Data construction task planning method based on knowledge graph

Abstract

The invention relates to a data construction task planning method based on a knowledge graph, and belongs to the technical field of data construction task planning in artificial intelligence. According to the method, the knowledge graph is constructed by adopting an ontology technology of the knowledge graph, and knowledge reasoning iteration is performed on the constructed knowledge graph to realize enhancement of the knowledge graph; then determining a data resource construction demand by using the enhanced knowledge graph, and identifying a specific collection object of the data resource construction collection demand; and determining a target data construction task according to the specific collection object, issuing the target data construction task to a corresponding operation unit, and completing planning of the data resource construction task after collection and reporting. Therefore, the isolated data are associated in a knowledge graph mode, the data are effectively managed and efficiently utilized, deep-level target information and features are mined, and support is provided for related requirements of task planning.

Description

Technical Field

[0001] The present invention relates to a data construction task planning method based on knowledge graph, and belongs to the technical field of data construction task planning in artificial intelligence. Background Art

[0002] Data construction task planning is an action based on task templates and data to execute established workflows, analyze and supplement battlefield information. Data construction task planning includes basic information management, task planning, task review, task management, etc. It can receive targets issued by command units and classify and decompose targets; based on existing data conditions and historical data analysis, it can predict which aspects of data may be missing or insufficient, and provide construction tasks for designated data based on the prediction results; after approval, it is automatically distributed to relevant data collection and compilation units; at the same time, it provides task tracking, which can understand the execution status and progress of tasks in real time, and make necessary adjustments and feedback. At present, the content of data construction is extensive, the requirements are complex, and there are many types of observation targets. There are large differences in the collection cycle and spatial distribution of different types of observation targets, and it is impossible to make independent data express the relationship between various elements. At present, the relevant technology of task planning is to obtain intelligence information and various environmental information, including natural, humanistic, transportation, economic activities and other information, and use new technologies such as knowledge base, rule base, artificial intelligence, cloud computing, etc. to analyze and predict data for the acquired intelligence data, and manually mine deep information and features from it to provide support for decision-making. The problem is that the existing knowledge bases and standard libraries are independent of each other and cannot simultaneously consider the relationship between each element and the whole; secondly, artificial intelligence data reasoning or semantic analysis lacks standardized model analysis for many data objects. Therefore, the data based on the current data construction task planning is mostly isolated and poorly correlated, making task planning difficult and difficult to manage. Summary of the invention

[0003] The purpose of this invention is to provide a data construction task planning method based on knowledge graph to solve the problem that current task planning is difficult and not easy to manage.

[0004] In order to solve the above technical problems, the present invention provides a data construction task planning method based on knowledge graph, which comprises the following steps:

[0005] 1) Use structured data and unstructured information in the target data resources to build a knowledge graph;

[0006] 2) Clarify data construction priorities based on data construction goals and knowledge graphs;

[0007] 3) Perform knowledge reasoning iteration on the constructed knowledge graph to obtain an enhanced knowledge graph for the data construction domain;

[0008] 4) Determine the data resource construction collection requirements based on the enhanced data construction domain knowledge graph and data construction priority list, and identify the specific collection objects of the data resource construction collection requirements;

[0009] 5) Determine the target data construction task according to the specific collection object and send it to the corresponding operating unit. After collection and reporting, the data resource construction task is completed.

[0010] Furthermore, the knowledge graph construction process in step 1) is as follows:

[0011] Acquire structured data and unstructured information text in the target data construction domain knowledge base;

[0012] Entity extraction, relationship extraction, and attribute extraction are performed on the structured data and unstructured information text, and the target data construction domain knowledge base is converted into a corresponding entity ontology, so as to establish a knowledge graph.

[0013] Furthermore, entity extraction, relationship extraction, and attribute extraction are implemented using a knowledge extraction model. The knowledge extraction model is a pre-trained model for extracting entity information in a knowledge base in the target data construction field, and includes an image feature conversion structure and a semantic feature conversion structure.

[0014] Furthermore, the constructed knowledge graph is a knowledge graph that has undergone knowledge fusion and disambiguation processing.

[0015] Furthermore, the method also includes storing the constructed knowledge graph using a database such as Neo4j or ArangoDB.

[0016] Furthermore, in step 3), semantic reasoning of knowledge representation, reasoning based on explicit rules, or knowledge discovery and acquisition based on specific mathematical models are used to iterate knowledge reasoning.

[0017] Furthermore, the method also includes real-time monitoring of the integrity, consistency, timeliness, security and availability of the knowledge ontology in the knowledge graph, and optimizing the data resource construction process based on the monitoring results.

[0018] Furthermore, the step 2) also includes target decomposition.

[0019] The beneficial effects of the present invention are as follows: as a pioneering invention, the present invention uses the ontology technology of knowledge graph to construct knowledge graph, and performs knowledge reasoning iteration on the constructed knowledge graph to enhance the knowledge graph; then the enhanced knowledge graph is used to determine the data resource construction needs, and the specific collection objects of the data resource construction collection needs are identified; the target data construction tasks are determined according to the specific collection objects and issued to the corresponding operating units, and after collection and reporting, the planning of the data resource construction tasks is completed. It can be seen that the present invention associates isolated data through the knowledge graph, effectively manages and efficiently utilizes the data, digs out deep target information and features, and provides support for the relevant needs of task planning. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 It is a flow chart of the data construction task planning method based on knowledge graph of the present invention;

[0021] Figure 2 is a flow chart of knowledge graph construction used in the embodiment of the present invention;

[0022] Figure 3 is a schematic diagram of the basic unit of the knowledge graph used in the embodiment of the present invention;

[0023] Figure 4 is a flowchart of a specific embodiment of the target library used in the embodiment of the present invention;

[0024] Figure 5 is a schematic diagram of a specific embodiment of a resource library used in an embodiment of the present invention;

[0025] Figure 6 is a schematic diagram of a specific embodiment of a task in a requirement library used in an embodiment of the present invention;

[0026] Figure 7 is a schematic diagram of the knowledge graph construction used in the embodiment of the present invention;

[0027] Figure 8 is a schematic diagram of target decomposition and priority setting adopted in an embodiment of the present invention;

[0028] Fig. 9 It is a schematic diagram of graph enhancement of knowledge reasoning adopted in an embodiment of the present invention. DETAILED DESCRIPTION

[0029] The specific implementation of the present invention will be further described below in conjunction with the accompanying drawings.

[0030] The present invention first constructs a knowledge graph using structured data and unstructured information in the target data resources, and clarifies the data construction priority according to the data construction goals and the knowledge graph; then, the constructed knowledge graph is iterated through knowledge reasoning to obtain an enhanced data construction field knowledge graph; then, based on the enhanced data construction field knowledge graph and the data construction priority list, the data resource construction acquisition requirements are determined, and the specific acquisition objects of the data resource acquisition construction requirements are identified; finally, the target data construction task is determined according to the specific acquisition object and issued to the corresponding work unit, and after collection and reporting, the data resource construction task is completed. The specific implementation process of the present invention is described in detail below using battlefield data construction task planning as an example. The implementation process of this method is as follows: Figure 1 As shown, the specific implementation steps are as follows.

[0031] 1. Use structured data and unstructured information in the target data resources to build a knowledge graph.

[0032] like Figure 2 As shown in the figure, knowledge graph construction requires two processes: knowledge extraction and knowledge fusion disambiguation. Knowledge extraction includes entity extraction, relationship extraction and attribute extraction. Knowledge fusion disambiguation is to use data integration results to fuse redundancy and correct ambiguity in knowledge extraction results.

[0033] Among them, knowledge extraction is to extract information from structured and semi-structured texts; use natural language processing technology, especially named entity recognition (NER), to identify entities with specific meanings in the knowledge graph from unstructured texts, identify semantic relationships between entities, and convert these relationships into edges in the knowledge graph. Define concepts, classes, attributes and relationships in the knowledge graph, select appropriate knowledge representation models, such as RDF, OWL or attribute graph models, and form an ontology model of battlefield data resources, that is, build a triple structured framework of entity-relationship-entity to provide a skeleton for the knowledge graph, which can convert complex information in the real world into a format that can be processed by computers. The purpose of knowledge fusion and disambiguation is to integrate duplicate knowledge, integrate and standardize ambiguous problems, and improve the quality of data through data cleaning technologies such as denoising, data normalization and missing value processing, laying the foundation for the efficiency and accuracy of subsequent processing; further, through real-time stream processing technology, real-time data can be quickly processed and analyzed to achieve immediate response to battlefield situations. The knowledge graph construction process provided in the embodiment of the present application is as follows: Figure 7 As shown, the specific steps include:

[0034] 1) Obtain target knowledge structured data and unstructured information text.

[0035] The target knowledge structured data includes battlefield information required for data construction. As an example, assuming that data construction requires geographic location information, it is necessary to obtain data containing battlefield geographic location information as target data by obtaining the geographic subject knowledge base in the field of data construction. Unstructured information text is used to indicate the semantic information that needs to be extracted from the target knowledge base as a target type. As an example, if the target information text is a certain oil depot, the entity information can be "stationary" (state classification) and "3000 (square meters)" (area).

[0036] The target type refers to the relationship information corresponding to the entity information. For example, Figure 3 As shown, for example, if the target text is an information text of a certain oil depot, and the entity information is "3000 (square meters)" (floor area), then the target type can be "floor area (m2)"; for example, if the target structured data is an infrared remote sensing image parameter, the entity information can be "12μm" (sensor resolution), then the target type can be "pixel size". The information text is used to indicate the target type of the entity information that needs to be extracted from the target knowledge base. As an example, assuming that the information in the target knowledge base is "a 160,000-ton oil tanker arrives at a certain offshore oil port", the entity information is "160,000 tons", and the target type is "tonnage", the information text is used to indicate the target type corresponding to the entity information. Among them, the information text can indicate one target type or multiple target types, which is not specifically limited here.

[0037] 2) Through the knowledge extraction model of entity extraction, relationship extraction and attribute extraction, the target knowledge base is converted into the corresponding entity ontology.

[0038] Ontology is something that is distinguishable and exists independently. It is a basic element in the knowledge graph, for example, "a certain oil depot", "stationary", "3000 (square meters)", etc.; relationships are semantic links that connect different entities and refer to the connection between entities. Relationships connect nodes in the knowledge graph through nodes to form a large graph, such as "longitude", "latitude", "area (square meters)", etc.; a class is a collection of entities with common attributes and behaviors, which is used to classify entities, such as the "energy reserve situation" class in the instance knowledge graph; an instance is a concretization and individual case of a class, which inherits the attributes of the class to which it belongs and can have its own unique attribute values. For example, "xx model drone" is an instance of the "mobility" class as a data construction capability resource, such as Figure 5 shown.

[0039] Therefore, the model layer built on the data layer is the core of the knowledge graph, which is usually managed by an ontology library and contains concept templates and hierarchical structures. Ontology provides a formal method for describing knowledge domains, consisting of a set of concepts, relationships between concepts, and a set of rules. The data model is an abstract representation of real-world information, used to organize, manage, and store data to ensure knowledge consistency, effective information retrieval, and efficient data management. Figure 6 As shown, the data layer consists of a series of facts, and knowledge is stored in units of facts. Figure 6 and Figure 4 An example of an entity relationship is "XX data resource construction task-energy reserve situation-stationary-a certain oil depot"; combined with Figure 6 and Figure 5 The instance relationship is "XX data resource construction task a certain oil depot-.TIFF-infrared sensor-xx model UAV".

[0040] This embodiment uses information extraction technology and natural language processing technology to perform entity extraction, relationship extraction, and attribute extraction. Natural language processing (Natural Language Processnng, NLP) is an important direction in the fields of computer science and artificial intelligence. It is a variety of theories and methods for realizing effective communication between people and computers using natural language. Natural language processing involves natural languages ​​used by humans in daily life, as well as computer science languages ​​and mathematical conversion models. The large language model in the field of NLP has also become an important technology for pre-training model training in the field of artificial intelligence. The fine-tuned large language model can be widely used in technologies including text processing, semantic understanding, machine translation, relationship extraction, knowledge graphs, etc.

[0041] The knowledge extraction model selected in this embodiment is pre-trained and is used to extract entity information from the knowledge base in the target data construction field. The knowledge extraction model in this embodiment can realize the extraction of image data and text data, and therefore includes an image feature conversion structure and a semantic feature conversion structure. The image feature conversion is to extract information from the image so that it can be used as an entity or attribute of the knowledge graph; the semantic feature conversion is to extract information from the text so that it can be used as an entity or relationship or attribute of the knowledge graph. The image feature conversion structure is a structure used to convert the target image into an image feature. Specifically, a structure such as the Transformer model of a convolutional neural network CNN can be used as the image feature conversion structure; the semantic feature conversion structure is used to convert the information text into corresponding semantic features. The BERT pre-trained model structure can be used as the semantic feature conversion structure. Its main model structure is composed of a pre-training framework and a large model fine-tuning framework for specific tasks. Image features are used to characterize the basic information of the target image and to identify and mark the image content. For example, in deep learning, taking the classification network as an example, the image is input into the classification network to obtain the category of the image. At this time, if the input of the last layer of the classifier is extracted (generally the output of the second to last layer), the output obtained at this time is the feature obtained by the classifier, or the embedding layer feature. Semantic features are feature words extracted from information text through semantic feature conversion structures, and the features of text information are represented by quantization.

[0042] It should be understood that the knowledge extraction model is based on an existing knowledge base, which has been preprocessed and cleaned, and contains databases, knowledge bases, standard libraries, etc. on various topics, and does not rely entirely on the results of manual recognition or other supervised classification interventions to improve the accuracy of the extracted entity information.

[0043] 3) Through the knowledge fusion disambiguation structure, redundant target entity information of the target type in the target data is fused according to data features and semantic features; ambiguous target entity information of the target type in the target data is corrected. The knowledge extraction model first extracts entity, attribute, and relationship information from the image or text, and integrates the above information after knowledge fusion disambiguation to become the final result output by the knowledge extraction model, which is used to construct a knowledge graph. The knowledge fusion disambiguation structure and the data integration structure are used to construct knowledge of structured data and semantic feature information to obtain the structure of the target entity information. The embodiment of the present application uses the knowledge fusion disambiguation structure and data integration to achieve that the knowledge extraction model can be directly based on the input image and query questions in actual applications. The specific composition of the graph entity and the entity ontology is as follows:

[0044] ID = Oname, which means the unique identifier of the instance primary key value corresponding to the entity;

[0045] <className rdf:about Oname / >That is, the class name corresponding to the instance body in the classification table;

[0046] Attr=attr represents the attribute value corresponding to each common attribute. The attribute value of the instance entity is<Attr rdf:datatype=”&xsd;type”> attr is composed of this;

[0047] The foreign key value PK corresponding to the foreign key attribute is pk, and the class instance composition rule associated with the foreign key correspondence is:<FKobjectProperty rdf:resource=“#pk instance”>

[0048] 2. Clarify the data construction priority and priority list according to the data construction goals and knowledge graph. This embodiment is aimed at battlefield data construction. Now the battlefield data construction goals are decomposed and prioritized. The process is as follows: Figure 8 As shown, this figure is only an example of an embodiment, not all priority indicators, and does not contain specific weight factors or parameters. First, it is necessary to conduct an in-depth analysis of the battlefield data construction target to determine the required various types of knowledge. This includes but is not limited to factors such as geography, meteorology, politics, economy and society. For this reason, relevant public data sets and thematic databases should be collected comprehensively and in-depth, such as geospatial data sets (Natural Earth, Global Land Cover Facility, OpenStreetMap, USGS, NASA Earth Observations) and meteorological data sets (data directory released by China Meteorological Administration, CAMELS data set). These data sets can provide detailed geographic and meteorological information, which is an important basis for battlefield mission planning.

[0049] After clarifying the battlefield data construction goals, the next step is to identify and determine the type, accuracy, update frequency, security, reliability, relevance, and integrity of the specific data required. For example, geospatial data sets need to include information such as topography, hydrology, and land use, while meteorological data sets need to provide meteorological parameters such as temperature, precipitation, and wind speed. At the same time, analyze and select appropriate and efficient data collection methods, such as remote sensing technology, sensor networks, and open data interfaces, to ensure the real-time and accuracy of data.

[0050] In the process of data collection and integration, it is necessary to conduct quality control on the integrity and consistency of the data to ensure the accuracy and reliability of the data. At the same time, considering the timeliness of the data, priority is given to collecting and updating dynamic data that has a greater impact on battlefield decision-making, such as real-time meteorological information and geographical environment changes. In addition, the structured and unstructured data required for data construction, such as text reports, images, videos and sensor data, and their application in the current battlefield data resource system are identified to meet the needs of comprehensive analysis of the battlefield environment. For example, the knowledge graph is applied to the data construction target, and the data source, data integrity, data accuracy, etc. are set as priority parameters. The indicators of data source include trend range and connection with other data sources. The indicators of data integrity include missing risk and influence factors on decision-making. The indicators of data accuracy include technical risk, timeliness and resource cost. The above indicators are used to realize the priority setting of current data construction needs. The indicators are all weighted factors. The larger the trend range, the lower the weight; the more connections with other data sources, the higher the weight; the higher the missing risk, the lower the weight; the greater the influence on decision-making, the higher the weight; the greater the technical risk, the lower the weight; the higher the timeliness requirement, the higher the weight; the higher the resource cost, the lower the weight. The purpose of priority is to form a priority list to ensure the delivery of specific content of subsequent task planning.

[0051] 3. Perform knowledge reasoning iterations on the constructed knowledge graph to obtain an enhanced data construction domain knowledge graph.

[0052] The graph enhancement process of knowledge reasoning adopted by the present invention is as follows: Fig. 9 As shown, the knowledge discovery and acquisition model is used to perform knowledge discovery and acquisition, and the knowledge graph is continuously enriched and improved through iteration, and entities, relationships and attributes are comprehensively and incrementally updated, as well as existing information is updated; a knowledge update mechanism is established to ensure the timeliness and accuracy of the knowledge graph.

[0053] The technical means used in this process include: first, semantic reasoning based on knowledge representation is used to construct the ontology model of battlefield data through knowledge representation methods such as RDF, RDFS and OWL. These models can express the relationship and attributes between entities and provide a basis for semantic reasoning. Then, explicit rule-based reasoning, such as SWRL (Semantic WebRule Language), is used to define more complex rules. These rules can express the complex logical relationship between entities in battlefield data. And knowledge discovery and acquisition methods based on specific mathematical models, such as probabilistic graph models or machine learning algorithms, can be used to automatically discover potential patterns and associations from data. Semantic models or Bayesian models suitable for the characteristics of battlefield data are selected, and the semantic rules of RDFS and OWL (Web Ontology Language) are used for reasoning. Battlefield data is represented in the form of RDF triples, and new knowledge and relationships are discovered through the reasoning rules of RDFS and OWL. At the same time, in view of the limitations of RDFS and OWL in expressing complex battlefield knowledge, tools such as SWRL rule language or Pellet reasoning engine are used for richer expression and reasoning.

[0054] Mathematical models are used to express and discover knowledge in specific fields, which is often difficult to obtain through generalized methods. For example, in battlefield data analysis, graph theory models can be applied to analyze the network structure of energy deployment and mobility, or time series analysis can be used to predict the development trend of battlefield events. In addition, statistical learning methods can be used to mine and analyze patterns and associations in large-scale battlefield data, such as identifying different battlefield behavior patterns through cluster analysis. Through these methods, valuable information can be extracted from complex battlefield data, enhancing the depth and breadth of the knowledge graph.

[0055] Based on the graph enhancement method of knowledge reasoning provided in the above embodiment, in order to further illustrate the process of graph enhancement of knowledge reasoning, the embodiment of the present application also provides a graph enhancement process decomposition of knowledge reasoning for energy reserve data construction task planning. Assuming that the target data construction task is energy reserve situation, the graph enhancement method of knowledge reasoning is applied to the task planning knowledge graph, and the process may include:

[0056] Step 1: Use the knowledge discovery and acquisition model and reasoning engine to discover the initial data construction domain knowledge graph established by the known target knowledge base. During this period, the rules of semantic reasoning based on knowledge representation, the rules of reasoning based on explicit rules, and the rules of knowledge discovery and acquisition based on specific mathematical models are specified to obtain the data construction domain knowledge graph after iterative enhancement of knowledge reasoning.

[0057] Step 2: The iteratively enhanced data construction domain knowledge graph proposes data construction demand suggestions, such as the need to supplement energy reserve data. One of the energy reserve targets is the infrared remote sensing image of an oil depot in a stationary building.

[0058] Step 3: Extract entities through the knowledge graph of data construction domain and determine the resources required to perform data construction tasks. For example, to collect infrared remote sensing images of a certain oil depot, the required resources may be a drone of model xx, because this drone is equipped with an infrared sensor of 7.68mm×6.144mm size. Figure 4 Optionally, by extracting entities from the knowledge graph in the data construction field, another method can be used to determine the resources required to perform the data construction task. For example, to collect infrared remote sensing images of a certain oil depot, the required resources may be the XX team of the XX Surveying and Mapping Bureau, because the target oil depot is a static building and its longitude and latitude are within the operating range of the XX team of the XX Surveying and Mapping Bureau.

[0059] It should be understood that there is not a single method for applying knowledge graphs. Due to the characteristics of knowledge graphs, relationships are established between the attribute information of each entity, and the connections between ontologies are also diverse. When applied, extraction is performed based on the distance of the association. The reasoning rules are determined by semantic reasoning based on knowledge representation, reasoning based on explicit rules, and knowledge discovery and acquisition based on specific mathematical models.

[0060] Step 4: Establish a priority list for target data construction and determine the resources required for the task. After reviewing the data resource construction task based on the knowledge graph in the data construction field, the target data construction task will be issued to the operating unit. The review will be conducted by setting permissions by specific users to determine whether the operating unit can know or execute the task. After compliant collection and reporting, the data resource construction task is completed, and the result of data construction task planning based on the knowledge graph is obtained.

[0061] It should be understood that the loss value can be determined based on the difference between the semantic features corresponding to the inferred entity information and the labeled entity information, and then a loss function can be constructed based on the loss value, and the model parameters of the text recognition model can be adjusted based on the loss function. In this way, the above operations are iteratively performed based on different training samples until the trained knowledge extraction model meets the training end conditions, for example, the number of training times of the knowledge discovery and acquisition model for the knowledge extraction model reaches a preset number of times, or the computing performance of the inference engine reaches the preset performance requirements.

[0062] 4. Monitor task progress and adjust tasks.

[0063] The main purpose of this process is to monitor the integrity, consistency, timeliness, security and availability of the knowledge ontology in real time, and adjust the task planning according to the changes in the battlefield data construction goals, as well as the real-time updating, compilation, analysis, distribution, reporting, withdrawal and other work of monitoring data construction. According to the steps of the present invention, a priority list will be established after the goal is decomposed. If there are major changes in the priority list, the task planning will be adjusted in time. At the same time, based on the monitoring results, the data resource construction process is continuously optimized to improve construction efficiency and data quality; further, data visualization tools are developed and applied to more intuitively understand the data status and monitoring results. As the task progresses, the knowledge graph is regularly updated to ensure that it reflects the latest battlefield data resource status.

[0064] This embodiment also provides a progress monitoring link, which uses data quality assurance principles, including completeness, accuracy, consistency and timeliness, to ensure the completeness of knowledge ontology records and information, the accuracy and consistency of data, and the timeliness of data output. Through the data governance and monitoring system, the real-time update, compilation, analysis, distribution, submission and withdrawal of data construction are continuously tracked to ensure the transparency and traceability of the data process.

[0065] The task planning method of the present invention is designed from the whole life cycle process of knowledge graph construction, battlefield data resource construction demand analysis, task decomposition and priority setting, knowledge graph enhancement, task planning progress monitoring and task adjustment. First, extract structured and semi-structured information; identify semantic relationships between entities; integrate repeated knowledge and integrate normative ambiguity problems; select a suitable storage method for the constructed knowledge graph; then determine the data resource construction requirements based on the knowledge graph, identify the specific collection objects of data resources (required resources or materials); set priorities according to the integrity, consistency, timeliness, security and availability of the data; reason and expand the knowledge graph data; finally, introduce a progress monitoring mechanism, and continuously optimize the data resource construction process based on the monitoring results. The knowledge graph makes task planning more intuitive and easy to manage. This solves the problem of poor correlation of data islands and improves the efficiency of data collection and convergence. Strengthen the compilation and unified storage management of battlefield data construction, ensure the integrity of information content elements and the accumulation and continuity of historical data, thereby improving the efficiency and quality of battlefield data resource construction; enhance the integration ability and real-time performance of data resources.

[0066] Therefore, the present invention utilizes knowledge graph technology to represent battlefield data, and through knowledge extraction, fusion and processing, it mines the potential relationships between multiple intelligent agents, realizes autonomous, flexible and accurate task planning, deepens the construction method of battlefield data, forms a distinctive data development technology route, and provides basic support for the construction of battlefield data resources.

Claims

1. A data construction task planning method based on knowledge graph, characterized in that the method comprises the following steps: 1) Use structured data and unstructured information in the target data resources to build a knowledge graph; 2) Clarify data construction priorities based on data construction goals and knowledge graphs; 3) Perform knowledge reasoning iteration on the constructed knowledge graph to obtain an enhanced knowledge graph in the data construction field; 4) Determine the data resource construction collection requirements based on the enhanced data construction domain knowledge graph and data construction priority list, and identify the specific collection objects of data resource construction requirements; 5) Determine the target data construction task according to the specific collection object and send it to the corresponding operating unit. After collection and reporting, the data resource construction task is completed.

2. The method for planning data construction tasks based on knowledge graph according to claim 1 is characterized in that the knowledge graph construction process in step 1) is as follows: Acquire structured data and unstructured information text in the target data construction domain knowledge base; Entity extraction, relationship extraction, and attribute extraction are performed on the structured data and unstructured information text, and the target data construction domain knowledge base is converted into a corresponding entity ontology, so as to establish a knowledge graph.

3. According to the data construction task planning method based on knowledge graph in claim 2, it is characterized in that entity extraction, relationship extraction, and attribute extraction are implemented by using a knowledge extraction model, and the knowledge extraction model is a pre-trained model for extracting entity information in the target data construction domain knowledge base, including an image feature conversion structure and a semantic feature conversion structure.

4. According to the data construction task planning method based on knowledge graph in claim 1, it is characterized in that the constructed knowledge graph is a knowledge graph after knowledge fusion and disambiguation processing.

5. According to the data construction task planning method based on knowledge graph according to claim 1, it is characterized in that the method also includes storing the constructed knowledge graph using the database Neo4j or ArangoDB.

6. According to the data construction task planning method based on knowledge graph in claim 1, it is characterized in that in step 3), semantic reasoning of knowledge representation, reasoning based on display rules, or knowledge discovery and acquisition based on specific mathematical models are used to perform knowledge reasoning iteration.

7. The data construction task planning method based on knowledge graph according to claim 1 is characterized in that the method also includes real-time monitoring of the integrity, consistency, timeliness, security and availability of the knowledge ontology in the knowledge graph, and optimizing the data resource construction process based on the monitoring results.

8. According to the knowledge graph-based data construction task planning method of claim 1, it is characterized in that the step 2) also includes goal decomposition.

Images