A multi-dimensional double-layer public opinion knowledge graph construction method, system, device and medium

Abstract

The application discloses a kind of multi-dimensional double-layer public opinion knowledge graph construction method, system, equipment and medium, multi-dimensional double-layer public opinion knowledge graph construction method includes: obtaining unstructured data and processing it, obtain data field;The data field is carried out entity extraction, obtain multi-dimensional entity, and the multi-dimensional entity is carried out relationship extraction, obtain multi-dimensional relationship structure, and then according to the multi-dimensional entity and the multi-dimensional relationship structure simultaneously constructs topic graph, event graph and space graph;Obtain external knowledge graph, and then the topic graph, the event graph, the space graph and the external knowledge graph are fused, obtain macroscopic layer public opinion knowledge graph;According to the macroscopic layer public opinion knowledge graph, microcosmic layer public opinion knowledge graph is constructed.The application can improve the comprehensiveness of system correlation design, meet multi-scene application and effectively reduce time overhead, easy to expand, can be widely applied in public opinion data management and monitoring field.

Description

Technical Field

[0001] This invention relates to the field of public opinion data management and monitoring, and in particular to a method, system, device and medium for constructing a multi-dimensional, two-layer public opinion knowledge graph. Background Technology

[0002] Knowledge graphs represent knowledge in a graph structure, describing the relationships between entities, concepts, relations, and attributes to form a semantic network that provides rich representations and semantic reasoning. In recent years, knowledge graphs have been increasingly applied to the field of online public opinion. By focusing on knowledge data such as social opinion, and based on the collection of public opinion data, knowledge graphs are constructed to conduct public opinion monitoring, sentiment analysis, and public opinion trend prediction, providing a more comprehensive and in-depth perspective for public opinion management, public opinion guidance, and decision support. However, current graph construction methods have some shortcomings in utilizing the information characteristics of online public opinion and in practical applications, mainly in the following aspects:

[0003] First, the design of the correlation between multiple dimensions is not comprehensive. Although the individual online public opinion topic graph contains complete topic data content, comments, etc., and has two dimensions of users and topics, some related events are missing. The design of the correlation between multiple dimensions is not comprehensive enough.

[0004] Second, it focuses more on the construction of network public opinion knowledge graphs in specific dimensions or from specific perspectives. Even if the graph involves multiple dimensions, very few network public opinion graph construction methods can meet the needs of multi-scenario applications and present public opinion information in a more comprehensive and complete way.

[0005] Third, it is difficult to meet the high efficiency and scalability required for processing massive amounts of public opinion data, and it ignores the problems of high time consumption, complex construction methods and difficulty in expansion. Summary of the Invention

[0006] To address the aforementioned technical problems, the present invention aims to provide a method, system, device, and medium for constructing a multi-dimensional, two-layer public opinion knowledge graph, which can improve the comprehensiveness of correlation design, meet the needs of multi-scenario applications, effectively reduce time overhead, and is easy to expand.

[0007] The technical solution adopted in one aspect of this invention is:

[0008] A method for constructing a multidimensional, two-layer public opinion knowledge graph includes the following steps:

[0009] Unstructured data is acquired and processed to obtain data fields;

[0010] Entity extraction is performed on the data fields to obtain multidimensional entities, and pairwise relationships are extracted from the multidimensional entities to obtain multidimensional relationship structures. Then, topic graphs, event graphs, and spatial graphs are constructed simultaneously based on the multidimensional entities and the multidimensional relationship structures.

[0011] An external knowledge graph is obtained, and then the topic graph, the event graph, the spatial graph, and the external knowledge graph are merged to obtain a macro-level public opinion knowledge graph;

[0012] Construct a micro-level public opinion knowledge graph based on the macro-level public opinion knowledge graph;

[0013] The multidimensional entities include user entities, topic entities, person entities, event entities, and spatial entities.

[0014] Furthermore, the unstructured data includes user data and public opinion topic data. The step of acquiring unstructured data and processing it to obtain data fields specifically includes:

[0015] Obtain the user data and public opinion topic data from the social network;

[0016] The user data and the public opinion topic data are cleaned to obtain data fields;

[0017] The data fields include user ID, topic name, topic text, and time information.

[0018] Furthermore, the step of extracting entities from the data fields to obtain multidimensional entities specifically includes:

[0019] The user entity is constructed based on the user identifier;

[0020] Construct the topic entity based on the topic name;

[0021] The topic text is segmented and named entity recognition is performed to obtain the name information entity, and then the name information entity is labeled to obtain the person entity;

[0022] The topic text is segmented, part-of-speech tagging is performed and syntactic analysis is performed to obtain the first event information. The first event information is then extracted according to the preset event framework, event trigger words and event type to obtain the event entity.

[0023] The topic text is subjected to named entity recognition to obtain spatial information entities. These spatial information entities are then segmented according to country / region and province to obtain spatial entities. Further, the step of extracting pairwise relationships from the multidimensional entities to obtain a multidimensional relationship structure, and then simultaneously constructing a topic graph, event graph, and spatial graph based on the multidimensional entities and the multidimensional relationship structure, specifically includes:

[0024] The relationships between each pair of the topic entity, the user entity, and the person entity are analyzed, and the relationships between multiple topic entities are also analyzed to obtain a first relationship structure. Then, the topic graph is constructed using the topic entity, the user entity, and the person entity as nodes and the first relationship structure as edges.

[0025] The association between the event entity and the topic entity is analyzed, and the association between multiple event entities is analyzed to obtain a second relationship structure. Then, the event graph is constructed with the event entity and the topic entity as nodes and the second relationship structure as edges.

[0026] The relationship between the spatial entity and the topic entity is analyzed to obtain a third relationship structure. Then, the spatial graph is constructed using the spatial entity and the topic entity as nodes and the third relationship structure as edges.

[0027] Furthermore, the step of acquiring an external knowledge graph and then fusing the topic graph, the event graph, the spatial graph, and the external knowledge graph to obtain a macro-level public opinion knowledge graph specifically includes:

[0028] The topic graph, event graph, and spatial graph are fused based on the topic entities to obtain a composite knowledge graph;

[0029] Obtain the external knowledge graph, extract information from the external knowledge graph, and obtain person information and second event information;

[0030] The composite knowledge graph is supplemented based on the person information and the second event information to obtain the macro-level public opinion knowledge graph.

[0031] Furthermore, the micro-level public opinion knowledge graph includes a topic propagation graph and an abstract reasoning graph.

[0032] Furthermore, the step of constructing a micro-level public opinion knowledge graph based on the macro-level public opinion knowledge graph specifically includes:

[0033] Based on the macro-level public opinion knowledge graph, a fine-grained micro-analysis is performed on the topic entities, user entities, and time information to obtain the topic propagation graph;

[0034] Based on the macro-level public opinion knowledge graph, the event graph is vectorized, merged, and abstracted to obtain the abstract event logic graph.

[0035] Another aspect of the technical solution adopted by the present invention is:

[0036] A multi-dimensional, two-layer public opinion knowledge graph construction system includes:

[0037] The data processing module is used to acquire unstructured data and process the unstructured data to obtain data fields;

[0038] The sub-graph parallel construction module is used to extract entities from the data fields to obtain multi-dimensional entities, and to extract pairwise relationships from the multi-dimensional entities to obtain multi-dimensional relationship structures. Then, topic graphs, event graphs and spatial graphs are constructed synchronously based on the multi-dimensional entities and the multi-dimensional relationship structures.

[0039] The macro-level construction module is used to acquire external knowledge graphs, and then integrate the topic graph, the event graph, the spatial graph, and the external knowledge graph to obtain a macro-level public opinion knowledge graph.

[0040] The micro-layer construction module is used to construct a micro-layer public opinion knowledge graph based on the macro-layer public opinion knowledge graph.

[0041] The multidimensional entities include user entities, topic entities, person entities, event entities, and spatial entities.

[0042] Another aspect of the technical solution adopted by the present invention is:

[0043] A computer device, comprising:

[0044] At least one processor;

[0045] At least one memory for storing at least one program;

[0046] When the at least one program is executed by the at least one processor, the at least one processor implements the multi-dimensional two-layer public opinion knowledge graph construction method described above.

[0047] Another aspect of the technical solution adopted by the present invention is:

[0048] A computer-readable storage medium storing a processor-executable program, characterized in that the processor-executable program, when executed by the processor, is used to implement the aforementioned multi-dimensional two-layer public opinion knowledge graph construction method.

[0049] The beneficial effects of this invention are as follows: The multi-dimensional, two-layer public opinion knowledge graph construction method, system, device, and medium of this invention, on the one hand, comprehensively consider the pairwise relationships between user entities, topic entities, person entities, event entities, and spatial entities, which can improve the comprehensiveness of the system's relational design; on the other hand, through the two-layer knowledge graph structure from both macro and micro perspectives, it can be applied to various scenarios such as daily macro-level rapid control, sudden situations, event evolution analysis, and key path and user analysis; in addition, by constructing the topic graph, event graph, and spatial graph in parallel, and by merging and supplementing the three sub-graphs with external knowledge graphs, it can reduce time overhead, simplify the construction process, and achieve high efficiency and easy scalability in the construction of public opinion knowledge graphs. Attached Figure Description

[0050] Figure 1 A flowchart illustrating the steps of a method for constructing a multidimensional, two-layer public opinion knowledge graph, as provided in this embodiment of the invention.

[0051] Figure 2 A schematic diagram of a multi-dimensional, two-layer public opinion knowledge graph provided in an embodiment of the present invention;

[0052] Figure 3 A flowchart illustrating a method for constructing a multidimensional, two-layer public opinion knowledge graph, provided in an embodiment of the present invention;

[0053] Figure 4 This is a schematic diagram of the structure of a multi-dimensional, two-layer public opinion knowledge graph construction system provided in an embodiment of the present invention;

[0054] Figure 5 This is a schematic diagram of the structure of a computer device provided in an embodiment of the present invention. Detailed Implementation

[0055] With the rapid development of the internet, people's traditional ways of obtaining and exchanging information have undergone fundamental changes. They express their attitudes towards social events through various internet means, including Weibo, WeChat, short videos, and news commentary, thus generating online public opinion. In recent years, knowledge graphs have been increasingly applied to the field of online public opinion. By focusing on knowledge data such as social opinion, and based on the collection of public opinion data, a public opinion knowledge graph is constructed for public opinion monitoring, sentiment analysis, and public opinion trend prediction, providing a more comprehensive and in-depth perspective for public opinion management, public opinion guidance, and decision support. However, existing methods for constructing online public opinion knowledge graphs, while considering multiple dimensions of public opinion, lack comprehensive design for the correlation between these dimensions. Furthermore, most methods only address a single perspective, failing to meet the needs of multi-scenario applications. Finally, current methods neglect issues such as high time consumption, complex construction methods, and difficulty in scalability.

[0056] To address this, this invention proposes a multi-dimensional, two-layer public opinion knowledge graph construction method. On one hand, it comprehensively considers the pairwise relationships between user entities, topic entities, person entities, event entities, and spatial entities, improving the comprehensiveness of the system's relational design. On the other hand, through a two-layer knowledge graph structure with both macro and micro perspectives, it is applicable to various scenarios, including daily macro-level rapid control, emergency situations, event evolution analysis, and key path and user analysis. Furthermore, by constructing the topic graph, event graph, and spatial graph in parallel and integrating these three sub-graphs with external knowledge graphs, it reduces time overhead, simplifies the construction process, and achieves high efficiency and scalability in public opinion knowledge graph construction. This multi-dimensional, two-layer public opinion knowledge graph construction method can be applied to public opinion hotspot prediction, government knowledge management, public opinion space early warning, public opinion monitoring and management, topic propagation path query, public opinion key node control, rumor tracing, emergency event tracing, and event decision-making, but is not limited to these applications.

[0057] This invention proposes a method for constructing a multi-dimensional, two-layer public opinion knowledge graph, such as... Figure 1 As shown, the method includes steps S101 to S104:

[0058] S101. Obtain unstructured data and process it to obtain data fields;

[0059] Specifically, unstructured data from social networks can be obtained through technologies such as web crawlers.

[0060] As an optional implementation, the unstructured data includes user data and public opinion topic data. Step S101 specifically includes the following steps S1011 and S1012:

[0061] S1011. Obtain user data and public opinion topic data from social networks;

[0062] Specifically, user data includes, but is not limited to, the number of posts, followers, followings, reposts, likes, comments, and user information. Public opinion topic data includes, but is not limited to, the text data corresponding to the topic, the topic description, and the topic start time.

[0063] S1012. Clean up user data and public opinion topic data to obtain data fields.

[0064] Specifically, missing values, duplicate fields, empty strings, and garbled data in user data and public opinion topic data are filtered and deleted, and text links related to images and videos are removed.

[0065] The data fields include topic name, topic text, topic content, topic type, topic description, time information, user ID, username, user authentication information, user type, user location data, number of followers, number of followings, number of posts, number of comments, number of reposts, and number of likes, but are not limited to these.

[0066] S102. Extract entities from data fields to obtain multidimensional entities, and extract pairwise relationships from multidimensional entities to obtain multidimensional relationship structures. Then, construct topic graphs, event graphs, and spatial graphs simultaneously based on multidimensional entities and multidimensional relationship structures.

[0067] Among them, multidimensional entities include user entities, topic entities, person entities, event entities, and spatial entities.

[0068] Specifically, five-dimensional entities—topics, people, users, events, and spaces—are extracted. These multi-dimensional entities are then divided into three parts: topics, users, people, and topics / events, as well as topics / spaces, all of which share a common topic entity. Based on this division, features related to these three dimensions are selected, and corresponding multi-dimensional sub-graphs are constructed in parallel. Parallel construction significantly reduces time consumption and is a prerequisite for efficiently generating macro-level public opinion knowledge graphs.

[0069] As an optional implementation, entity extraction of the data fields to obtain multidimensional entities can be divided into the following steps S1021 to S1025:

[0070] S1021. Construct the user entity based on the user identifier;

[0071] The core attributes of a user entity include: content, user authentication information, user type, number of followers, number of followings, time information, and location data.

[0072] S1022. Construct a topic entity based on the topic name;

[0073] The core attributes of a topic entity include: time information, topic text, topic content, topic description, and topic type.

[0074] S1023. Perform word segmentation and named entity recognition on the topic text to obtain personal name information entities, and then label the personal name information entities to obtain person entities;

[0075] Specifically, Chinese word segmentation tools such as jieba and LTP can be used to segment the topic text, and then named entity recognition can be performed on the segmented topic text. The identified personal name information entities can be labeled to obtain the person entity. The core attributes of the person entity include: person profile.

[0076] S1024. Perform word segmentation, part-of-speech tagging, and syntactic analysis on the topic text to obtain the first event information, and extract the first event information according to the preset event framework, event trigger words, and event type to obtain the event entity;

[0077] Specifically, events are extracted from a corpus containing numerous topics. The text in the corpus undergoes preprocessing, primarily including word segmentation, part-of-speech tagging, and syntactic analysis. Based on a predefined event framework, and combining event trigger words and event types, the BILSTM+CRF algorithm is used to extract event elements. BILSTM+CRF is a popular sequence labeling algorithm that combines BILSTM and CRF, allowing the model to consider the correlation between sequences like CRF while possessing the feature extraction and fitting capabilities of LSTM. The core attributes of event entities include: time information, event type, and event duration.

[0078] S1025. Perform named entity recognition on the topic text to obtain spatial information entities, and then divide the spatial information entities according to country, region and province to obtain spatial entities.

[0079] Specifically, the Named Entity Recognition (NER) method is used to extract and identify spatial information entities in a topic text. The identified spatial information entities are then divided according to country, region and province to obtain spatial entities, such as "Sichuan" and "France".

[0080] As an optional implementation, pairwise relationships are extracted from the multidimensional entities to obtain a multidimensional relationship structure. The step of simultaneously constructing a topic graph, event graph, and spatial graph based on the multidimensional entities and the multidimensional relationship structure can be divided into the following steps S1026-S1028:

[0081] S1026. Analyze the pairwise relationships between topic entities, user entities, and person entities, and analyze the relationships between multiple topic entities to obtain the first relationship structure. Then, construct a topic graph with topic entities, user entities, and person entities as nodes and the first relationship structure as edges.

[0082] Specifically, such as Figure 2 As shown, the relationship between user entities and topic entities is divided into four types based on the user's behavior information in a topic: posting, forwarding, commenting, and liking.

[0083] Regarding the association between user entities and person entities: Since the relationship between actual people and users is a complex issue that requires the combination of multiple methods and data sources, this embodiment of the invention only targets people involved in the topic. It performs keyword searches in the social network and then matches people based on user authentication information. If a match is found, the person is associated with the user. If a match is found but the user does not exist in the created graph, the user entity node is created and associated with the person entity node.

[0084] Regarding the association between topic entities and person entities: After analyzing the topic text and extracting the person using the Named Entity Recognition (NER) method, the topic is associated with the extracted person;

[0085] Regarding the relationship between topic entities: The topic text is vectorized, and then word segmentation data is obtained by processing it according to the Word2Vec model. The vectors of the words that make up the topic text and their average values ​​are used as the vectors of the topic entities. The similarity between topic entities is calculated and compared with a threshold. If the similarity is higher than the predetermined threshold, there is a relationship between the topic entities, that is, a logical or derivative relationship.

[0086] In summary, the extracted topic entities, user entities, and person entities are used as nodes, and the above relationship structure is used as edges to construct a topic graph.

[0087] S1027. Analyze the relationship between event entities and topic entities, and analyze the relationship between multiple event entities to obtain the second relationship structure. Then, construct an event graph with event entities and topic entities as nodes and the second relationship structure as edges.

[0088] Specifically, such as Figure 2As shown, regarding the association between topic entities and event entities: the topic text is analyzed and clustered using the K-means clustering algorithm. Then, the event attributes are extracted from the clustered text results using the Named Entity Recognition (NER) method: time, event process (i.e., brief event text and keywords). The topic is then associated with the events extracted from it. At the same time, if there is a relationship between topics (logical or derivative relationship), the event entity node with the earliest time attribute is selected from the events extracted from both and associated with the associated topic entity node.

[0089] Regarding the relationships between event entities: The event logic relationships are extracted from the text results obtained by the topic text clustering algorithm using pattern matching. First, temporal, causal, and sequential relationship templates are constructed, and related words are extracted. Then, logical event pairs in the topic are identified and extracted to establish logical relationships between event entities. At the same time, new event entities are generated. The new event entities generated in this topic are then associated with the existing event entities in this topic.

[0090] In summary, the extracted event entities and topic entities are used as nodes, and the relationships between topic entities and event entities, as well as the logical relationships between event entities, are used as edges to construct an event graph.

[0091] S1028. Analyze the relationship between spatial entities and topic entities to obtain the third relationship structure, and then construct a spatial graph with spatial entities and topic entities as nodes and the third relationship structure as edges.

[0092] Specifically, such as Figure 2 As shown, the relationship between spatial entities and topic entities is determined by classifying spatial entities according to country, region, and province. For example, if there are many data entries in a topic, and the spatial information of one data entry is "Chengdu", which can be matched with the spatial entity "Sichuan", then spatial information is extracted from all data entries in that topic. If the number of successful matching with "Sichuan" exceeds a threshold, then the topic in the graph is associated with the spatial entity "Sichuan".

[0093] In summary, the extracted spatial entities and topic entities are used as nodes, and the relationships between spatial entities and topic entities are used as edges to construct a spatial graph.

[0094] S103. Obtain external knowledge graphs, and then integrate topic graphs, event graphs, spatial graphs and external knowledge graphs to obtain macro-level public opinion knowledge graphs;

[0095] Specifically, the generated topic graph, event graph, and spatial graph are fused into sub-graphs to complete the important steps of entity alignment and relationship alignment. Then, the acquired external knowledge graph is supplemented and fused to construct a macro-level public opinion knowledge graph. Through the macro-level public opinion knowledge graph, real-time macro-level public opinion monitoring applications such as public opinion heat prediction, government knowledge management, and public opinion space early warning are focused.

[0096] As an optional implementation, step S103 can be further divided into the following steps S1031 to S1033:

[0097] S1031. Based on the topic entities, the topic graph, event graph, and spatial graph are fused to obtain a composite knowledge graph;

[0098] Specifically, since there are topic entities of the same type in the three subgraphs, entity alignment can be achieved through rule matching. This will match and align the entities, eliminate duplicate topic entity nodes, and connect the entities in different subgraphs. User entities in the topic graph have location data attributes, which can be connected to spatial entity nodes in the spatial graph, so that each user is associated with at least one spatial node.

[0099] S1032. Obtain an external knowledge graph, extract information from the external knowledge graph, and obtain information about the person and the second event.

[0100] Specifically, after the three sub-graphs complete entity alignment and relation alignment to obtain a composite knowledge graph, information about people and second events is extracted from publicly available external knowledge graphs such as CN-Dbpedia. The information about people and second events includes people's occupations, family relationships, personality traits, and experiences.

[0101] S1033. Supplement the composite knowledge graph with information on the person and the second event to obtain a macro-level public opinion knowledge graph.

[0102] Specifically, the person and event information extracted from the external knowledge graph will be added to the person entity attributes of the macro-level public opinion knowledge graph to increase more event information and features, forming a larger macro-level public opinion knowledge graph based on multi-dimensional entities of topics, events, users, people, and space. This will enable information between different subgraphs to access and interact with each other, facilitating relationship reasoning in the public opinion knowledge graph.

[0103] As mentioned earlier, the macro-level public opinion knowledge graph is constructed from the dimensions of topics, events, users, people, and spatial entities. It is used for public opinion heat prediction, government knowledge management, and public opinion space early warning. For example, from a macro perspective, it can perform knowledge reasoning on the correlation between users' participation in multiple topics and the correlation between topics and spaces.

[0104] S104. Construct a micro-level public opinion knowledge graph based on the macro-level public opinion knowledge graph.

[0105] Specifically, the micro-level public opinion knowledge graph performs micro-level fine-grained analysis on the macro-level public opinion knowledge graph, which can cover more details and features to describe public opinion more comprehensively and precisely. Through the micro-level public opinion knowledge graph, we can focus on the source tracing of emergencies such as topic dissemination paths, key node control, rumor tracing, and response-type knowledge management, as well as the micro-level control of decision-making scenarios.

[0106] As an optional implementation, the micro-level public opinion knowledge graph includes a topic propagation graph and an abstract reasoning graph.

[0107] Among them, the topic propagation graph is constructed based on topic entities, user entities, and time information, which can further analyze the topic propagation path, filter time information, and understand the topic's propagation trend and influence in mainland China at different time periods.

[0108] Abstract event graphs are constructed based on event entities, enabling higher-level expression and revelation of events. They represent the laws and patterns of event evolution, providing decision-making basis for the actual management and control of major emergencies, and enabling the development of efficient response plans that conform to the laws of event evolution.

[0109] As an optional implementation, step S104 can be further divided into the following steps S1041 and S1042:

[0110] S1041. Based on the macro-level public opinion knowledge graph, perform micro-level fine-grained analysis on topic entities, user entities, and time information to obtain the topic propagation graph;

[0111] Specifically, such as Figure 2 As shown, the nodes are topic entities and user entities in the social network. User entity nodes include the poster, liker, forwarder, and commenter of the topic. The edges are the propagation relationships of the topic among users, such as forwarding relationships, liking relationships, and commenting relationships between users. The edges have attributes, representing information such as propagation path, propagation intensity, and timestamp. The edges connecting topics indicate that the topic is derived from the topics connected to it. The edges connecting topics and users indicate that the user participated in the topic and posted in the topic. This completes the construction of the micro-level topic propagation graph.

[0112] To further refine the time dimension, the nodes and edges in the topic propagation graph often change over time. By filtering the user entity nodes and the edges between users in the graph by time range, we can focus the analysis by hour or other time periods.

[0113] S1042. Based on the macro-level public opinion knowledge graph, the event graph is vectorized, merged, and abstracted to obtain an abstract event graph.

[0114] Specifically, such as Figure 2 As shown, text data in the event graph can be vectorized using Tencent AILab's Chinese word vector data. The event vector is obtained by averaging the word vectors. Events with high similarity are then merged, and similar events are represented by an abstract event. The K-means++ algorithm is then used to select good initial cluster centers and initialize the K-means clustering algorithm. After abstract generalization, the high-frequency words of similar nodes are used as the overall label of the event cluster to obtain an abstract description of the event, thus completing the construction of the micro-level abstract event graph.

[0115] As mentioned earlier, the process of constructing a multi-dimensional, two-layer public opinion knowledge graph is as follows: Figure 3 As shown:

[0116] Step 1: Obtain unstructured data and process it;

[0117] Step 2 involves constructing three types of multidimensional sub-graphs in parallel: topic graph, event graph, and spatial graph;

[0118] Step 3: Merge the sub-graphs and supplement them with external knowledge graphs to build a macro-level public opinion knowledge graph based on topics, users, people, events and space.

[0119] Step 4 involves conducting micro-analysis of the macro-level public opinion knowledge graph to construct a micro-level public opinion knowledge graph based on topics, users, time, and events: a topic propagation graph based on topics, users, and time, and an abstract event logic graph based on events.

[0120] The above describes the method for constructing a multi-dimensional, two-layer public opinion knowledge graph according to embodiments of the present invention. It can be recognized that, compared with traditional methods for constructing public opinion knowledge graphs, embodiments of the present invention have the following characteristics:

[0121] First, it fully considers the relationships between topic entities, between topic entities and spatial entities, and between user entities and spatial entities. In the macro-level public opinion knowledge graph, based on multiple dimensions such as topics, events, users, people, and spaces, it directly associates topic nodes with users who post, forward, like, and comment. In the micro-level public opinion knowledge graph, based on multiple dimensions such as topics, users, time, and events, it conducts fine-grained analysis of public opinion, which can improve the comprehensiveness of the system's correlation design.

[0122] Second, the system constructs both macro and micro perspectives on online public opinion, enabling the simultaneous capture and mining of multi-dimensional public opinion characteristics from both perspectives, each suitable for different scenarios. The macro layer associates event entities with entities such as topics, users, and spaces, focusing on real-time macro-level public opinion monitoring applications such as public opinion heat prediction, government knowledge management, and public opinion space early warning. The micro layer, based on the multi-dimensional graph of topics, events, users, people, and spaces from the macro layer, designs topic propagation graphs and abstract event logic graphs, focusing on topic propagation paths and abstract event evolution patterns.

[0123] Third, by constructing topic graphs, event graphs, and spatial graphs in parallel and designing specific nodes and edges for sub-graphs, and supplementing them with information on individuals and events from external knowledge graphs, the construction speed of the multi-dimensional, two-layer public opinion knowledge graph can be further accelerated, making the construction process more efficient and the system easier to expand.

[0124] Reference Figure 4 This invention also provides a multi-dimensional, two-layer public opinion knowledge graph construction system, including:

[0125] The data processing module is used to acquire unstructured data and process it to obtain data fields.

[0126] The subgraph parallel construction module is used to extract entities from data fields to obtain multi-dimensional entities, and to extract pairwise relationships from multi-dimensional entities to obtain multi-dimensional relationship structures. Then, topic graphs, event graphs and spatial graphs are constructed synchronously based on multi-dimensional entities and multi-dimensional relationship structures.

[0127] The macro-level construction module is used to acquire external knowledge graphs, and then integrate topic graphs, event graphs, spatial graphs and external knowledge graphs to obtain a macro-level public opinion knowledge graph;

[0128] The micro-level construction module is used to construct a micro-level public opinion knowledge graph based on the macro-level public opinion knowledge graph.

[0129] Among them, multidimensional entities include user entities, topic entities, person entities, event entities, and spatial entities.

[0130] The content of the above-described multi-dimensional two-layer public opinion knowledge graph construction method embodiments is applicable to the multi-dimensional two-layer public opinion knowledge graph construction system embodiments. The specific functions implemented by the multi-dimensional two-layer public opinion knowledge graph construction system embodiments are the same as those of the above-described multi-dimensional two-layer public opinion knowledge graph construction method embodiments, and the beneficial effects achieved are also the same as those achieved by the above-described multi-dimensional two-layer public opinion knowledge graph construction method embodiments.

[0131] Reference Figure 5 This invention also provides a computer device, comprising:

[0132] At least one processor;

[0133] At least one memory for storing at least one program;

[0134] When at least one program is executed by at least one processor, such that at least one processor achieves Figure 1 The method for constructing a multidimensional, two-layer public opinion knowledge graph is shown.

[0135] The content of the above-described multi-dimensional two-layer public opinion knowledge graph construction method embodiments is applicable to the computer device embodiments. The specific functions implemented by the computer device embodiments are the same as those of the above-described multi-dimensional two-layer public opinion knowledge graph construction method embodiments, and the beneficial effects achieved are also the same as those achieved by the above-described multi-dimensional two-layer public opinion knowledge graph construction method embodiments.

[0136] This invention also provides a computer-readable storage medium storing a processor-executable program, which, when executed by a processor, is used to implement... Figure 1 The method for constructing a multidimensional, two-layer public opinion knowledge graph is shown.

[0137] This invention provides a computer-readable storage medium that can execute the multi-dimensional, two-layer public opinion knowledge graph construction method provided in the method embodiments of this invention. It can execute any combination of the implementation steps of the method embodiments and has the corresponding functions and beneficial effects of the method.

[0138] This invention also discloses a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device can read the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, causing the computer device to perform... Figure 1 The method shown.

[0139] In some alternative embodiments, the functions / operations mentioned in the block diagrams may not occur in the order shown in the operation diagrams. For example, depending on the functions / operations involved, two consecutively shown blocks may actually be executed substantially simultaneously, or the aforementioned blocks may sometimes be executed in reverse order. Furthermore, the embodiments presented and described in the flowcharts of this invention are provided by way of example to provide a more comprehensive understanding of the technology. The disclosed methods are not limited to the operations and logic flows presented herein. Alternative embodiments are contemplated in which the order of various operations is changed and sub-operations described as part of a larger operation are executed independently.

[0140] Furthermore, although the invention has been described in the context of functional modules, it should be understood that, unless otherwise stated, one or more of the aforementioned functions and / or features may be integrated into a single physical device and / or software module, or one or more functions and / or features may be implemented in a separate physical device or software module. It is also understood that a detailed discussion of the actual implementation of each module is unnecessary for understanding the invention. Rather, given the properties, functions, and internal relationships of the various functional modules in the apparatus disclosed herein, the actual implementation of the module will be understood within the scope of conventional skill of an engineer. Therefore, those skilled in the art can implement the invention as set forth in the claims using ordinary techniques without excessive experimentation. It is also understood that the specific concepts disclosed are merely illustrative and not intended to limit the scope of the invention, which is determined by the full scope of the appended claims and their equivalents.

[0141] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. 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.

[0142] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0143] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the aforementioned program can be printed, because the aforementioned program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or, if necessary, processing in other suitable ways, and then stored in computer memory.

[0144] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0145] In the foregoing description of this specification, references to terms such as "one embodiment," "another embodiment," or "some embodiments" indicate that a specific feature, structure, material, or characteristic described in connection with an embodiment or example is included in at least one embodiment or example of the present invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0146] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

[0147] The above is a detailed description of the preferred embodiments of the present invention. However, the present invention is not limited to the embodiments. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of the present invention. All such equivalent modifications or substitutions are included within the scope defined by the claims of this application.

Claims

1. A method for constructing a multidimensional, two-layer public opinion knowledge graph, characterized in that, Includes the following steps: Unstructured data is acquired and processed to obtain data fields, which include time information. Entity extraction is performed on the data fields to obtain multidimensional entities, and pairwise relationships are extracted from the multidimensional entities to obtain multidimensional relationship structures. Then, topic graphs, event graphs, and spatial graphs are constructed simultaneously based on the multidimensional entities and the multidimensional relationship structures. An external knowledge graph is obtained, and then the topic graph, the event graph, the spatial graph, and the external knowledge graph are merged to obtain a macro-level public opinion knowledge graph; Construct a micro-level public opinion knowledge graph based on the macro-level public opinion knowledge graph; The multidimensional entities include user entities, topic entities, person entities, event entities, and spatial entities; The step of acquiring an external knowledge graph and then fusing the topic graph, the event graph, the spatial graph, and the external knowledge graph to obtain a macro-level public opinion knowledge graph specifically includes: The topic graph, event graph, and spatial graph are fused based on the topic entities to obtain a composite knowledge graph; Obtain the external knowledge graph, extract information from the external knowledge graph, and obtain person information and second event information; The composite knowledge graph is supplemented based on the person information and the second event information to obtain the macro-level public opinion knowledge graph; The micro-level public opinion knowledge graph includes a topic propagation graph and an abstract event logic graph. The step of constructing the micro-level public opinion knowledge graph based on the macro-level public opinion knowledge graph specifically includes: Based on the macro-level public opinion knowledge graph, a fine-grained micro-analysis is performed on the topic entities, user entities, and time information to obtain the topic propagation graph; Based on the macro-level public opinion knowledge graph, the event graph is vectorized, merged, and abstracted to obtain the abstract event graph. The multidimensional relationship structure includes a first relationship structure, a second relationship structure, and a third relationship structure. The topic graph uses the topic entity, the user entity, and the person entity as nodes and the first relationship structure as edges. The event graph uses the event entity and the topic entity as nodes and the second relationship structure as edges. The spatial graph uses the spatial entity and the topic entity as nodes and the third relationship structure as edges.

2. The method for constructing a multi-dimensional, two-layer public opinion knowledge graph according to claim 1, characterized in that, The unstructured data includes user data and public opinion topic data. The step of acquiring and processing the unstructured data to obtain data fields specifically includes: Obtain the user data and public opinion topic data from the social network; The user data and the public opinion topic data are cleaned to obtain data fields; The data fields also include user ID, topic name, and topic text.

3. The method for constructing a multi-dimensional, two-layer public opinion knowledge graph according to claim 2, characterized in that, The step of extracting entities from the data fields to obtain multidimensional entities specifically includes: The user entity is constructed based on the user identifier; Construct the topic entity based on the topic name; The topic text is segmented and named entity recognition is performed to obtain the name information entity, and then the name information entity is labeled to obtain the person entity; The topic text is segmented, part-of-speech tagging is performed and syntactic analysis is performed to obtain the first event information. The first event information is then extracted according to the preset event framework, event trigger words and event type to obtain the event entity. Named entity recognition is performed on the topic text to obtain spatial information entities. These spatial information entities are then divided according to country, region, and province to obtain the spatial entities.

4. The method for constructing a multi-dimensional, two-layer public opinion knowledge graph according to claim 1, characterized in that, The step of extracting pairwise relationships from the multidimensional entities to obtain a multidimensional relationship structure, and then simultaneously constructing a topic graph, an event graph, and a spatial graph based on the multidimensional entities and the multidimensional relationship structure, specifically includes: The relationships between each pair of the topic entity, the user entity, and the person entity are analyzed, and the relationships between multiple topic entities are also analyzed to obtain a first relationship structure. Then, the topic graph is constructed using the topic entity, the user entity, and the person entity as nodes and the first relationship structure as edges. The association between the event entity and the topic entity is analyzed, and the association between multiple event entities is analyzed to obtain a second relationship structure. Then, the event graph is constructed with the event entity and the topic entity as nodes and the second relationship structure as edges. The relationship between the spatial entity and the topic entity is analyzed to obtain a third relationship structure. Then, the spatial graph is constructed using the spatial entity and the topic entity as nodes and the third relationship structure as edges.

5. A multi-dimensional, two-layer public opinion knowledge graph construction system, characterized in that: include: A data processing module is used to acquire unstructured data and process the unstructured data to obtain data fields, the data fields including time information; The sub-graph parallel construction module is used to extract entities from the data fields to obtain multi-dimensional entities, and to extract pairwise relationships from the multi-dimensional entities to obtain multi-dimensional relationship structures. Then, topic graphs, event graphs and spatial graphs are constructed synchronously based on the multi-dimensional entities and the multi-dimensional relationship structures. The macro-level construction module is used to acquire external knowledge graphs, and then integrate the topic graph, the event graph, the spatial graph, and the external knowledge graph to obtain a macro-level public opinion knowledge graph. The micro-layer construction module is used to construct a micro-layer public opinion knowledge graph based on the macro-layer public opinion knowledge graph. The multidimensional entities include user entities, topic entities, person entities, event entities, and spatial entities; The step of acquiring an external knowledge graph and then fusing the topic graph, the event graph, the spatial graph, and the external knowledge graph to obtain a macro-level public opinion knowledge graph specifically includes: The topic graph, event graph, and spatial graph are fused based on the topic entities to obtain a composite knowledge graph; Obtain the external knowledge graph, extract information from the external knowledge graph, and obtain person information and second event information; The composite knowledge graph is supplemented based on the person information and the second event information to obtain the macro-level public opinion knowledge graph; The micro-level public opinion knowledge graph includes a topic propagation graph and an abstract event logic graph. The step of constructing the micro-level public opinion knowledge graph based on the macro-level public opinion knowledge graph specifically includes: Based on the macro-level public opinion knowledge graph, a fine-grained micro-analysis is performed on the topic entities, user entities, and time information to obtain the topic propagation graph; Based on the macro-level public opinion knowledge graph, the event graph is vectorized, merged, and abstracted to obtain the abstract event graph. The multidimensional relationship structure includes a first relationship structure, a second relationship structure, and a third relationship structure. The topic graph uses the topic entity, the user entity, and the person entity as nodes and the first relationship structure as edges. The event graph uses the event entity and the topic entity as nodes and the second relationship structure as edges. The spatial graph uses the spatial entity and the topic entity as nodes and the third relationship structure as edges.

6. A computer device, characterized in that, include: At least one processor; At least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements the method as described in any one of claims 1-4.

7. A computer-readable storage medium storing a processor-executable program, characterized in that, The processor-executable program, when executed by the processor, is used to implement the method of any one of claims 1-4.

Images