A method for constructing and system implementing a global knowledge graph of anti-terrorism data

Abstract

This invention relates to a method for constructing and implementing a global knowledge graph for counterterrorism data. The method includes fusing and optimizing the Maryland Global Counterterrorism Database with a custom database to obtain a global database and a global knowledge graph; extracting resource description framework triples; embedding the resource description framework triples into a semantic space; training a deep learning network through the semantic space to obtain a multi-task joint learning model; inputting the word vector space to be identified into the multi-task joint learning model to obtain new resource description framework triples; and adding the new resource description framework triples to the global knowledge graph. This invention utilizes a global knowledge graph, significantly reducing the size of the knowledge base during training, testing, and querying, while greatly improving accuracy. It also fills gaps in the global knowledge graph. The technical solution provided in this application utilizes the correlation between things to quickly determine and analyze terrorist situations, greatly meeting current application needs.

Description

Technical Field

[0001] This invention belongs to the field of big data analysis technology, specifically relating to a method for constructing and implementing a global knowledge graph of counter-terrorism data. Background Technology

[0002] Knowledge graphs, also known as knowledge domain visualizations or knowledge domain mapping maps, are a series of various graphs that display the development process and structural relationships of knowledge. They use visualization techniques to describe knowledge resources and their carriers, and to mine, analyze, construct, draw, and display knowledge and the relationships between them. A knowledge graph is a collection of triples, where each triple takes the form (subject, predicate, object), where the subject and object are entities, and the predicate is a relation. Each triple (e.g., (name, birthplace, Honolulu)) represents a fact. Knowledge graphs can be applied to many scenarios, such as information recommendation systems or text classification, and are an emerging and important intelligent method for processing various types of data across industries. Knowledge graphs are divided into general knowledge graphs and domain knowledge graphs. Domain knowledge graphs gather professional knowledge within a specific domain, allowing the construction of intelligent systems that handle specific problems within that domain. The global knowledge graph for counterterrorism data is a domain-specific knowledge graph constructed by integrating and optimizing the Maryland Global Counterterrorism Database with a custom database, and supplementing it with knowledge graph completion technology. Unlike other domain-specific knowledge graphs, the counterterrorism data knowledge graph is more complex. Specifically, it has a complex structure, potential connections between different counterterrorism data, a high degree of information hiding, and is difficult and less accurate in extracting effective information.

[0003] Among related technologies, existing methods for building global counterterrorism data knowledge graphs based on counterterrorism data knowledge graphs lack corresponding methods in terms of completion and application, resulting in low accuracy and slow judgment speed of counterterrorism data knowledge graphs, making it difficult to meet current usage needs. Summary of the Invention

[0004] In view of this, the purpose of this invention is to overcome the shortcomings of the prior art and provide a method for constructing and implementing a global knowledge graph of counter-terrorism data, so as to solve the problems of low accuracy and slow judgment speed of the existing counter-terrorism data knowledge graph, which makes it difficult to meet the current usage needs.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a method for constructing and implementing a global knowledge graph for counter-terrorism data, comprising:

[0006] Using a data source based on a counterterrorism database, the data source is extracted based on preset entity words to obtain triples related to the entity words;

[0007] A global database is established by fusing the anti-terrorism database and a custom database using the triples, a global knowledge graph is obtained using the global database, and resource description framework triples are extracted based on the global knowledge graph.

[0008] The resource description framework triples are embedded into the semantic space, and a training set is established based on the semantic space. The deep learning network is trained using the training set to obtain a multi-task joint learning model. The multi-task joint learning model is used to identify resource description framework triples.

[0009] The word vector space to be identified is input into the multi-task joint learning model to obtain a new resource description framework triplet, which is then added to the global knowledge graph.

[0010] Furthermore, the data source includes:

[0011] Structured data, semi-structured data, and unstructured data.

[0012] Furthermore, the step of extracting the data source based on preset entity words includes:

[0013] For structured data, the D2R method is used to extract knowledge from the structured data or the graph mapping method is used to extract knowledge from linked data;

[0014] For semi-structured data, a wrapper is used to extract knowledge from the semi-structured data;

[0015] For unstructured data, information extraction methods are used to extract knowledge from free text.

[0016] Furthermore, based on the global knowledge graph, resource description framework triples are extracted, including:

[0017] Calculate the Hamming distance between entity strings in the global database;

[0018] Based on the Hamming distance similarity, entity pairs that match between the anti-terrorism database and the custom database are selected, and an equivalent mapping between entities is constructed.

[0019] Based on the selected entity pairs, a breadth-first search is performed on the global anti-terrorism database to obtain entities related to the selected entity pairs.

[0020] Based on the obtained entities, extract resource description framework triples.

[0021] Furthermore, the step of inputting the word vector space to be identified into the multi-task joint learning model to obtain a new resource description framework triplet includes:

[0022] The resource description framework triplet, mapped to the word vector space, is input into the multi-task joint learning model. The knowledge embedded in the word vector matrix space is then remapped to the resource description framework triplet of the text representation, thus obtaining a new resource description framework triplet.

[0023] Furthermore, the new resource description framework triple includes multiple elements, and after obtaining the new resource description framework triple, it also includes:

[0024] The newly generated resource description framework triples are organized and merged.

[0025] Furthermore, it also includes:

[0026] Normalization processing of strings in the global anti-terrorism database.

[0027] This application provides an apparatus for constructing and implementing a global knowledge graph of counter-terrorism data, comprising:

[0028] The extraction module is used to extract data from the data source based on the anti-terrorism database and based on preset entity words to obtain triples related to the entity words.

[0029] A module is established by using the triples to fuse the anti-terrorism database and the custom database to establish a global database, using the global database to obtain a global knowledge graph, and extracting resource description framework triples based on the global knowledge graph.

[0030] The training module is used to embed the resource description framework triples into the semantic space, establish a training set based on the semantic space, and train the deep learning network using the training set to obtain a multi-task joint learning model; the multi-task joint learning model is used to identify resource description framework triples.

[0031] The completion module is used to input the word vector space to be identified into the multi-task joint learning model to obtain a new resource description framework triplet, and to fill the new resource description framework triplet into the global knowledge graph.

[0032] The beneficial effects that can be achieved by adopting the above technical solution in this invention include:

[0033] This invention provides a method for constructing and implementing a global knowledge graph for counterterrorism data. This application integrates and optimizes the Maryland Global Counterterrorism Database with a custom database to obtain a global database. A global knowledge graph is then derived from this global database, and resource description framework triples are extracted from the global knowledge graph. These resource description framework triples are embedded into a semantic space, and a deep learning network is trained using this semantic space to obtain a multi-task joint learning model. The word vector space to be identified is input into the multi-task joint learning model to obtain new resource description framework triples, which are then added to the global knowledge graph. This invention utilizes a global knowledge graph, significantly reducing the size of the knowledge base during training, testing, and querying, while greatly improving accuracy. It also automatically mines hidden knowledge within the global knowledge graph of counterterrorism data, filling a gap in knowledge graph-based methods for completing global knowledge graphs for counterterrorism data. Furthermore, the implementation of the counterterrorism data analysis system can maximize and rapidly extract valuable information from massive amounts of fragmented data, quickly identifying and analyzing terrorist situations using the relationships between things, greatly satisfying current application needs. Attached Figure Description

[0034] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0035] Figure 1 This is a schematic diagram illustrating the steps of the method for constructing and implementing a global knowledge graph of counter-terrorism data according to the present invention.

[0036] Figure 2 This is a schematic diagram of the device for constructing and implementing a global knowledge graph of counter-terrorism data according to the present invention. Detailed Implementation

[0037] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be described in detail below. Obviously, the described embodiments are merely some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other implementation methods obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0038] Terrorism refers to the claims and actions that use violence, sabotage, intimidation, or other means to create social panic, endanger public safety, infringe upon personal and property rights, or coerce state organs or international organizations to achieve their political, ideological, or other objectives. Terrorism has become the biggest challenge threatening social stability and national security, and has gradually spread from one country or region to all countries in the world, escalating into an international problem. In the internet age, terrorists use the network to transmit various kinds of information. During transmission, a great deal of information is left behind. Mastering this information allows for real-time monitoring of the basic situation of terrorist organizations and the activities of terrorists, dynamically grasping changes within and outside terrorist organizations, studying their organizational patterns, understanding their activity patterns, issuing early warnings, and proactively protecting relevant attack targets or infrastructure. Against this backdrop, using knowledge graph technology and utilizing global knowledge graphs of counterterrorism data to prevent and combat terrorism will be an important direction in counterterrorism research and a crucial means of obtaining counterterrorism intelligence. Global knowledge graphs can provide in-depth understanding of the evolution of terrorist networks and analyze their network structural characteristics, all of which will provide technical support for counterterrorism efforts.

[0039] The following describes, with reference to the accompanying drawings, a specific method for constructing and implementing a global knowledge graph for counter-terrorism data provided in an embodiment of this application.

[0040] like Figure 1 As shown in the embodiments of this application, the method for constructing and implementing a global knowledge graph of counter-terrorism data includes:

[0041] S101, using a data source based on an anti-terrorism database, extracting the data source based on preset entity words to obtain triplets related to the entity words;

[0042] S102, using the triples to fuse the anti-terrorism database and the custom database to establish a global database, using the global database to obtain a global knowledge graph, and extracting resource description framework triples based on the global knowledge graph;

[0043] S103, embed the resource description framework triples into the semantic space, establish a training set based on the semantic space, and train the deep learning network using the training set to obtain a multi-task joint learning model; the multi-task joint learning model is used to identify resource description framework triples.

[0044] The input data for deep learning networks is a set of known triples, or a defined binary tensor A[1:|R|]. For relational fact extraction models, the input data consists of sentences mentioning various entity pairs, each sentence having its own internal structure, and the co-occurrence of various elements carrying frequency information. Therefore, the input data for both is heterogeneous.

[0045] Multi-task joint learning models learn a fixed-dimensional continuous representation for each entity from an existing knowledge graph. These continuous representations are then used for computational reasoning. In contrast, relational fact extraction models can use either hand-designed traditional features or deep learning models combined with word vectors and other embeddings to encode sentences. The two also differ significantly at the feature level.

[0046] Multi-task joint learning models typically learn by optimizing a margin-based ranking criterion. During the optimization process, the model continuously distinguishes between correct entities (either the head or the tail) and incorrect entities in a triple (h, r, t). For relational fact extraction, however, probabilistic models are usually used to optimize maximum likelihood, and the objects being distinguished are various different relations. Therefore, the two models differ significantly in their application.

[0047] S104, input the word vector space to be identified into the multi-task joint learning model to obtain a new resource description framework triplet, and add the new resource description framework triplet into the global knowledge graph.

[0048] The working principle of the method for constructing and implementing a global knowledge graph for counterterrorism data is as follows: This application integrates and optimizes the Maryland Global Counterterrorism Database with a custom database to obtain a global database. A global knowledge graph is obtained based on the global database, and resource description framework triples are extracted based on the global knowledge graph. The resource description framework triples are embedded into the semantic space, and a deep learning network is trained through the semantic space to obtain a multi-task joint learning model. The word vector space to be identified is input into the multi-task joint learning model to obtain new resource description framework triples, and the new resource description framework triples are added to the global knowledge graph.

[0049] It should be noted that knowledge graph completion involves adding new triples to an existing knowledge graph, and the added triples must be objective facts. There are two main sources of information that can be used to complete a knowledge graph: 1. Inferring new triples from existing triples in the knowledge graph. 2. Extracting new entities and new triples from text. A comparative analysis of the advantages and disadvantages of these two methods leads to the proposal of a multi-task joint learning framework model to embed entities from the knowledge graph and words from the corpus into the same vector space. The purpose of this model is to address several difficulties and challenges encountered in merging knowledge graph embedding models and relational fact extraction models, ultimately aiming to solve the problem of global knowledge graph completion for counter-terrorism data. The multi-task joint learning model can be decomposed into three parts: a knowledge model, a text model, and an alignment model. By defining the knowledge model and text model, a unified, compatible method for jointly learning entity embedding representations and word vectors is provided. By defining the alignment model, the difficulty of embedding entity embedding representations and word vectors into the same vector space is solved. Multi-task joint learning models can utilize the learned vector representations to calculate scores for candidate triples where entity links fail. Models within this framework simultaneously learn the embedding representations of entities in the knowledge graph and words in the text.

[0050] This application, by constructing a counter-terrorism knowledge graph, can effectively expand the means of data acquisition and collection, effectively compensate for the shortcomings of existing intelligence, surveillance, and information systems, and, relying on technologies such as data mining and correlation analysis, can maximize the extraction of valuable information from massive and fragmented data. By utilizing the relationships between things, it can quickly identify and analyze terrorist situations. The counter-terrorism data analysis system, based on data from the Maryland database, constructs a college-standard counter-terrorism database. Through the construction of a knowledge system and knowledge graph, it achieves the goals of teaching, learning, and scientific research, further broadening the scope of counter-terrorism work, improving efficiency, judgment, and early warning capabilities, while effectively meeting current application needs.

[0051] In some embodiments, the data source includes:

[0052] Structured data, semi-structured data, and unstructured data.

[0053] In some embodiments, the step of extracting the data source based on preset entity words includes:

[0054] For structured data, the D2R method is used to extract knowledge from the structured data or the graph mapping method is used to extract knowledge from linked data;

[0055] For semi-structured data, a wrapper is used to extract knowledge from the semi-structured data;

[0056] For unstructured data, information extraction methods are used to extract knowledge from free text.

[0057] It should be noted that the data sources associated with specific entity terms in this application include structured data, semi-structured data, and unstructured data. Correspondingly, knowledge extraction from the data sources associated with specific entity terms includes: extracting knowledge from the data sources associated with specific entity terms using knowledge extraction methods corresponding to the data source type; different data sources correspond to different knowledge extraction methods. If the data source type is structured data, the D2R method is used to extract knowledge from a relational database or the graph mapping method is used to extract knowledge from linked data; if the data source type is semi-structured data, a wrapper is used to extract knowledge from the semi-structured data; if the data source type is unstructured data, an information extraction method is used to extract knowledge from free text.

[0058] In some embodiments, resource description framework triples are extracted based on the global knowledge graph, including:

[0059] Calculate the Hamming distance between entity strings in the global database;

[0060] Based on the Hamming distance similarity, entity pairs that match between the anti-terrorism database and the custom database are selected, and an equivalent mapping between entities is constructed.

[0061] Based on the selected entity pairs, a breadth-first search is performed on the global anti-terrorism database to obtain entities related to the selected entity pairs.

[0062] Based on the obtained entities, extract resource description framework triples.

[0063] Specifically, a global database is established based on the identical and related entities between the Maryland Global Counterterrorism Database and the custom database. Strings in the global counterterrorism database are normalized. Hamming distances of entity strings in the global counterterrorism database are calculated. Based on the Hamming distance similarity, entity pairs matching between the Maryland Global Counterterrorism Database and the custom database are selected, and equivalent mappings between entities are constructed. Based on the selected entity pairs, a breadth-first search is performed on the global counterterrorism database to obtain entities related to the selected entity pairs. Based on the obtained entities, resource description framework triples are extracted.

[0064] The step of inputting the word vector space to be identified into the multi-task joint learning model to obtain a new resource description framework triplet includes:

[0065] The resource description framework triplet, mapped to the word vector space, is input into the multi-task joint learning model. The knowledge embedded in the word vector matrix space is then remapped to the resource description framework triplet of the text representation, thus obtaining a new resource description framework triplet.

[0066] In one implementation, the new resource description framework triple includes multiple elements, and after obtaining the new resource description framework triple, it also includes:

[0067] The newly generated resource description framework triples are organized and merged.

[0068] Understandably, the specific training process of the deep learning network involves training the model by embedding the resource description framework triples in the global knowledge graph of counterterrorism data, performing word mapping on the resource description framework triples in the global knowledge graph of counterterrorism data according to the embedding mapping rules, using the entities in the global knowledge graph of counterterrorism data as head entities, generating tail entities for all relationships that have appeared in the global knowledge graph of counterterrorism data, and determining the relationship between variables through the positional relationship in the vector space between the head entities and the tail entities.

[0069] Finally, the process of back-completing the global knowledge graph is as follows: the resource description framework triples mapped to the word vector space are fed into the trained multi-task joint learning model, and the knowledge embedded in the word vector matrix space is remapped to the resource description framework triples represented by the text, thus obtaining new resource description framework triples; the newly generated resource description framework triples are reorganized and merged, and then added to the incomplete global knowledge graph of counterterrorism data, resulting in the completed global knowledge graph of counterterrorism data.

[0070] like Figure 2 As shown in the figure, this application provides an apparatus for constructing and implementing a global knowledge graph of counter-terrorism data, including:

[0071] The extraction module 201 is used to extract the data source based on the data source in the anti-terrorism database and based on preset entity words to obtain triplets related to the entity words.

[0072] Module 202 is established by using the triples to fuse the anti-terrorism database and the custom database to establish a global database, using the global database to obtain a global knowledge graph, and extracting resource description framework triples based on the global knowledge graph.

[0073] Training module 203 is used to embed the resource description framework triples into the semantic space, establish a training set based on the semantic space, and use the training set to train the deep learning network to obtain a multi-task joint learning model; the multi-task joint learning model is used to identify resource description framework triples.

[0074] The completion module 204 is used to input the word vector space to be identified into the multi-task joint learning model to obtain a new resource description framework triplet, and to fill the new resource description framework triplet into the global knowledge graph.

[0075] The working principle of the device for constructing and implementing a global knowledge graph of counter-terrorism data provided in this application is as follows: Extraction module 201 uses data sources based on the counter-terrorism database to extract data sources based on preset entity words, obtaining triples related to the entity words; Establishment module 202 integrates the triples from the counter-terrorism database and a custom database to establish a global database, uses the global database to obtain a global knowledge graph, and extracts resource description framework triples based on the global knowledge graph; Training module 203 embeds the resource description framework triples into a semantic space, establishes a training set based on the semantic space, and uses the training set to train a deep learning network to obtain a multi-task joint learning model; The multi-task joint learning model is used to identify resource description framework triples; Completion module 204 inputs the word vector space to be identified into the multi-task joint learning model to obtain new resource description framework triples, and fills the new resource description framework triples into the global knowledge graph.

[0076] Based on the aforementioned global knowledge base, a counter-terrorism data analysis system is constructed. The system includes entry points, client terminals, and supports platforms such as UNIX, Linux, and Windows NT. It is primarily developed using the Java programming language and built upon a distributed computing and storage system. A distributed microservice architecture and related components are used to establish the core support system, ensuring that services corresponding to each business function can be independently compiled and deployed, communicate with each other through service interfaces, unconditionally support data extraction and push, and have no limit on the number of users or concurrent connections. The database content is mainly based on a global counter-terrorism data knowledge base. It includes functional modules for data acquisition, data editing, data management, user management, access control, visual interface management, and interface management.

[0077] The counter-terrorism data analysis system uses the HTTP protocol for access, and the corresponding port is configured on the UNIX server to wait for client connections.

[0078] It should be noted that the client operation steps are as follows:

[0079] The first step is the login page: Enter your account and password, click "Login," and log in. The backend homepage mainly displays basic server information and virtual machine information.

[0080] The second step is to select a module from the left-hand menu and then select the corresponding operation: such as the counterterrorism knowledge graph module.

[0081] Third step, click Attribute Management: Administrators can manage attribute tags and make corresponding modifications;

[0082] Fourth step, click Relationship Management: Administrators can modify the names of related relationships;

[0083] Step 5, click on the triplet: Administrators can add, modify, and query relevant data based on the start and end nodes;

[0084] Step 6: Click on the graph search: Administrators can retrieve the relationship graph and event summary based on the query event.

[0085] The seventh step involves analyzing different dimensions, such as country, criminal group, administrative division, target type, weapon type, attack type, and casualty situation, to form a corresponding counter-terrorism knowledge graph and generate corresponding reports for users to access.

[0086] The client-side module includes: basic data module, large screen management module, business data module, anti-terrorism knowledge graph module, report management module, system settings module, permission management module, and log management module.

[0087] The basic data module mainly includes: (1) Country: Administrators can query the Chinese and English names of various countries. (2) Criminal Groups: Administrators can add, delete, and modify information about criminal groups on this page. (3) Administrative Divisions: Administrators can add, delete, modify, and query information about administrative regions on this page. (4) Course Management: Administrators can add and query courses on this page. (5) Target Types: Administrators can add, delete, modify, and query basic target types. (6) Weapon Types: Administrators can add, delete, modify, and query basic weapon types on this page. (7) Attack Types: Administrators can add, modify, delete, and query basic attack types on this page. (8) Regions: Administrators can add, modify, delete, and query regions in Chinese / English. (9) Combat Types: Administrators can add, modify, delete, and query basic combat types on this page.

[0088] The large screen management module mainly includes: (1) Large screen backend management: Administrators can add screen modules, upload background images for each screen, modify colors, etc. (2) Tag management: Administrators can perform tag management for the entire system, including adding, modifying, deleting, and querying. (3) Data display: Administrators can display pie charts and line charts based on data types, and generate partial text reports based on the year.

[0089] The business data module mainly includes: (1) Internal Cases: Administrators can add, modify, delete, and query data related to internal cases. (2) Maryland Data: Administrators can add, modify, delete, and query data, and can export the required Excel reports.

[0090] The counter-terrorism knowledge graph module mainly includes: (1) Attribute management: Administrators can manage attribute tags and make corresponding modifications. (2) Relationship management: Administrators can modify the names of related relationships. (3) Triples: Administrators can add, modify, and query related data based on start and end nodes. (4) Graph retrieval: Administrators can query relationship graphs and event descriptions based on query events.

[0091] The report management module mainly includes: Reports: Administrators can upload PDF files, preview them based on the title and tags, and perform add, modify, and query operations.

[0092] The system settings module mainly includes: (1) Menu management: Administrators can adjust various modules according to this page; (2) Scheduled tasks: Administrators can synchronize knowledge graph data on a scheduled basis.

[0093] The access control module mainly includes: (1) User management: Administrators can add members who use this system. (2) Department management: Administrators can add departments that use this system. (3) Role management: Administrators can add users who log in to the backend and set permissions.

[0094] The log management module mainly includes: (1) Login log: Login and logout records of backend users; (2) Operation log: Records of which modules backend users viewed and the usage report exported to Excel; (3) Exception log: Exceptions encountered by backend users when using the system.

[0095] The counter-terrorism data analysis system provided in this application includes an entry point, client side, and platforms such as UNIX, Linux, and Windows NT. It is primarily developed using the JAVA programming language and built upon a distributed computing and storage system. A distributed microservice architecture and related components are used to establish the core support system, ensuring that services corresponding to each business can be independently compiled and deployed, and communicate with each other through service interfaces. Each system is fully open, unconditionally supporting data extraction and push, without limiting the number of users or concurrent connections. The database content is mainly based on a global counter-terrorism data knowledge base. Addressing different users and needs, it analyzes data from various dimensions such as country, criminal group, administrative division, target type, weapon type, attack type, and casualty situation to form corresponding counter-terrorism knowledge graphs and generate corresponding reports for user review. The client side includes: a basic data module, a large-screen management module, a business data module, a counter-terrorism knowledge graph module, a report management module, a system settings module, a permission management module, and a log management module.

[0096] In summary, this invention provides a method for constructing and implementing a global knowledge graph for counter-terrorism data. This application integrates and optimizes the Maryland Global Counter-Terrorism Database with a custom database to obtain a global database. A global knowledge graph is then derived from this global database, and resource description framework triples are extracted from the global knowledge graph. These resource description framework triples are embedded into a semantic space, and a deep learning network is trained using this semantic space to obtain a multi-task joint learning model. The word vector space to be identified is input into the multi-task joint learning model to obtain new resource description framework triples, which are then added to the global knowledge graph. This invention utilizes a global knowledge graph, significantly reducing the size of the knowledge base during training, testing, and querying, while greatly improving accuracy. It also automatically mines hidden knowledge within the global knowledge graph of counter-terrorism data, filling the gap in knowledge graph-based methods for completing global knowledge graphs for counter-terrorism data. Furthermore, the implementation of the counter-terrorism data analysis system can maximize and rapidly extract valuable information from massive amounts of fragmented data, quickly identifying and analyzing terrorist situations using the relationships between things, greatly satisfying current application needs.

[0097] It is understood that the method embodiments provided above correspond to the device embodiments described above, and the specific details can be referred to each other, which will not be repeated here.

[0098] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage and optical storage) containing computer-usable program code.

[0099] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0100] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction methods implemented in a process. Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0101] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0102] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for constructing and implementing a global knowledge graph for counter-terrorism data, characterized in that, include: Using a data source based on a counterterrorism database, the data source is extracted based on preset entity words to obtain triples related to the entity words; A global database is established by fusing the anti-terrorism database and a custom database using the triples, a global knowledge graph is obtained using the global database, and resource description framework triples are extracted based on the global knowledge graph. The resource description framework triples are embedded into the semantic space, and a training set is established based on the semantic space. The deep learning network is trained using the training set to obtain a multi-task joint learning model. The multi-task joint learning model is used to identify resource description framework triples. The word vector space to be identified is input into the multi-task joint learning model to obtain a new resource description framework triplet, which is then added to the global knowledge graph.

2. The method according to claim 1, characterized in that, The data source includes: Structured data, semi-structured data, and unstructured data.

3. The method according to claim 2, characterized in that, The step of extracting the data source based on preset entity words includes: For structured data, the D2R method is used to extract knowledge from the structured data or the graph mapping method is used to extract knowledge from linked data; For semi-structured data, a wrapper is used to extract knowledge from the semi-structured data; For unstructured data, information extraction methods are used to extract knowledge from free text.

4. The method according to claim 1, characterized in that, Based on the global knowledge graph, a resource description framework triple is extracted, including: Calculate the Hamming distance between entity strings in the global database; Based on the Hamming distance similarity, entity pairs that match between the anti-terrorism database and the custom database are selected, and an equivalent mapping between entities is constructed. Based on the selected entity pairs, a breadth-first search is performed on the global anti-terrorism database to obtain entities related to the selected entity pairs. Based on the obtained entities, extract resource description framework triples.

5. The method according to claim 4, characterized in that, The step of inputting the word vector space to be identified into the multi-task joint learning model to obtain a new resource description framework triplet includes: The resource description framework triplet, mapped to the word vector space, is input into the multi-task joint learning model. The knowledge embedded in the word vector matrix space is then remapped to the resource description framework triplet of the text representation, thus obtaining a new resource description framework triplet.

6. The method according to claim 5, characterized in that, The new resource description framework triples include multiple elements, and after obtaining the new resource description framework triples, they also include: The newly generated resource description framework triples are organized and merged.

7. The method according to claim 4, characterized in that, Also includes: Normalization processing of strings in the global anti-terrorism database.

8. An apparatus for constructing and implementing a global knowledge graph of counter-terrorism data, characterized in that, include: The extraction module is used to extract data from the data source based on the anti-terrorism database and based on preset entity words to obtain triples related to the entity words. A module is established by using the triples to fuse the anti-terrorism database and the custom database to establish a global database, using the global database to obtain a global knowledge graph, and extracting resource description framework triples based on the global knowledge graph. The training module is used to embed the resource description framework triples into the semantic space, establish a training set based on the semantic space, and train the deep learning network using the training set to obtain a multi-task joint learning model; the multi-task joint learning model is used to identify resource description framework triples. The completion module is used to input the word vector space to be identified into the multi-task joint learning model to obtain a new resource description framework triplet, and to fill the new resource description framework triplet into the global knowledge graph.

Images