, the prompt is spot prompt + trigger word, and associate prompt + relation category.

[0046] S212: Establishing a UIE model, using the sequence data of step S211 as the input and output of the UIE model, and establishing a Transformer including an Encoder and a Decoder as a basis to define the UIE model;

[0047] ;

[0048] ;

[0049] ;

[0050] ;

[0051] S213: concatenate the SSI information and the sentence in step S212 and encode them using an encoder;

[0052] ;

[0053] ;

[0054] ;

[0055] S214: decoding the output defined in step S213 by autoregression to form extracted information;

[0056] ;

[0057] .

[0058] The parameters are trained through the above steps and the UIE model is finally generated.

[0059] S22: Sort the documents in the event by the number of interactions. If the document has forwarded comments, the sum of the forwarded comments is used as the sorting basis. Select the top N documents as the material for feature extraction.

[0060] S23: The text selected in sub-step S22 is called the UIE model trained in sub-step S21 to extract at least one information element from the document, including time, place, person, organization, company, keyword, behavior, result, and information source. The irregular information is corrected or deleted through the information filter, and finally the information is merged, and the same element in the same document is retained only once.

[0061] S24: All elements extracted from the event in sub-step S23 are combined and aggregated to form an event element set.

[0062] The event fusion step S3 includes the following sub-steps:

[0063] S31: Establish a correlation between two events based on the common elements that appear simultaneously in the two events.

[0064] S31 specifically includes the following sub-steps:

[0065] S311: Grouping events according to single key elements, merging all events, and forming a data structure of a list of events corresponding to key elements;

[0066] S312: Traverse all events under the same key element to form event combinations in pairs, and associate the key element;

[0067] S313: grouping all the above two-event combinations to obtain a set of all key elements of the two-event combinations;

[0068] S314: Assume that the key element set of event A is {a1, a2, ..., an}, the key element set of event B is {b1, b2, ..., bm}, and the common element set of event A and event B is {c1, c2, ..., ck}. This triplet can be expressed as (A, {c1, c2, ..., ck}, B), where A represents event A, B represents event B, and {c1, c2, ..., ck} represents the common element list of event A and event B.

[0069] S32: Calculate the association weights of two events according to certain rules for all key element sets of two-by-two event combinations, and obtain triple data of "event A->several key elements, weight<-event B".

[0070] S32 specifically includes the following sub-steps:

[0071] S321: Assign weights to each type of factor; for example: time->1, person->2, result->0.5, etc.

[0072] S322: Calculate the weight of such triples, using the type weight and the number of occurrences of each element in the common element list, and then perform weighted summation; possible weight calculation formulas are as follows:

[0073] Triple weight = ∑ element type weight (ci) × element occurrence number (ci);

[0074] S33: The full amount of triple data is organized into graph structure data with events as nodes and several key elements as edges, such as Figure 3 .

[0075] S34: Use community mining algorithm to divide the above graph data into communities, and obtain several independent communities in the graph. The events corresponding to the points in each independent community are the same event cluster, which actually describes the same event in the real world.

[0076] S34 specifically includes the following sub-steps:

[0077] S341: The formula of the community division algorithm for graph data involves the definition of modularity, where the calculation formula of modularity Q is as follows:

[0078]

[0079] In this formula, A ij represents the connection weight between node i and node j in the graph, k i and k j is the degree of node i and node j, m is the sum of the weights of all edges in the graph, c i and c j denote the communities where nodes i and j are located, δ(c i ,c j ) is the Kronecker delta function, if c i =c j If yes, it is 1, otherwise it is 0;

[0080] The meaning of this formula is that the modularity Q measures the difference between the tightness of the connection between the nodes in the graph and the expected connection within the community. By maximizing Q, the optimal community structure of the graph can be found;

[0081] S342: Each event point joins neighboring nodes to form a community, and calculates the modularity, compares the modularity before and after Q gain, and iteratively calculates all nodes to join the community to obtain the maximum modularity gain;

[0082] S343: Reconstruct the community network with the community as the node, repeat the process of step S342, the weight of the community node is the sum of the weights of all edges between communities, the community node joins the neighbor community node, and calculates the modularity, iteratively calculates all community nodes to join the neighbor community to obtain the maximum modularity gain;

[0083] S344: Output the final community structure, such as Figure 3 All events within each community are considered a complete event.

[0084] S35: Merge the measurement information of multiple events in the above event cluster, such as the element list, the number of documents of the event, the influence, etc., and store the merged events persistently to form an event knowledge base, providing query and matching functions to the outside world.

[0085] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0086] Based on the same inventive concept, an embodiment of the present application also provides a public opinion event fusion device based on an event knowledge graph, which includes modules corresponding to the steps in the embodiment of the public opinion event fusion method based on the event knowledge graph, and is used to implement the public opinion event fusion method based on the event knowledge graph.

[0087] The public opinion event fusion device based on event knowledge graph includes:

[0088] Event recognition module: used to segment social media and news document data, extract frequent item sets, and extract summaries of important articles as events to identify real-world events;

[0089] Event element extraction module: used to extract key elements of important documents in an event. Key elements include at least one of time, place, person, organization, behavior, and result.

[0090] Event fusion module: It is used to establish connections between multiple events based on common key elements to form an event graph; it uses the community mining algorithm of graph computing to introduce modularity to mine events, and finally integrates related events to form an event knowledge base.

[0091] It should be understood that each module of the public opinion event fusion device based on the event knowledge graph is used to execute each step in the embodiment of the corresponding method, and each step in the embodiment of the corresponding method has been explained in detail in the above embodiment. Please refer to the relevant description in the embodiment of the corresponding method for details, and no further details will be given here.

[0092] Based on the same inventive concept, the embodiment of the present application also provides a computer device for implementing the above-mentioned method for fusion of public opinion events based on event knowledge graph. The implementation scheme for solving the problem provided by the computer device is similar to the implementation scheme recorded in the above method, so the specific limitations in the computer device embodiment provided below can refer to the limitations of the method for fusion of public opinion events based on event knowledge graph above, and will not be repeated here.

[0093] In one embodiment, a computer device is provided, which may be a terminal, and includes a processor, a memory, a communication interface, a display screen, and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a mobile cellular network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, a method for fusion of public opinion events based on event knowledge graphs is implemented. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covered on the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0094] In one embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the method for fusion of public opinion events based on event knowledge graph as described in the above embodiment is implemented.

[0095] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the method for merging public opinion events based on event knowledge graph as described in the above embodiment is implemented.

[0096] In one embodiment, a computer program product is provided, including a computer program, which, when executed by a processor, implements the public opinion event fusion method based on event knowledge graph as described in the above embodiment.

[0097] The method, device, computer equipment, computer-readable storage medium and computer program product for integrating public opinion events based on event knowledge graph of the present invention have the following beneficial effects:

[0098] 1. The implementation process of the present invention is fully automated. It only requires annotating a very small amount of sample data at the beginning of the process to train the model, so as to achieve the purpose of extracting accurate key elements of events from actual data. Other links are based on unsupervised algorithms, and basically no human participation in the calculation process is required.

[0099] 2. The data calculated in the three stages of the present invention are from original document data to event key element data, and then to event data fusion. The amount of data calculated decreases exponentially, and the calculation performance is excellent.

[0100] 3. The present invention performs calculations based on a graph structure, can dynamically increase the association relationship of events, and can perform calculations based on incremental data without being affected by the data source.

[0101] 4. With the increase of data, the present invention can make the relationship between events more complete; adding the time dimension to event data can achieve the effect of event tracing and event context; the data source is constructed from various vertical field data, and it can generate a vertical field event library, such as: financial risk event library, listed company behavior event library, etc., which has a very broad application scenario.

[0102] 5. The present invention can automatically divide and identify the development period of ongoing events, historical events, etc. in real time, and the entire process does not require human intervention, which can reduce human subjective errors, achieve rapid response, and provide assistance in locating and finding the internal causes of event mutations.

[0103] The present invention is based on the analysis and mining of massive document data, and uses NLP technology to segment, aggregate, and generate summaries of document content, so as to quickly identify and extract events on the entire network or events in a specific field. In terms of event key element extraction, the use of UIE technology can support a small amount of sample data training to extract event elements in event-related documents, thereby improving the performance and accuracy of event element extraction. In terms of traditional event extraction, only event recognition can be achieved, and several expressions of the same event or different stages of the same event cannot be merged. The present invention can achieve unsupervised event fusion based on event graph technology and graph community mining algorithm, and calculate event weights, thereby having a comprehensive and accurate measurement index for the event. The system corresponding to the present invention continuously runs the above logic to build an event knowledge base, and only a small amount of manual participation is required to complete the entire process of event discovery, event element extraction, event fusion, and event management.

[0104] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.

[0105] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0106] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A method for fusion of public opinion events based on event knowledge graph, characterized in that: The following steps are involved: Event identification step S1: Segment the social media and news document data, extract frequent item sets, and extract documents from important articles as events to identify events in the real world; Event element extraction step S2: extracting key elements of the event from important documents in the event; Event fusion step S3: Multiple events are connected based on common key elements to form an event graph. The community mining algorithm of graph computing is used to introduce modularity to mine events, and finally related events are fused to form an event knowledge base. The event fusion step S3 includes the following sub-steps: S31: Establishing the association between two events based on the common elements that appear simultaneously in the two events; S32: Calculate the association weights of two events according to certain rules for all key element sets of two-by-two event combinations, and obtain triple data of "event A->several key elements, weight<-event B"; S33: The full amount of triple data is organized into graph structure data with events as points and several key elements as edges; S34: Use community mining algorithm to divide the above graph structure data into communities, and obtain several independent communities in the graph. The events corresponding to the points in each independent community are the same event cluster, which actually describes the same event in the real world; S35: Merge the measurement information of multiple events in the above event cluster, and persistently store the merged events to form an event knowledge base; The S31 includes the following steps: S311: Grouping events according to single key elements, merging all events, and forming a data structure of a list of events corresponding to key elements; S312: Traverse all events under the same key element to form event combinations in pairs, and associate the key element; S313: grouping all the above two-event combinations to obtain a set of all key elements of the two-event combinations; S314: Assume that the key element set of event A is {a1, a2, ..., an}, the key element set of event B is {b1, b2, ..., bm}, and the common key element set of event A and event B is {c1, c2, ..., ck}. This triple is expressed as (A, {c1, c2, ..., ck}, B), where A represents event A and B represents event B.

2. The method for fusion of public opinion events based on event knowledge graph according to claim 1 is characterized in that: The event identification step S1 includes the following sub-steps: S11: Segment all social media and news data in the form of documents to obtain document and word data; S12: Use the FPtree algorithm to construct the FPTre structure for the word lists in the documents in the same time period, and collect frequent word groups and their corresponding document relationships; S13: For each frequent item set, select the top N documents with high interaction counts from social media documents and the top N documents with high reading counts from news documents. Extract summary information from the documents as event names and event description information, and record the association between events and documents. S14: merging events with the same content based on the degree of text overlap between the event name and summary information; After the above four steps, we get several events and their corresponding document lists.

3. According to claim 1, the method for fusion of public opinion events based on event knowledge graph is characterized in that: The event element extraction step S2 includes the following sub-steps: S21: Train an extractive UIE model based on plain text and document scenarios, use uie-base as the base model, set the schema, and fine-tune the feature extraction model suitable for events; S22: Sort the documents in the event according to the number of interactions, and select the topN documents as the material for feature extraction; S23: The document selected in sub-step S22 is called the UIE model trained in sub-step S21, and at least one information element among the time, place, person, organization, company, keyword, behavior, result, and information source in the document is extracted, and irregular information is corrected or deleted through information filter, and finally the information is merged, and the same element in the same document is retained only once; S24: All elements extracted from the event in sub-step S23 are combined and aggregated to form an event element set.

4. The method for fusion of public opinion events based on event knowledge graph according to claim 3 is characterized in that: The S21 comprises the following sub-steps: S211: Given a structured pattern, annotate the training data and parse the annotated data into a structured pattern instructor SSI and a structured extraction language SEL representation; S212: Establishing a UIE model, using the sequence data of step S211 as the input and output of the UIE model, and establishing a Transformer including an Encoder and a Decoder as a basis to define the UIE model; S213: concatenate the SSI information and the sentence in step S211 and encode them using an encoder; S214: decoding the output defined in step S213 by autoregression to form extracted information; The parameters are trained through the above steps and the UIE model is finally generated.

5. The method for fusion of public opinion events based on event knowledge graph according to claim 1 is characterized in that: The S32 comprises the following sub-steps: S321: Assign weights to each type of element; S322: Calculate the weight of such triples, using the type weight and the number of occurrences of each element in the common key element set, and then perform weighted summation.

6. The method for fusion of public opinion events based on event knowledge graph according to claim 1 is characterized in that: The S34 comprises the following sub-steps: S341: The formula of the community division algorithm for graph data involves the definition of modularity, where the calculation formula for modularity Q is as follows: ； Among them, A ij represents the connection weight between node i and node j in the graph, k i and k j is the degree of node i and node j, m is the sum of the weights of all edges in the graph, c i and c j denote the communities where nodes i and j are located, δ(c i ,c j ) is the Kronecker delta function, if c i =c j If yes, it is 1, otherwise it is 0; S342: Each event point joins neighboring nodes to form a community, and calculates the modularity, compares the modularity before and after Q gain, and iteratively calculates all nodes to join the community to obtain the maximum modularity gain; S343: Reconstruct the community network with the community as the node, repeat the process of step S342, the weight of the community node is the sum of the weights of all edges between communities, the community node joins the neighbor community node, and calculates the modularity, iteratively calculates all community nodes to join the neighbor community to obtain the maximum modularity gain; S344: Output the final community structure, with all events in each community treated as a complete event.

7. A public opinion event fusion device based on event knowledge graph, characterized in that: Used to implement the public opinion event fusion method based on event knowledge graph as described in any one of claims 1-6.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the public opinion event fusion method based on event knowledge graph described in any one of claims 1 to 6 are implemented.

Images