Intelligence collaboration processing platform and potential relationship prediction method based on knowledge graph

Abstract

The application discloses an intelligence collaborative processing platform and a potential relation prediction method based on a knowledge graph, and relates to the technical field of intelligence processing. The application discloses an intelligence collaborative processing platform and a potential relation prediction method based on a knowledge graph, and relates to the technical field of intelligence processing. The data acquisition processing layer comprises a data acquisition module, a data format conversion module and a knowledge base management module. The knowledge extraction intelligent analysis layer comprises a keyword extraction module, a keyword hierarchical module, a semantic fusion module and a knowledge ontology dynamic evolution module. The knowledge application decision layer comprises an evolution state discrimination module, a military strategy graph generation module, an evaluation result acquisition module and a military strategy guidance information updating module. The application is oriented to real-time military strategies, integrates a large number of scattered and isolated intelligence, realizes the correlation of military intelligence information from the semantic level, provides strong support for the comprehensive analysis, research and display of intelligence information, improves analysis efficiency, and realizes the effective collaboration of military strategies and intelligence data processing.

Description

Technical Field

[0001] This invention relates to the field of military intelligence data processing technology, specifically to an intelligence collaborative processing platform based on knowledge graphs and a method for predicting potential relationships. Background Technology

[0002] Thanks to advancements in information technology, existing battlefield intelligence information is experiencing explosive growth, complex intelligence sources, and diverse representations. Traditional search engines and query methods are insufficient to capture and understand battlefield information and vast amounts of military knowledge.

[0003] The invention disclosed in CN115878812A presents a method for strategic intent analysis based on open-source intelligence. This method includes acquiring open-source intelligence text and preprocessing it; extracting multiple event elements and core words of the events; extracting event dependency subgraphs containing meta-event pairs; obtaining the labels corresponding to each meta-event in the meta-event pairs; constructing a preliminary event graph; obtaining an optimized event graph; and using a strategic event prediction model to predict subsequent events in the optimized event graph and obtain the causes of events in the optimized event graph. The method provided in this application mines information from unstructured text, integrates relevant data and resources, and expands the open-source intelligence text dataset. It extracts event elements and event relationships from the open-source intelligence text dataset, analyzes the enemy's strategic intent through event graph analysis, and realizes the transformation of traditional intelligence work towards informatization and intelligence, providing auxiliary decision-making capabilities for intelligence analysis.

[0004] The invention disclosed in publication number CN115878811A presents a method for intelligent analysis and deduction of military intelligence based on event-driven intelligence graphs. It proposes an event-centric approach to intelligence analysis and deduction, elaborating on the specific processes from automated construction of event-driven intelligence graphs to event-centric visual reasoning. This invention uses event-driven intelligence analysis technology to comprehensively monitor and visualize battlefield intelligence data, providing managers with a visible, manageable, and controllable military command and decision-making platform. This helps users uncover the value of military intelligence data and improve the efficiency of command and decision-making for combat personnel.

[0005] The invention disclosed in CN114896387A provides a method, apparatus, and computer-readable storage medium for military intelligence analysis visualization. After acquiring and preprocessing event texts from various fields, the method categorizes the event texts into event classes based on an event ontology library and extracts event elements. Based on these event elements, the method extracts the main content and event class relationships of the event texts. A Bayesian network classification model is then used to predict and deduce the event evolution path, and the event class relationships and evolution path are visualized. This allows the overall context of the intelligence event to be presented graphically, and a machine learning classification model can be used to predict the relevant impacts of the event and push customized domain-specific event content to relevant users.

[0006] However, the aforementioned invention requires the construction of a contextual map and optimization based on it to analyze the enemy's strategic intentions. It can only provide some information for military strategy as a reference. The information provided depends on the construction of the contextual map, ignores the real-time changes in military strategy, cannot achieve effective coordination between military strategy and intelligence data processing, and is difficult to uncover potential relationships in military intelligence.

[0007] The invention disclosed in CN115408532A presents a method for constructing a weapon and equipment knowledge graph based on open-source intelligence. This method involves acquiring military text data from open-source resources, preprocessing the military text data to obtain standardized military text data, and then labeling the standardized military text data to obtain a training set (a weapon and equipment entity recognition dataset, a weapon and equipment attribute extraction dataset, and a weapon and equipment entity link dataset). The model is improved through training, and based on the improved model, weapon and equipment are identified and attributes are extracted. Finally, a knowledge graph is constructed. This method solves the problems of low utilization rate, time-consuming and laborious querying, and inconvenience caused by the current scattered, inconsistent quality, and large volume of open-source military information. However, this invention can only be used to construct a weapon and equipment knowledge graph and cannot achieve effective collaboration between military strategy and intelligence data processing, nor does it address the potential relationships within military intelligence. Summary of the Invention

[0008] The purpose of this invention is to provide an intelligence collaborative processing platform and a potential relationship prediction method based on knowledge graphs. Starting from actual military operational data and guided by real-time military strategies, this invention studies military knowledge graphs for the field of military intelligence information, integrates a large amount of scattered and isolated intelligence, realizes the association of military intelligence information at the semantic level, provides strong support for the comprehensive analysis, judgment and display of intelligence information, improves analysis efficiency, realizes effective collaboration between military strategies and intelligence data processing, and effectively predicts potential relationships in military intelligence under the real-time guidance of military strategies.

[0009] To achieve the above-mentioned technical objectives, the technical solution adopted by the present invention is as follows:

[0010] A knowledge graph-based intelligence collaborative processing platform, comprising a data acquisition and processing layer, a knowledge extraction and intelligent analysis layer, and a knowledge application decision-making layer;

[0011] The data acquisition and processing layer includes a data acquisition module, a data format conversion module, and a knowledge base management module;

[0012] The data acquisition module obtains military intelligence business data from multiple data sources, cleans and classifies it, and stores it in the corresponding database. This military intelligence business data includes various military intelligence results analyzed and researched in recent years, as well as intelligence information reported in real-time by various levels of departments. The data format conversion module is used to uniformly represent the multi-source, heterogeneous military intelligence business data collected by the data acquisition module. The knowledge base management module generates a semantic network knowledge base based on the format-converted military intelligence business data, dynamically expands the semantic network knowledge base, and integrates duplicate and conflicting content in the knowledge ontology construction. The semantic network knowledge base includes four types of domain knowledge ontology: people, organizations, resources, and military scenario events. Each domain knowledge entity has corresponding attribute information, and each domain knowledge entity includes an active status attribute.

[0013] The knowledge extraction intelligent analysis layer includes a keyword extraction module, a keyword layering module, a semantic fusion module, and a knowledge ontology dynamic evolution module.

[0014] The keyword extraction module is used to obtain several words or phrases that are highly relevant to the entity as basic keywords;

[0015] The keyword layering module constructs a core keyword set for the current period based on the evolution trend of historical military strategy-related information and the military strategy guidance information of the current period; then, according to the dependency relationship of syntactic analysis and the co-occurrence frequency of keywords, it layers and classifies the remaining basic keywords to construct a spatial distribution model between keywords and entities, and between keywords themselves.

[0016] The semantic fusion module takes the relationship between keywords and knowledge ontology of various domains as the main feature, establishes feature vectors of knowledge ontology of various domains, calculates the similarity between each pair of knowledge ontology of various domains, and obtains the relationship between knowledge ontology of various domains based on a general threshold; starting from keyword frequency and distance features, it mines the hierarchical and sequential relationship between knowledge entities of various domains, constructs a knowledge ontology evolution semantic network, and performs semantic fusion on multi-source heterogeneous data in the semantic network knowledge base.

[0017] The knowledge ontology dynamic evolution module starts from the dynamic evolution of knowledge ontology itself and between each domain, performs upper and lower layer semantic reasoning on the fused semantic network knowledge base, finds the link between two knowledge ontology and extends the link, discovers the potential semantics or broader relationships on the link, and mines the evolution trend of multiple relationships between people, organizations, resource behavior sequences and military scenario events, and constructs a multi-dimensional knowledge semantic network; it introduces the time slice method to cumulatively expand the network topology decoupling, and sets position labels for each node and the edge of the link in each time slice to indicate whether there is a state transition of network nodes and edges in the corresponding time slice, discovers and reasons about knowledge groups, and updates the active state attributes of each domain knowledge entity;

[0018] The knowledge application decision layer includes an evolution state discrimination module, a military strategy map generation module, an evaluation result collection module, and a military strategy guidance information update module.

[0019] The evolution status discrimination module, based on the discovery and reasoning results of the knowledge group, discerns the evolution status and trend of military strategy-related information from several aspects, including individuals, organizations, resources, and military scenario events. Simultaneously, it updates the active status attributes for each domain knowledge entity. The military strategy map generation module, based on the updated active status attributes, selects several key individuals, organizations, resources, and events of interest for the current period and generates a military strategy map. The evaluation result collection module collects the evaluation results of military experts on military strategy-related information. The military strategy guidance information update module dynamically updates the military strategy guidance information for the next period based on the evolution status and evaluation results of the military strategy-related information.

[0020] Furthermore, the basic structure of the data in the semantic network knowledge base is: SKB = (C, E, A, V), where C = {c1, c2, ..., c...} n} represents the category set, E = {e1, e2, ..., e m Let A = {a1, a2, ..., a} represent a set of entities. k} represents the set of entity attributes, where a i ∈A represents a basic attribute. This represents the value corresponding to the attribute; in the category set {C}, there is a parent-child relationship between categories, i.e., c j ∈c S , indicating category c j It is category c S Subclasses of.

[0021] Furthermore, the knowledge base management module includes an entity construction component, a semantic network knowledge base generation component, and a conflict fusion component;

[0022] The entity construction component is used to study the characteristics of some unstructured text, manually construct an initial military domain knowledge entity attribute library, and then expand the military domain knowledge entity attribute library based on word similarity. At the same time, the unstructured text is labeled with entities and attributes using a dictionary method combined with part-of-speech attribute discrimination.

[0023] The semantic network knowledge base generation component uses keyword spatial location co-occurrence features as a training set, employs a deep learning model to train the labeled entity and attribute data, uses the trained deep learning model to extract new domain knowledge data, evaluates the effectiveness of the extracted entities and attributes, dynamically expands the evaluated effective entities and attributes into the military domain knowledge entity attribute database, and then identifies domain knowledge entities and attributes in massive texts based on reinforcement machine learning methods.

[0024] The conflict fusion component analyzes terminological conflicts, semantic conflicts, and predicate conflicts, and uses a hybrid approach of logic tree fusion, frequency fusion, and syntactic fusion to select and fuse repetitive and conflicting content in the knowledge ontology construction.

[0025] Furthermore, the conflict fusion component includes a terminology database, a predicate database, an ontology database, a terminology conflict processing unit, a semantic conflict processing unit, a predicate conflict processing unit, and a fusion unit;

[0026] The terminology database and predicate database extract relevant knowledge ontology from the military domain knowledge entity attribute database, respectively, and construct terminology sets, predicate sets, and semantic sets, which are then transmitted to the ontology database. The ontology database analyzes the terminology sets, predicate sets, and semantic sets to identify existing terminology conflicts, predicate conflicts, and semantic conflicts. The terminology conflict processing unit calls a logic tree fusion model, a frequency fusion model, and a syntactic fusion model to process terminology conflicts; the semantic conflict processing unit calls a syntactic fusion model to process semantic conflicts; and the predicate conflict processing unit calls a frequency fusion model and a syntactic fusion model to process predicates. The fusion unit integrates the processing results of the terminology conflict processing unit, the semantic conflict processing unit, and the predicate conflict processing unit to generate fused knowledge items.

[0027] The logic tree fusion model fuses conflicting items by determining the logic tree relationship between two conflicting items; the frequency fusion model uses high-frequency co-occurrence knowledge items as the highest score output; the syntactic fusion model uses syntactic dependency tree discriminant to determine the semantic space dependency relationship between entities, attribute words, and attribute value words, and selects and discards multiple conflicting items.

[0028] Furthermore, the keyword extraction module obtains a word set D-terms by segmenting the document set D of each domain knowledge entity content, maps the document set D to a graph G, where vertices are words and edges represent the co-occurrence frequency of two words, calculates the network node centrality between nodes, sorts the network node centrality of each word, merges some words into phrases using an n-gram algorithm, and outputs the top-ranked results as basic keywords related to the domain knowledge entity.

[0029] Furthermore, the keyword layering module filters out entities that are active in the current period based on the evolution trend of historical military strategy-related information. Then, for each entity and keyword association, it selects the first keyword related to the entity from the basic keywords. It acquires and analyzes the military strategy guidance information of the current period, and selects the second keyword related to the entity corresponding to the military strategy guidance information from the basic keywords. The first keyword and the second keyword are combined to form the core keyword set of the current period.

[0030] Furthermore, the semantic fusion module employs an entity similarity calculation model and a category network fusion model to fuse the semantic knowledge network;

[0031] The entity similarity calculation model uses the following formula to fuse entities between the semantic network knowledge bases SKB1 and SKB2:

[0032] f:(C i E i A i V i )×(C j E j A j V j → (C, E, A, V)

[0033] Specifically, if the entity names in two knowledge bases are the same, then the two knowledge bases are considered to describe the same content; if the entity name in SKB1 is e i No matching entity name was found in SKB2. j However, it finds e that has the same domain knowledge description or a similarity greater than the first preset similarity threshold. j If the entity name in SKB1 is the same as the entity name in the other knowledge base, then the entities in the two knowledge bases are considered to be the same; if the entity name in SKB1 is the same as the entity name in the other knowledge base, then the entities in the two knowledge bases are considered to be the same. i In SKB2, only pairs of characters with a similarity less than the first preset similarity threshold can be found. j However, if the similarity of other attributes and attribute values ​​is greater than the second preset similarity threshold, then the entities in the two knowledge bases are considered to be the same; when two entities are determined to be the same, let e = e i Or e = e j C = (Ci ∪C j ), A = (A i ∪A j ),in

[0034] When two entities are judged to be identical or two semantic network category nodes are merged, the category network fusion model is used to execute C = (C i ∪C j When an entity has only one parent class, determine whether the parent classes of the two knowledge bases are the same:

[0035] If c i With c j If the terms used in the category are the same, then the parent class is considered to be the same.

[0036] If c i With c j If the category terms are inconsistent, but the domain knowledge descriptions are identical or the similarity exceeds a third preset threshold, then the categories in the two knowledge bases are considered identical. If the category terms and domain content descriptions are inconsistent, then the similarity of other attributes and attribute values ​​of the category exceeds a fourth preset similarity threshold, and the categories in the two knowledge bases are considered identical. When an entity has multiple parent classes in different knowledge bases, and... Then C i ∩C j The common parent class of the entities is used, while other parent classes invoke the deep inheritance algorithm and the fusion algorithm.

[0037] Furthermore, the knowledge ontology dynamic evolution module includes a unit knowledge reasoning component, a multi-relationship reasoning component, and a dynamic knowledge reasoning component;

[0038] The unit knowledge reasoning component loads knowledge network data and establishes a data structure for each node's neighbor nodes and label information. For each node in the knowledge network, based on the label attributes of the node and its neighbor nodes, and the specified adjustable parameter p for the label information ratio, it calculates the probability value of the neighbor nodes of the node being visited, and randomly selects several neighbor nodes from the node's neighbors, with the probability of each neighbor node being selected matching the calculated probability value. Then, based on the obtained probability value and the visiting parameter, it initiates a visit to obtain several visiting paths. Based on the visiting paths, it calls the Word2vec method for training to obtain word vectors, and performs a multi-label classification task on the knowledge network nodes.

[0039] The multi-relation reasoning component constructs possible candidate LE patterns based on the combination of all edge types on the knowledge network; it traverses the entire graph to find all instances corresponding to different candidate LE patterns, calculates the weights and values ​​of different candidate LE patterns, solves for reliable LE patterns using a greedy algorithm, and uses the selected reliable LE patterns to perform matching in the knowledge network to obtain the reasoning results.

[0040] The dynamic knowledge reasoning component introduces a time-slice approach to cumulatively extend the decoupling of network topology. By setting a position label for each node and the linked edge in each time slice, it indicates whether there is a state transition of network nodes and edges in the corresponding time slice, thereby discovering and reasoning about knowledge groups.

[0041] The present invention also discloses a potential relationship prediction method based on knowledge graphs, wherein the potential relationship prediction method is executed based on the knowledge graph-based intelligence collaborative processing platform described above;

[0042] The potential relationship prediction method includes the following steps:

[0043] Military intelligence business data is obtained from multiple data sources, cleaned and classified, and then stored in the corresponding databases. The military intelligence business data includes various military intelligence results data analyzed and researched in recent years and intelligence information reported by departments at all levels in real time.

[0044] The system unifies the representation of multi-source heterogeneous military intelligence business data collected by the data acquisition module; it generates a semantic network knowledge base based on the format-converted military intelligence business data, dynamically expands the semantic network knowledge base, and integrates duplicate and conflicting content in the knowledge ontology construction; the semantic network knowledge base includes four types of domain knowledge ontology: people, organizations, resources, and military scenario events, each domain knowledge entity has corresponding attribute information, and each domain knowledge entity includes an active status attribute;

[0045] Several words or phrases that are highly relevant to the entity are selected as basic keywords;

[0046] Based on the evolution trend of historical military strategy information and the military strategy guidance information of the current period, a core keyword set for the current period is constructed; then, according to the dependency relationship of syntactic analysis and the co-occurrence frequency of keywords, the remaining basic keywords are hierarchically and classified to construct a spatial distribution model between keywords and entities, and between keywords themselves.

[0047] Using the association between keywords and knowledge ontology in various domains as the main feature, feature vectors of knowledge ontology in various domains are established, and the similarity between each pair of knowledge ontology in various domains is calculated. The relationship between knowledge ontology in various domains is obtained according to a general threshold. Starting from keyword frequency and distance features, the hierarchical and sequential relationships between knowledge entities in various domains are mined, a knowledge ontology evolution semantic network is constructed, and semantic fusion is performed on multi-source heterogeneous data in the semantic network knowledge base.

[0048] Starting from the dynamic evolution of knowledge ontology itself and between different domains, semantic reasoning is performed on the integrated semantic network knowledge base at both upper and lower layers. After finding the links between two knowledge ontology, the links are extended to discover potential semantics or broader relationships on the links. The evolutionary trends of multiple relationships between people, organizations, resource behavior sequences, and military scenario events are mined to construct a multi-dimensional knowledge semantic network. The network topology is decoupled and expanded cumulatively by introducing a time-slice approach. Each node and the edge of the link are labeled with a position label in each time slice to indicate whether there is a state transition of network nodes and edges in the corresponding time slice. Knowledge groups are discovered and reasoned, and the active state attributes of knowledge entities in each domain are updated.

[0049] Based on the findings and reasoning of the knowledge community, the evolution status and trends of military strategy-related information are determined from several aspects, including people, organizations, resources, and military scenario events. At the same time, the active status attributes of each knowledge entity are updated. Based on the updated active status attributes, several key people, organizations, resources, and events of the current period are selected to generate a military strategy map.

[0050] Collect assessment results from military experts on military strategy-related information; dynamically update military strategy guidance information for the next cycle based on the evolution of military strategy-related information and assessment results.

[0051] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0052] First, the knowledge graph-based intelligence collaborative processing platform and potential relationship prediction method of this invention start from actual military business data, are guided by real-time military strategies, study military knowledge graphs for the field of military intelligence information, integrate a large amount of scattered and isolated intelligence, realize the association of military intelligence information at the semantic level, provide strong support for the comprehensive analysis, judgment and display of intelligence information, improve analysis efficiency, realize effective collaboration between military strategy and intelligence data processing, and effectively predict potential relationships in military intelligence under the real-time guidance of military strategy.

[0053] Secondly, the knowledge graph-based intelligence collaborative processing platform and potential relationship prediction method of the present invention can realize the correlation analysis and aggregation collaboration of military intelligence data oriented towards military strategy, with military strategy as the center, and provide effective basis for military personnel to make strategy adjustments through the visualization processing of military strategy-related information. Attached Figure Description

[0054] Figure 1 This is a schematic diagram of the knowledge graph-based intelligence collaborative processing platform structure of the present invention;

[0055] Figure 2 A schematic diagram illustrating the principles of entity recognition and attribute extraction;

[0056] Figure 3 Diagram of conflict resolution principle;

[0057] Figure 4 This is a schematic diagram of the spatial distribution model structure between keywords and entities, and between keywords themselves. Detailed Implementation

[0058] The embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.

[0059] See Figure 1 This embodiment discloses an intelligence collaborative processing platform based on knowledge graphs. The intelligence collaborative processing platform includes a data acquisition and processing layer, a knowledge extraction and intelligent analysis layer, and a knowledge application decision-making layer.

[0060] The data acquisition and processing layer includes a data acquisition module, a data format conversion module, and a knowledge base management module;

[0061] The data acquisition module obtains military intelligence business data from multiple data sources, cleans and classifies it, and stores it in the corresponding database. This military intelligence business data includes various military intelligence results analyzed and researched in recent years, as well as intelligence information reported in real-time by various levels of departments. The data format conversion module is used to uniformly represent the multi-source, heterogeneous military intelligence business data collected by the data acquisition module. The knowledge base management module generates a semantic network knowledge base based on the format-converted military intelligence business data, dynamically expands the semantic network knowledge base, and integrates duplicate and conflicting content in the knowledge ontology construction. The semantic network knowledge base includes four types of domain knowledge ontology: people, organizations, resources, and military scenario events. Each domain knowledge entity has corresponding attribute information, and each domain knowledge entity includes an active status attribute.

[0062] The knowledge extraction intelligent analysis layer includes a keyword extraction module, a keyword layering module, a semantic fusion module, and a knowledge ontology dynamic evolution module.

[0063] The keyword extraction module is used to obtain several words or phrases that are highly relevant to the entity as basic keywords;

[0064] The keyword layering module constructs a core keyword set for the current period based on the evolution trend of historical military strategy-related information and the military strategy guidance information of the current period; then, according to the dependency relationship of syntactic analysis and the co-occurrence frequency of keywords, it layers and classifies the remaining basic keywords to construct a spatial distribution model between keywords and entities, and between keywords themselves.

[0065] The semantic fusion module takes the relationship between keywords and knowledge ontology of various domains as the main feature, establishes feature vectors of knowledge ontology of various domains, calculates the similarity between each pair of knowledge ontology of various domains, and obtains the relationship between knowledge ontology of various domains based on a general threshold; starting from keyword frequency and distance features, it mines the hierarchical and sequential relationship between knowledge entities of various domains, constructs a knowledge ontology evolution semantic network, and performs semantic fusion on multi-source heterogeneous data in the semantic network knowledge base.

[0066] The knowledge ontology dynamic evolution module starts from the dynamic evolution of knowledge ontology itself and between each domain, performs upper and lower layer semantic reasoning on the fused semantic network knowledge base, finds the link between two knowledge ontology and extends the link, discovers the potential semantics or broader relationships on the link, and mines the evolution trend of multiple relationships between people, organizations, resource behavior sequences and military scenario events, and constructs a multi-dimensional knowledge semantic network; it introduces the time slice method to cumulatively expand the network topology decoupling, and sets position labels for each node and the edge of the link in each time slice to indicate whether there is a state transition of network nodes and edges in the corresponding time slice, discovers and reasons about knowledge groups, and updates the active state attributes of each domain knowledge entity;

[0067] The knowledge application decision layer includes an evolution state discrimination module, a military strategy map generation module, an evaluation result collection module, and a military strategy guidance information update module.

[0068] The evolution status discrimination module, based on the discovery and reasoning results of the knowledge group, discerns the evolution status and trend of military strategy-related information from several aspects, including individuals, organizations, resources, and military scenario events. Simultaneously, it updates the active status attributes for each domain knowledge entity. The military strategy map generation module, based on the updated active status attributes, selects several key individuals, organizations, resources, and events of interest for the current period and generates a military strategy map. The evaluation result collection module collects the evaluation results of military experts on military strategy-related information. The military strategy guidance information update module dynamically updates the military strategy guidance information for the next period based on the evolution status and evaluation results of the military strategy-related information.

[0069] I. Data Acquisition and Processing Layer

[0070] The data acquisition and processing layer is used to collect, preprocess, and perform basic data analysis on multi-source heterogeneous data. The basic data sources mainly include massive amounts of real-time access to military intelligence business data, various intelligence results analyzed and researched in recent years, and intelligence information reported by troops at all levels. Based on this, the layer studies the fundamental theories of domain-oriented knowledge ontology models, constructs various relationships and their definitions, and organizes, manages, and stores various data resources around the domain-oriented knowledge ontology model. The data acquisition and processing layer includes a data acquisition module, a data format conversion module, and a knowledge base management module.

[0071] To enable effective data collection, this invention provides functions for building and displaying a basic database. The basic database supports batch entry of existing basic data, integration of basic data and trend intelligence data on the same platform, uploading, storage, and classification management of basic data, display and viewing of basic data messages, and full-text search of basic database data, etc.

[0072] To address the inconsistency in knowledge description during the fusion of multi-source semantic network knowledge bases, this invention employs a unified multi-source semantic knowledge base fusion framework. The fused semantic network knowledge base primarily consists of four parts: knowledge ontology / entity, category, attribute, and attribute value. An entity, as a leaf node in the entire semantic network, is an instance. The category hierarchy comprises categories, sub-categories (parent categories), and entities under each category. An entity or category can belong to multiple categories or parent categories, as shown in Table 1.

[0073] Table 1. Schematic diagram of the time-series and spatial content models of knowledge ontology.

[0074]

[0075]

[0076] An entity is the smallest unit of various knowledge bases. It is a name for a certain thing and can be classified into categories. It is a specific instance of a category.

[0077] The semantic network knowledge base also provides a simple documentation explanation for the category, including the parent class, subclasses, and entities of the category.

[0078] Attributes are used to describe entities and categories. For each entity or category, the attribute is a term, category, or domain knowledge description obtained from the semantic network knowledge base.

[0079] An attribute value is a detailed description of an attribute in an entity.

[0080] Based on the above four-tuples, this project presents the basic structure of a semantic knowledge base (SKB):

[0081] SKB = (C, E, A, V)

[0082] Where C = {c1, c2, ..., c} n} represents the category set, E = {e1, e2, ..., e m Let A = {a1, a2, ..., a} represent a set of entities. k} represents the set of entity attributes, where a i ∈A represents a basic attribute. This represents the value corresponding to the attribute. In the category set {C}, there are parent-child relationships between categories, i.e., c j ∈c S , indicating category c j It is category c S Subclasses of (Chile-Category).

[0083] Entity recognition classifies entities in text into a specific semantic type. This invention primarily employs three methods for entity recognition: dictionary-based, statistical, and rule-based methods. Dictionary-based methods mainly search for named entities in a lexicon through string matching. Rule-based methods incorporate lexical, syntactic, and semantic rules into the entity recognition process, then identify various types of named entities through rule matching. Statistical methods utilize manually labeled or raw corpora for training, first establishing a language model, then estimating model parameters on training data, which is beneficial for portability to different languages ​​and new domains. Statistical methods primarily utilize statistical models such as Hidden Markov Models, Maximum Entropy Models, Support Vector Machines, Conditional Random Fields (CRFs), and deep learning. Attribute extraction involves constructing an attribute table for each entity's semantic class and extracting attribute values. The main method for attribute extraction is pattern matching. See the diagrams for the principles of entity recognition and attribute extraction. Figure 2 In this invention, the semantic network knowledge base includes four domain knowledge ontology categories: people, organizations, resources, and military scenario events. Each domain category has subcategories, and in practical applications, further subdivision of these subcategories is required. These four domain knowledge ontology categories can be used to display four types of information in the final military strategy map. The domains are also closely related, and the selection of core keywords is also of positive significance.

[0084] The knowledge base management module includes an entity construction component, a semantic network knowledge base generation component, and a conflict fusion component;

[0085] The entity construction component is used to study the characteristics of some unstructured texts. It manually builds an initial military domain knowledge entity attribute database, and then uses the deep learning LSTM sequence mining algorithm to establish word similarity to expand the military domain knowledge entity attribute database. At the same time, it uses a dictionary method combined with part-of-speech attribute discrimination to annotate entities and attributes for the unstructured texts that need to be processed.

[0086] In terms of entity and attribute feature extraction, the semantic network knowledge base generation component uses keyword spatial location co-occurrence features as the training set, employs a deep learning model to train on the labeled entity and attribute data, selects effective parameter models to extract new domain knowledge data, evaluates the effectiveness of the extracted entities and attributes, and dynamically expands the evaluated entities and attributes into the military domain knowledge entity and attribute database. Based on this, the corpus is continuously added, and the domain knowledge entity and attribute database is dynamically expanded. Finally, reinforcement learning-based methods are used to identify domain knowledge entities and attributes in massive amounts of text.

[0087] During the construction of entity attribute sets, conflicts may arise in the credibility and expression of various knowledge item corpora. The conflict fusion component analyzes terminological conflicts, semantic conflicts, and predicate conflicts, employing a hybrid approach of logic tree fusion, frequency fusion, and syntactic fusion to select and fuse duplicate and conflicting content in the knowledge ontology construction. For example... Figure 3 As shown, the conflict fusion component includes a terminology database, a predicate database, an ontology database, a terminology conflict processing unit, a semantic conflict processing unit, a predicate conflict processing unit, and a fusion unit.

[0088] The terminology database and predicate database extract relevant knowledge ontology from the military domain knowledge entity attribute database, respectively construct terminology sets, predicate sets, and semantic sets, and transmit them to the ontology database. The ontology database analyzes the terminology sets, predicate sets, and semantic sets to identify existing terminology conflicts, predicate conflicts, and semantic conflicts. The terminology conflict processing unit calls the logic tree fusion model, frequency fusion model, and syntactic fusion model to process terminology conflicts; the semantic conflict processing unit calls the syntactic fusion model to process semantic conflicts; and the predicate conflict processing unit calls the frequency fusion model and syntactic fusion model to process predicates. The fusion unit integrates the processing results of the terminology conflict processing unit, semantic conflict processing unit, and predicate conflict processing unit to generate fused knowledge items. The logic tree fusion model fuses conflicting items by determining the logic tree relationship between two conflicting items. For example, for the entity Entity0 extracted from knowledge items, two values, value1 and value2, are extracted. Without considering the time state transition between value1 and value2, there is a terminology logic conflict between value1 and value2. This project first determines the logic tree relationship between value1 and value2. If value1 contains value2, then value1 is treated as a second-level attribute attribute12 of attribute1 of entity0; otherwise, the same treatment is performed. The frequency fusion model uses high-frequency co-occurrence knowledge items as the highest score output; the syntactic fusion model uses syntactic dependency tree discriminant to determine the semantic space dependency relationship between entities, attribute words, and attribute value words, and selects and discards multiple conflicting items to achieve knowledge fusion and authentication.

[0089] II. Knowledge Extraction Intelligent Analysis Layer

[0090] The knowledge extraction and intelligent analysis layer provides the system with goal-centered, multi-modal data processing methods and capabilities. Combined with the system's data storage model, it offers frameworks for offline, real-time, and graph computing to meet various data analysis application needs. It implements deep learning-based knowledge ontology and event extraction, network structure evolution, and event correlation analysis services, constructing a domain knowledge ontology library and multi-domain knowledge ontology network relationships. It also enables behavioral pattern discovery of domain knowledge ontology, builds relationships between multi-domain knowledge ontology, and facilitates reasoning applications during the dynamic evolution of domain knowledge ontology. The knowledge extraction and intelligent analysis layer includes a keyword extraction module, a keyword layering module, a semantic fusion module, and a knowledge ontology dynamic evolution module.

[0091] (2.1) Keyword Extraction Module

[0092] The keyword extraction module is primarily used to obtain several words or phrases that are highly relevant to the entity. First, the document set D containing the content of each domain knowledge entity is segmented to obtain a word set D-terms. Document D is then mapped to a graph G, where vertices are words and edges represent the co-occurrence frequency of two words. Next, the BC (Breakpoint Centrality) values ​​between nodes are calculated. Then, the BC values ​​of each word are sorted, and some words are merged into phrases using an n-gram algorithm. Finally, the top-ranked results are output as keywords relevant to the domain knowledge entity.

[0093] (2.2) Keyword Hierarchy Module

[0094] To further illustrate the hierarchical relationships of the extracted keywords, based on the patterns of keyword occurrence within the military domain knowledge entity document set, this invention categorizes keywords into core keywords, important keywords, and general keywords. Then, it constructs the relationships between domain knowledge entities and keywords, resulting in a spatial distribution model between keywords and entities, and between keywords themselves, as shown below. Figure 4 As shown.

[0095] In this invention, the guidance of military strategy on intelligence data aggregation is mainly achieved through the selection of core keywords. Specifically, the keyword layering module selects a first keyword from basic keywords based on the evolution trend of historical military strategy-related information and the correlation between keywords; it acquires and analyzes military strategy guidance information for the current period, and selects a second keyword from basic keywords based on this information; finally, it combines the first and second keywords to form the core keyword set for the current period. It should be noted that the number of first and second keywords selected here is usually multiple.

[0096] The primary keyword, as a common keyword, is significantly influenced by the evolutionary trends and real-time status of historical military strategy-related information, and is less affected by the subjective biases of military experts, thus possessing higher objectivity. To ensure that military personnel receive the most complete intelligence information possible, after identifying the primary keyword, an LSTM deep sequence neural network can be used to train a broader set of core keywords. The secondary keyword, as a characteristic keyword, is more relevant to the current period's military strategy guidance information, enabling subsequent knowledge reasoning processes to be more targeted at real-time military strategy changes and achieving a synergistic relationship between military strategy and data processing. As a preferred example, the selection of the primary keyword can be achieved by setting keyword selection rules or introducing machine learning algorithms. First, the evolutionary trends of historical military strategy-related information are analyzed, and important figures, organizations, resources, or military scenarios are screened. Then, the correlation between the screened important figures, organizations, resources, or military scenarios and the keywords is used to select the primary keyword. As for the selection of the secondary keyword, data on military personnel's viewing preferences for military strategy-related information and their evaluation results can be collected. By analyzing keywords related to these preferences or evaluation results and adjusting priorities, some secondary keywords can be selected or eliminated. Therefore, the core keywords obtained by combining the first and second keywords are not only objective and comprehensive, but also real-time and targeted under the guidance of strategy, so that the subsequent knowledge reasoning process can also be smoothly coordinated with the changing process of military strategy.

[0097] Based on the determination of core keywords, the keywords extracted by the keyword extraction module are divided into two categories related to domain knowledge: important and general keywords, based on the dependency relationship of syntactic analysis and the co-occurrence frequency of keywords. A spatial distribution model between keywords and entities, and between keywords themselves, is then constructed.

[0098] (2.3) Semantic Fusion Module

[0099] The semantic fusion module includes a domain knowledge entity association extraction component, a causal relationship discrimination component, and a semantic fusion component for multi-source heterogeneous data.

[0100] The domain knowledge entity association extraction component uses the association between keywords and domain knowledge entities as the main feature, establishes feature vectors for domain knowledge entities, calculates the pairwise similarity between domain knowledge entities, and obtains the relationships between domain knowledge entities based on a general threshold. Considering the sparsity of domain knowledge entities, a hierarchical training strategy, HAM (hierarchical abstract machines), is adopted to obtain models with different levels of accuracy. The top-level model is used to remove a large amount of irrelevant data, while the bottom-level models use a multi-classification method to accurately identify the associations between domain knowledge entities.

[0101] The prior causality discriminant component is used to determine the prior causal relationships in a domain knowledge entity network. Specifically, determining the prior relationships between domain knowledge entities that have already established connections can be treated as a binary classification problem, i.e., it is transformed into a feature selection and classification algorithm selection problem. The prior causality discriminant component mainly starts from keyword frequency and distance features. Considering that there may be correlations between the above feature vectors, for example, SVM can be used to identify the prior relationships between domain knowledge entities.

[0102] Given the definitions of entities and categories, the semantic fusion components of multi-source heterogeneous data, namely the entity similarity calculation model and the category network fusion model, fuse multiple semantic knowledge networks.

[0103] The entity similarity calculation model uses the following formula to fuse entities between the semantic network knowledge bases SKB1 and SKB2:

[0104] f:(C i E i A i V i )×(C j E j A j V j → (C, E, A, V)

[0105] Specifically, if the entity names in two knowledge bases are the same, then the two knowledge bases are considered to describe the same content; if the entity name in SKB1 is e i No matching entity name was found in SKB2. j However, it finds e that has the same domain knowledge description or a similarity greater than the first preset similarity threshold. j If the entity name in SKB1 is the same as the entity name in the other knowledge base, then the entities in the two knowledge bases are considered to be the same; if the entity name in SKB1 is the same as the entity name in the other knowledge base, then the entities in the two knowledge bases are considered to be the same. i In SKB2, only pairs of characters with a similarity less than the first preset similarity threshold can be found. jHowever, if the similarity of other attributes and attribute values ​​is greater than the second preset similarity threshold, then the entities in the two knowledge bases are considered to be the same; when two entities are determined to be the same, let e = e i Or e = e j C = (C i ∪C j ), A = (A i ∪A j ),in

[0106] When two entities are judged to be identical or two semantic network category nodes are merged, the category network fusion model is used to execute C = (C i ∪C j When an entity has only one parent class, determine whether the parent classes of the two knowledge bases are the same:

[0107] If c i With c j If the terms used in the category are the same, then the parent class is considered to be the same.

[0108] If c i With c j If the category terms are inconsistent, but the domain knowledge descriptions are identical or the similarity exceeds a third preset threshold, then the categories in the two knowledge bases are considered identical. If the category terms and domain content descriptions are inconsistent, then the similarity of other attributes and attribute values ​​of the category exceeds a fourth preset similarity threshold, and the categories in the two knowledge bases are considered identical. When an entity has multiple parent classes in different knowledge bases, and... Then C i ∩C j The common parent class of the entities is used, while other parent classes invoke the deep inheritance algorithm and the fusion algorithm.

[0109] (2.4) Knowledge Ontology Dynamic Evolution Module

[0110] Semantic network knowledge reasoning techniques, starting from the self-generation and dynamic evolution of knowledge ontology, can be divided into reasoning of semantics between upper and lower layers of unit knowledge nodes, reasoning of relationships between multiple knowledge nodes, and reasoning of knowledge changes in the dynamic evolution of the entire network.

[0111] In terms of unit knowledge reasoning, this invention improves upon existing random walk algorithms to construct a semantic network reasoning algorithm for military intelligence content. Assume the knowledge network data to be processed contains N nodes, and the current node in the walk is C. The next node to be selected from C's neighbors is E, representing this as follows:

[0112] neighbors(C)={n1,n2,n3,…,n E}, 0≤E <N

[0113] At the same time, the neighbor nodes in neighbors(C) that have the same label as node C are represented as:

[0114] common(C) = {m1,m2,m3,…,m} k}, 0≤k≤E

[0115] Clearly, common(C) is a subset of neighbors(C). Let node D be selected as the next node after node C and be traversed, where D belongs to the set neighbors(C). The probability that node D belongs to the set common(C) must satisfy the following condition.

[0116] P(D∈common(C))=p,D∈neighbors(C)

[0117] Where probability p is the walking parameter set before the node walk begins. Calculate the new variables for node C:

[0118]

[0119] Assign a probability of being visited to each of node C's neighbors. Pass this set of probability values ​​to AliasMethod. The overall calculation process for unit knowledge reasoning is given below:

[0120] Step 1: Load the knowledge network data and establish a data structure for the neighbor nodes and label information corresponding to each node;

[0121] Step 2: For each node in the knowledge network, based on the label attributes of the node and its neighboring nodes, and the adjustable parameter p of the specified label information ratio, calculate the probability value of the neighboring nodes of the node being visited, and use AliasMethod to randomly select several neighboring nodes from the node, with the probability of each neighboring node being selected matching the calculated probability value.

[0122] Step 3: Based on the probability value obtained in the previous step and other walking parameters, such as walking length and the number of times to walk from each node, start walking to obtain several walking paths;

[0123] Step 4: Based on the traversal path, call the Word2vec method to train and obtain word vectors;

[0124] Step 5: Perform multi-label classification on the knowledge network nodes to test the classification performance of the algorithm.

[0125] Multi-relation reasoning can be specifically described as extending the link (Link Extendable, LE) after finding a link between two knowledge ontologies to discover latent semantics or broader relationships on the link. This invention employs a greedy algorithm from reinforcement learning to achieve fast convergence reasoning in multi-relationship processing. The specific steps are as follows:

[0126] 1) Construct possible LE patterns. For all edge types on the knowledge network GT / S, any combination of three edge types can form a candidate LE pattern.

[0127] 2) Reinforcement learning implements mode selection by traversing the entire graph to find all instances corresponding to different modes, calculating the weight and value of different modes, and solving for the reliable LE mode through a greedy algorithm.

[0128] 3) Use the selected pattern to perform matching in the network to obtain the reasoning result.

[0129] In terms of knowledge change reasoning in the dynamic evolution of the entire network, knowledge discovery in the military intelligence field must address the rapid changes in knowledge. This invention employs the Spark platform's Streaming framework to achieve dynamic reasoning and knowledge discovery for knowledge networks. To reduce the incremental algorithmic pressure on the platform from sparse data streams, a time-slice approach is introduced to cumulatively expand the network topology decoupling. Each node and connecting edge is labeled with a position tag in each time slice, indicating the existence or state transition of network nodes and edges in the corresponding time slice, thus enabling the discovery and reasoning of knowledge groups. The knowledge ontology dynamic evolution module also updates the activity status attributes of each domain knowledge entity. The update rules for the activity status of different domain knowledge entities can be customized. For example, some weapons will remain active as long as they are under research or exist, while some events, such as military scenarios like meetings, can be adjusted to an inactive state at the end of the meeting. Preferably, different domain knowledge entities have different activity values ​​to distinguish between key and non-key events. For example, military experts can adjust the activity value based on their assessment of military scenarios. The activity value of some resource entities is related to their attribute values, while the activity value of some individuals or organizations is related to the complexity of their relationships or the activity value of their associated nodes, and so on.

[0130] This invention constructs a knowledge graph based on extracted entities, attributes, and relationships. For example, it builds a semantic network based on a graph data structure, consisting of nodes ("entities") and edges ("relationships"), connecting all different types of information (people, organizations, weapons, countries, etc.) into a relational network. This allows for analysis and problem-solving from a "relationship" perspective. To facilitate operation, the knowledge extraction intelligent analysis layer also provides several plugins to support knowledge graph storage, querying, and front-end display functions; automatic updates of the knowledge graph; and manual editing of the knowledge graph.

[0131] III. Knowledge Application Decision-Making Level

[0132] The knowledge application decision-making level provides business support for knowledge application in the military intelligence field. On the one hand, it provides military personnel with business applications related to military intelligence knowledge, such as providing business service capabilities like military strategy map display, intelligent search, ontology target profiling, relationship network analysis, time series analysis, and knowledge recommendation, as well as providing historical event display functions and providing data and computing service support for upper-level applications. On the other hand, the knowledge application decision-making level also needs to judge the evolution status and trends of military strategy-related information, collect the real-time military strategy intentions of military personnel, dynamically update the military strategy guidance information for the next cycle based on the evolution status and evaluation results of military strategy-related information, automatically optimize core keywords using military strategy guidance information, and guide the knowledge extraction intelligent analysis layer to optimize the knowledge reasoning process so that the knowledge reasoning results are more in line with the real-time military strategy needs.

[0133] The knowledge application decision-making layer includes an evolution state discrimination module, a military strategy map generation module, an evaluation result collection module, and a military strategy guidance information update module. The evolution state discrimination module, based on the findings and reasoning results of the knowledge group, discerns the evolution state and trends of military strategy-related information from several aspects, including individuals, organizations, resources, and military scenario events. The military strategy map generation module generates a military strategy map based on the output of the evolution state discrimination module. The evaluation result collection module collects the evaluation results of military experts on military strategy-related information. The military strategy guidance information update module dynamically updates the military strategy guidance information for the next cycle based on the evolution state and evaluation results of the military strategy-related information.

[0134] The knowledge application decision-making layer also includes a database module, an entity and relation extraction module, an event topic database, an organization database, and other query modules.

[0135] This invention, guided by military strategy, explores the application of knowledge in the field of military intelligence, focusing on knowledge retrieval, recommendation, and visualization. Taking knowledge retrieval as an example, user queries are often represented by sets of keywords, while the answers are typically sets of connected nodes, represented in the knowledge network as subgraphs matching the search content. Due to the semantic completeness of keyword matching, a large knowledge network can have a wide variety of subgraphs that meet the criteria, exhibiting different topologies or keyword distributions. During the query process, the relevance of nodes and edges is first calculated. A heuristic search algorithm is used to generate search results, and a ranking algorithm is then used to sort the top-k subgraphs, providing the final top-k search results. The knowledge application decision layer, tailored to the specificities of military strategy, provides query modules such as a database, a unary relation query component, a multivariate relation query component, an event topic database, and an organization database. Preferably, the entity and relation extraction module also includes unary and multivariate relation query components to provide unary and multivariate relation query functions.

[0136] The database module includes a database management component, a database query component, and a database display component. The database management component stores the cleaned and categorized data from the data acquisition module into two databases: a basic database and a trend database. It also adds entry time, publication date, and keyword attributes to each incoming message. The database query component provides data query functionality, including a publication date time range plugin, a keyword query plugin, and database type options. These plugins allow users to search for messages in the selected databases based on their input publication date range and keywords. The database display component then pushes the search results to the main display screen.

[0137] The entity and relation extraction module includes an entity recognition component, an entity graph component, and a relation recognition component. The entity recognition component queries relevant entities based on input keywords and pushes the relevant attributes of the queried entities to a large display screen. The entity graph component includes a first entity name plugin, which queries the relevant nodes of the entity in the knowledge graph based on the input single entity name, obtaining and displaying its associated attributes and relation query results. The relation recognition component includes a relation selection plugin and multiple second entity name plugins. Based on the input multiple relation types, the relation recognition component infers and displays the knowledge graph of specified relationships between multiple input entities through multi-layer penetration between multiple entities. The specified relationships include associations, associated organizations, and associated events between entities.

[0138] For example, inputting a single entity into the unary relation query component allows you to query related nodes of that entity in the knowledge graph, obtaining query results such as associated attributes and relationships. Examples include unary relation (person-person relationship), unary relation (organization-person relationship), and unary relation (person-organization relationship). Inputting multiple entities into the multi-dimensional relation query component allows you to obtain query results such as the relationships between entities, associated organizations, and associated events through multiple layers of inter-entity permeation. Examples include multi-dimensional relation (person-person relationship) and multi-dimensional relation (person-organization relationship) queries, etc.

[0139] The following examples illustrate the extraction results of entities, attributes, etc.

[0140] Entity extraction types: people, organizations, military units, countries, cities, regions, weapons, etc.

[0141] Attribute extraction type:

[0142] The types of attributes extracted from individuals include: Chinese name, English name, nationality, date of birth, place of birth, gender, position, and education. The types of attributes extracted from organizations include: nature, establishment time, and establishment location.

[0143] Relation extraction type:

[0144] Relationships between characters:

[0145] <People> Colleagues<People>, <People> Relatives and Friends<People>, <People> Classmates<People>, <People> Fellow Villagers<People>, <People> Superiors<People>, <People> Partners<People>, <People> Collaborators<People>, <People> Supporters<People>, <People> Assisters<People>.

[0146] The relationship between individuals and organizations:

[0147] <Person> serves <Organization>, <Person> leader <Organization>, <Person> core member <Organization>, <Person> member <Organization>, <Person> member <Organization>, <Person> supports <Organization>, <Person> funds <Organization>, <Person> establishes <Organization>, <Person> cooperates with <Organization>.

[0148] Relationships between organizations:

[0149] <Organization> cooperation, <Organization> funding, <Organization> affiliation, <Organization> competition, <Organization> hostility, <Organization> support.

[0150] The relationship between individuals and nations:

[0151] <Person>Born in <Country>, <Person>Nationality in <Country>, <Person>Resides in <Country>, <Person>Studied in <Country>, <Person>Works in <Country>, <Person>Serves in <Country>, <Person>Visited <Country>.

[0152] The relationship between individuals and military units:

[0153] The person belongs to the military unit, the person serves the military unit, and the person visits the military unit.

[0154] The relationship between people and cities:

[0155] <People> were born in <City>, <People> live in <City>, <People> work in <City>, <People> visit <City>.

[0156] The relationship between weapons and the state:

[0157] The nation develops weapons, the nation possesses weapons, the nation purchases weapons, the nation intercepts weapons, the nation uses weapons, and the nation destroys weapons.

[0158] Existing multi-relational queries often calculate the subgraph with the highest weight in the entire domain knowledge network based on the set of keywords input by the user. Taking two keywords as an example, this involves calculating the shortest path or the path with the optimal meaning between the two nodes. Two situations may arise in the graph: first, a node may connect to many nodes, but the node and its nearest neighbor may have very low relevance; second, although the distance between two nodes is long, the content of the node and the nodes along the path may have very high similarity, resulting in a high relevance for these nodes when calculated together. Therefore, this invention, after calculating the shortest path between two nodes, uses the Steiner node representation method to determine potential hub nodes on the shortest path (the HITS algorithm can be used to calculate the hub value of each node). If a hub node exists, a second shortest path is searched; if no next shortest path exists, the previous shortest path containing a hub node is used. This method can relatively compensate for the shortcomings of directly calculating the shortest path between each pair of nodes.

[0159] The event topic database analyzes key events from multiple dimensions, including basic event information, event introduction, development history, related messages, relationship graphs, and trending keywords. It supports the addition and configuration management of key events, summarizing and analyzing different event topics arising from the evolution of events, aggregating related messages and performing timeline analysis based on event topics, and viewing related messages for event topics. Specifically, the event topic database includes an event analysis component, an event addition component, an event association component, a related message aggregation component, and a related message query component. The event analysis component analyzes key events from multiple dimensions, including basic event information, event introduction, development history, related messages, relationship graphs, and trending keywords. The event addition component allows manual addition of key events. The event association component summarizes and analyzes different event topics arising from the evolution of events. The related message aggregation component aggregates related messages for event topics according to a timeline. The related message query component provides the function of viewing related messages.

[0160] The organization library manages key organizations from dimensions such as basic organizational information, organizational introduction, organizational history, organizational structure, and relationship graph, and supports the addition and configuration management of organizations of particular interest. Specifically, the organization library includes an organization basic information management component, an organization history management component, an associated event management component, an associated message management component, and a relationship graph generation component. The organization basic information management component manages the basic information of all organizational entities in the organization library; the organization history management component manages the changes in the organization's related personnel structure and resource information; the associated event management component adds event types related to all organizational entities in real time based on knowledge reasoning results; the associated message management component manages all messages related to the organization; and the relationship graph generation component organizes and generates a binary relationship knowledge graph related to the organization.

[0161] 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, CD-ROM, optical storage, etc.) containing computer-usable program code. The solutions in the embodiments of this application can be implemented in various computer languages, such as the object-oriented programming language Java and the interpreted scripting language JavaScript.

[0162] 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, produce 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.

[0163] 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 function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0164] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment, causing a series of operational steps to be executed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that run 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.

[0165] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0166] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A knowledge graph-based intelligence collaborative processing platform, characterized in that, The intelligence collaborative processing platform includes a data acquisition and processing layer, a knowledge extraction and intelligent analysis layer, and a knowledge application decision-making layer. The data acquisition and processing layer includes a data acquisition module, a data format conversion module, and a knowledge base management module; The data acquisition module obtains military intelligence business data from multiple data sources, cleans and classifies it, and stores it in the corresponding database. This military intelligence business data includes various military intelligence results analyzed and researched in recent years, as well as intelligence information reported in real-time by various levels of departments. The data format conversion module is used to uniformly represent the multi-source, heterogeneous military intelligence business data collected by the data acquisition module. The knowledge base management module generates a semantic network knowledge base based on the format-converted military intelligence business data, dynamically expands the semantic network knowledge base, and integrates duplicate and conflicting content in the knowledge ontology construction. The semantic network knowledge base includes four types of domain knowledge ontology: people, organizations, resources, and military scenario events. Each domain knowledge entity has corresponding attribute information, and each domain knowledge entity includes an active status attribute. The knowledge extraction intelligent analysis layer includes a keyword extraction module, a keyword layering module, a semantic fusion module, and a knowledge ontology dynamic evolution module. The keyword extraction module is used to obtain several words or phrases that are highly relevant to the entity as basic keywords; The keyword layering module constructs a core keyword set for the current period based on the evolution trend of historical military strategy-related information and the military strategy guidance information of the current period; then, according to the dependency relationship of syntactic analysis and the co-occurrence frequency of keywords, it layers and classifies the remaining basic keywords to construct a spatial distribution model between keywords and entities, and between keywords themselves. The semantic fusion module takes the relationship between keywords and knowledge ontology of various domains as the main feature, establishes feature vectors of knowledge ontology of various domains, calculates the similarity between each pair of knowledge ontology of various domains, and obtains the relationship between knowledge ontology of various domains based on a general threshold; starting from keyword frequency and distance features, it mines the hierarchical and sequential relationship between knowledge entities of various domains, constructs a knowledge ontology evolution semantic network, and performs semantic fusion on multi-source heterogeneous data in the semantic network knowledge base. The knowledge ontology dynamic evolution module starts from the dynamic evolution of knowledge ontology itself and between each domain, performs upper and lower layer semantic reasoning on the fused semantic network knowledge base, finds the link between two knowledge ontology and extends the link, discovers the potential semantics or broader relationships on the link, and mines the evolution trend of multiple relationships between people, organizations, resource behavior sequences and military scenario events, and constructs a multi-dimensional knowledge semantic network; it introduces the time slice method to cumulatively expand the network topology decoupling, and sets position labels for each node and the edge of the link in each time slice to indicate whether there is a state transition of network nodes and edges in the corresponding time slice, discovers and reasons about knowledge groups, and updates the active state attributes of each domain knowledge entity; The knowledge application decision layer includes an evolution state discrimination module, a military strategy map generation module, an evaluation result collection module, and a military strategy guidance information update module. The evolution status discrimination module, based on the discovery and reasoning results of the knowledge group, discerns the evolution status and trend of military strategy-related information from several aspects, including individuals, organizations, resources, and military scenario events. Simultaneously, it updates the active status attributes for each domain knowledge entity. The military strategy map generation module, based on the updated active status attributes, selects several key individuals, organizations, resources, and events of interest for the current period and generates a military strategy map. The evaluation result collection module collects the evaluation results of military experts on military strategy-related information. The military strategy guidance information update module dynamically updates the military strategy guidance information for the next period based on the evolution status and evaluation results of the military strategy-related information.

2. The knowledge graph-based intelligence collaborative processing platform according to claim 1, characterized in that, The basic structure of the data in the semantic network knowledge base is: SKB = (C, E, A, V), where C = {c1, c2, ..., c...} n } represents the category set, E = {e1, e2, ..., e m Let A = {a1, a2, ..., a} represent a set of entities. k } represents the set of entity attributes, where a i ∈A represents a basic attribute. This represents the value corresponding to the attribute; in the category set {C}, there is a parent-child relationship between categories, i.e., c j ∈c s , indicating category c j It is category c S Subclasses of.

3. The knowledge graph-based intelligence collaborative processing platform according to claim 1, characterized in that, The knowledge base management module includes an entity construction component, a semantic network knowledge base generation component, and a conflict fusion component. The entity construction component is used to study the characteristics of some unstructured text, manually construct an initial military domain knowledge entity attribute library, and then expand the military domain knowledge entity attribute library based on word similarity. At the same time, the unstructured text is labeled with entities and attributes using a dictionary method combined with part-of-speech attribute discrimination. The semantic network knowledge base generation component uses keyword spatial location co-occurrence features as a training set, employs a deep learning model to train the labeled entity and attribute data, uses the trained deep learning model to extract new domain knowledge data, evaluates the effectiveness of the extracted entities and attributes, dynamically expands the evaluated effective entities and attributes into the military domain knowledge entity attribute database, and then identifies domain knowledge entities and attributes in massive texts based on reinforcement machine learning methods. The conflict fusion component analyzes terminological conflicts, semantic conflicts, and predicate conflicts, and uses a hybrid approach of logic tree fusion, frequency fusion, and syntactic fusion to select and fuse repetitive and conflicting content in the knowledge ontology construction.

4. The knowledge graph-based intelligence collaborative processing platform according to claim 3, characterized in that, The conflict fusion component includes a terminology database, a predicate database, an ontology database, a terminology conflict processing unit, a semantic conflict processing unit, a predicate conflict processing unit, and a fusion unit; The terminology database and predicate database extract relevant knowledge ontology from the military domain knowledge entity attribute database, respectively, and construct terminology sets, predicate sets, and semantic sets, which are then transmitted to the ontology database. The ontology database analyzes the terminology sets, predicate sets, and semantic sets to identify existing terminology conflicts, predicate conflicts, and semantic conflicts. The terminology conflict processing unit calls a logic tree fusion model, a frequency fusion model, and a syntax fusion model to process terminology conflicts; the semantic conflict processing unit calls a syntax fusion model to process semantic conflicts; and the predicate conflict processing unit calls a frequency fusion model and a syntax fusion model to process predicates. The fusion unit integrates the processing results of the terminology conflict processing unit, semantic conflict processing unit, and predicate conflict processing unit to generate fused knowledge items; The logic tree fusion model fuses conflicting items by determining the logic tree relationship between them. The frequency fusion model uses high-frequency co-occurrence knowledge items as the highest score output; the syntactic fusion model uses syntactic dependency tree discriminant to determine the semantic space dependency relationship between entities, attribute words and attribute value words, and to select and discard multiple items that cause conflict.

5. The knowledge graph-based intelligence collaborative processing platform according to claim 1, characterized in that, The keyword extraction module obtains a word set D-terms by segmenting the document set D of each domain knowledge entity content. The document set D is then mapped to a graph G, where vertices are words and edges represent the co-occurrence frequency of two words. The network node centrality between nodes is then calculated, and the network node centrality of each word is sorted. Some words are merged into phrases using an n-gram algorithm, and the top-ranked results are output as basic keywords related to the domain knowledge entity.

6. The knowledge graph-based intelligence collaborative processing platform according to claim 1, characterized in that, The keyword layering module filters entities that are active in the current period based on the evolution trend of historical military strategy-related information. Then, for each entity and keyword association, it selects the first keyword related to the entity from the basic keywords. It acquires and analyzes the military strategy guidance information of the current period, and selects the second keyword related to the entity corresponding to the military strategy guidance information from the basic keywords. The first keyword and the second keyword are combined to form the core keyword set of the current period.

7. The knowledge graph-based intelligence collaborative processing platform according to claim 1, characterized in that, The semantic fusion module uses an entity similarity calculation model and a category network fusion model to fuse the semantic knowledge network; The entity similarity calculation model uses the following formula to fuse entities between the semantic network knowledge bases SKB1 and SKB2: f:(C i ,E i ,A i ,V i )×(C j ,E j ,A j ,V j )→(C,E,A,V) Specifically, if the entity names in two knowledge bases are the same, then the two knowledge bases are considered to describe the same content; if the entity name in SKB1 is e i No matching entity name was found in SKB2. j However, it finds e that has the same domain knowledge description or a similarity greater than the first preset similarity threshold. j If the entity name in SKB1 is the same as the entity name in the other knowledge base, then the entities in the two knowledge bases are considered to be the same; if the entity name in SKB1 is the same as the entity name in the other knowledge base, then the entities in the two knowledge bases are considered to be the same. i In SKB2, only pairs of characters with a similarity less than the first preset similarity threshold can be found. j However, if the similarity of other attributes and attribute values ​​is greater than the second preset similarity threshold, then the entities in the two knowledge bases are considered to be the same; when two entities are determined to be the same, let e = e i Or e = e j C = (C i ∪C j ), A = (A i ∪A j ),in When two entities are judged to be identical or two semantic network category nodes are merged, the category network fusion model is used to execute C = (C i ∪C j When an entity has only one parent class, determine whether the parent classes of the two knowledge bases are the same: If c i With c j If the terms used in the category are the same, then the parent class is considered to be the same. If c i With c j If the category terms are inconsistent, but the domain knowledge descriptions are identical or the similarity exceeds a third preset threshold, then the categories in the two knowledge bases are considered identical. If the category terms and domain content descriptions are inconsistent, then the similarity of other attributes and attribute values ​​of the category exceeds a fourth preset similarity threshold, and the categories in the two knowledge bases are considered identical. When an entity has multiple parent classes in different knowledge bases, and... Then C i ∩C j The common parent class of the entities is used, while other parent classes invoke the deep inheritance algorithm and the fusion algorithm.

8. The knowledge graph-based intelligence collaborative processing platform according to claim 1, characterized in that, The knowledge ontology dynamic evolution module includes a unit knowledge reasoning component, a multi-relationship reasoning component, and a dynamic knowledge reasoning component; The unit knowledge reasoning component loads knowledge network data and establishes a data structure for each node's neighbor nodes and label information. For each node in the knowledge network, based on the label attributes of the node and its neighbor nodes, and the specified adjustable parameter p for the label information ratio, it calculates the probability value of the node's neighbor nodes being traversed, and randomly selects several neighbor nodes from the node's neighbors, with the probability of each neighbor node being selected matching the calculated probability value. Then, based on the obtained probability value and the traversal parameters, it initiates traversal to obtain several traversal paths. Based on the traversal path, the Word2vec method is called for training to obtain word vectors, which are then used to perform multi-label classification tasks on the knowledge network nodes. The multi-relation reasoning component constructs possible candidate LE patterns based on the combination of all edge types on the knowledge network; it traverses the entire graph to find all instances corresponding to different candidate LE patterns, calculates the weights and values ​​of different candidate LE patterns, solves for reliable LE patterns using a greedy algorithm, and uses the selected reliable LE patterns to perform matching in the knowledge network to obtain the reasoning results. The dynamic knowledge reasoning component introduces a time-slice approach to cumulatively extend the decoupling of network topology. By setting a position label for each node and the linked edge in each time slice, it indicates whether there is a state transition of network nodes and edges in the corresponding time slice, thereby discovering and reasoning about knowledge groups.

9. A latent relationship prediction method based on knowledge graphs, characterized in that, The potential relationship prediction method is performed based on the knowledge graph-based intelligence collaborative processing platform as described in any one of claims 1-8; The potential relationship prediction method includes the following steps: Military intelligence business data is obtained from multiple data sources, cleaned and classified, and then stored in the corresponding databases. The military intelligence business data includes various military intelligence results data analyzed and researched in recent years and intelligence information reported by departments at all levels in real time. The system unifies the representation of multi-source heterogeneous military intelligence business data collected by the data acquisition module; it generates a semantic network knowledge base based on the format-converted military intelligence business data, dynamically expands the semantic network knowledge base, and integrates duplicate and conflicting content in the knowledge ontology construction; the semantic network knowledge base includes four types of domain knowledge ontology: people, organizations, resources, and military scenario events, each domain knowledge entity has corresponding attribute information, and each domain knowledge entity includes an active state attribute; Several words or phrases that are highly relevant to the entity are selected as basic keywords; Based on the evolution trend of historical military strategy information and the military strategy guidance information of the current period, a core keyword set for the current period is constructed; then, according to the dependency relationship of syntactic analysis and the co-occurrence frequency of keywords, the remaining basic keywords are hierarchically and classified to construct a spatial distribution model between keywords and entities, and between keywords themselves. Using the association between keywords and knowledge ontology in various domains as the main feature, feature vectors of knowledge ontology in various domains are established, and the similarity between each pair of knowledge ontology in various domains is calculated. The relationship between knowledge ontology in various domains is obtained according to a general threshold. Starting from keyword frequency and distance features, the hierarchical and sequential relationships between knowledge entities in various domains are mined, a knowledge ontology evolution semantic network is constructed, and semantic fusion is performed on multi-source heterogeneous data in the semantic network knowledge base. Starting from the dynamic evolution of knowledge ontology itself and between different domains, semantic reasoning is performed on the integrated semantic network knowledge base at both upper and lower layers. After finding the links between two knowledge ontology, the links are extended to discover potential semantics or broader relationships on the links. The evolutionary trends of multiple relationships between people, organizations, resource behavior sequences, and military scenario events are mined to construct a multi-dimensional knowledge semantic network. The network topology is decoupled and expanded cumulatively by introducing a time-slice approach. Each node and the edge of the link are labeled with a position label in each time slice to indicate whether there is a state transition of network nodes and edges in the corresponding time slice. Knowledge groups are discovered and reasoned, and the active state attributes of knowledge entities in each domain are updated. Based on the findings and reasoning of the knowledge community, the evolution status and trends of military strategy-related information are determined from several aspects, including people, organizations, resources, and military scenario events. At the same time, the active status attributes of each knowledge entity are updated. Based on the updated active status attributes, several key people, organizations, resources, and events of the current period are selected to generate a military strategy map. Collect assessment results from military experts on military strategy-related information; dynamically update military strategy guidance information for the next cycle based on the evolution of military strategy-related information and assessment results.

Images