Knowledge recommendation method and device based on space remote sensing atlas, equipment and medium
By constructing an aerospace remote sensing atlas and utilizing a generative large model, the problem of low accuracy in knowledge recommendation caused by the diversity of remote sensing paper structures and the variability of terminology was solved, thus achieving efficient and accurate recommendation of remote sensing knowledge.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- AEROSPACE INFORMATION RES INST CAS
- Filing Date
- 2025-06-13
- Publication Date
- 2026-07-31
AI Technical Summary
Existing remote sensing-related papers are often lengthy, structurally diverse, and use a variety of terminology, leading to low accuracy in recommending aerospace application knowledge.
A knowledge recommendation method based on aerospace remote sensing maps rewrites and breaks down user queries to extract keywords, constructs a knowledge graph by combining multi-source geographic data, generates query statements, and inputs them into a generative large model to obtain answers.
It achieves efficient and accurate recommendation of remote sensing knowledge, and solves the problem of recommendation accuracy caused by the diversity of remote sensing paper structure and the variability of terminology.
Smart Images
Figure CN120804304B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of artificial intelligence technology, and in particular to a knowledge recommendation method, apparatus, equipment and medium based on aerospace remote sensing maps. Background Technology
[0002] Over the years, the field of remote sensing has accumulated massive amounts of data and rich knowledge. The data includes raw remote sensing imagery and various products created based on it; the knowledge encompasses the fundamental theories of remote sensing, as well as innovative achievements resulting from the application of artificial intelligence technologies such as machine learning and deep learning in remote sensing. These remote sensing technologies have not only been continuously deepened within this field but have also been widely applied to many other areas, generating a large number of knowledge outputs in the form of academic papers. These outputs are reflected in two aspects: firstly, the research results in these papers can directly supplement the knowledge of specific research areas; secondly, the research methods used in these papers have strong transferability and can be referenced and applied to other areas, thereby promoting the widespread application and transformation of aerospace information knowledge.
[0003] While large language models possess the ability to understand and process unstructured text, using remote sensing-related papers as a knowledge base presents a more complex challenge than processing general documents. Specifically, these papers are typically lengthy, structurally diverse, use varied terminology, and are extremely knowledge-intensive. Furthermore, many papers exhibit overlap, fragmentation, and redundancy in their research topics, data usage, and methodological choices. These characteristics not only increase the difficulty of text processing but also reduce the accuracy of knowledge recommendation for aerospace applications. Summary of the Invention
[0004] This invention provides a knowledge recommendation method, apparatus, device, and medium based on aerospace remote sensing maps to address the shortcomings of existing technologies that use remote sensing-related papers as knowledge bases. These papers are characterized by their length, diverse structures, varied terminology, and extremely high knowledge density, leading to low accuracy in recommending aerospace application knowledge. This invention achieves efficient recommendation of remote sensing knowledge.
[0005] This invention provides a knowledge recommendation method based on aerospace remote sensing maps, comprising the following steps: In response to a user's input query question, the query question is rewritten and split according to the question template, and at least one keyword is extracted from the rewritten and split query question; In the aerospace remote sensing map, a first target node matching the at least one keyword is determined; the aerospace remote sensing map is a knowledge graph constructed based on entities, relationships and attributes extracted from papers related to remote sensing data, combined with supplementary attributes of regional entities extracted from multi-source geographic data, and constructed according to a preset ontology framework. Obtain a query template that matches the at least one keyword, fill the first target node into the query template, and generate a query statement; Contextual data related to the query statement is obtained from the aerospace remote sensing map, and the contextual data and the query question are input as prompt information into the generative big model to obtain the knowledge answer output by the generative big model.
[0006] According to the present invention, a knowledge recommendation method based on aerospace remote sensing maps is provided, wherein determining a first target node in the aerospace remote sensing map that matches the at least one keyword includes: For each keyword, based on the embedding similarity between the first embedding vector of the keyword and the second embedding vector of each first node in the aerospace remote sensing map, a first target node matching the keyword is determined; the first node is a node with the same entity type as the keyword. Specifically, when the entity type corresponding to the keyword is a task, the first target node for keyword matching is determined based on the embedding similarity between the first embedding vector of the keyword and the second embedding vector of each second node in the aerospace remote sensing map; the second node includes task nodes and process nodes.
[0007] According to the knowledge recommendation method based on aerospace remote sensing maps provided by the present invention, obtaining a query template matching at least one keyword includes: If the at least one keyword is a single keyword and there is no parent keyword for the at least one keyword, obtain the first query template matching the at least one keyword; If the at least one keyword is a single keyword and the at least one keyword has a parent keyword, obtain a second query template matching the at least one keyword and a third query template matching the parent keyword of the at least one keyword.
[0008] According to the knowledge recommendation method based on aerospace remote sensing maps provided by the present invention, obtaining a query template matching at least one keyword includes: When at least one keyword is a first keyword pair, a fourth query template, a fifth query template, and a sixth query template matching the at least one keyword pair are obtained; the first keyword pair includes a target region and a task; the fourth query template is used to query papers that match the target region and the task; the fifth query template is used to query solutions in the papers; and the sixth query template is used to find solutions in papers from regions similar to the target region when no papers belonging to the target region exist. When at least one keyword is a second keyword pair, obtain the seventh, eighth, and ninth query templates matching the at least one keyword pair; the second keyword pair includes a target region and data, the seventh query template is used to query numerical data in the target region, the eighth query template is used to query whether the spatial range corresponding to the spatial data includes the target region, and the ninth query template is used to query whether there is an inclusion relationship between regions.
[0009] According to the present invention, a knowledge recommendation method based on aerospace remote sensing maps is provided, wherein the aerospace remote sensing maps are constructed in the following manner: Acquire multiple papers related to remote sensing data and standardize the text format of the papers; Each paper with a uniform text format and the first prompt information are input into the large language model to obtain the paper summary corresponding to each paper output by the large language model; the first prompt information includes the preset ontology framework and the remote sensing application methodology logic chain. The summary of each paper and the second prompt information are input into the large language model to obtain the entities, relations and attributes in each paper output by the large language model; the second prompt information includes the preset ontology framework and structured knowledge extraction prompt information. For regional entities within an entity, supplementary attributes of the regional entity are extracted by combining multi-source geographic data; Based on the entities, relationships, attributes, and supplementary attributes of the entities in each paper, an aerospace remote sensing map is constructed.
[0010] According to the present invention, a knowledge recommendation method based on aerospace remote sensing maps is provided, wherein the aerospace remote sensing map is constructed based on the entities, relationships, and attributes in each paper, including: For each entity in the paper, create entity-related labels and names; Based on entity-related labels and names, determine whether the current graph database contains the same second target node; If the second target node does not exist in the current graph database, create a new node corresponding to the entity; If the second target node exists in the current graph database, determine whether the second target node has an attribute corresponding to the entity; If the second target node does not have an attribute corresponding to the entity, assign the attribute corresponding to the entity to the second target node; If the second target node has attributes corresponding to the entity, the attribute values are merged. For each relation in the paper, determine the head node and tail node corresponding to the relation, establish the edge between the head node and the tail node, and assign the attribute corresponding to the relation to the edge between the head node and the tail node.
[0011] According to the knowledge recommendation method based on aerospace remote sensing maps provided by the present invention, the preset ontology framework is designed in the following manner: Identify the paper as the primary core entity type, and create associated authors and associated regions for the paper. The task is identified as the second core entity type. A paper-related task is created, and a task-related process, a process-related data processing tool, and data processing tool-related parameters are created. The data processing tool includes at least one of an algorithm model or a software tool.
[0012] The present invention also provides a knowledge recommendation device based on aerospace remote sensing maps, comprising: The first knowledge recommendation module is used to respond to the query questions input by the user, rewrite and split the query questions according to the question template, and extract at least one keyword from the rewritten and split query questions; The second knowledge recommendation module is used to determine the first target node that matches the at least one keyword in the aerospace remote sensing map; the aerospace remote sensing map is a knowledge graph constructed according to a preset ontology framework based on entities, relationships and attributes extracted from papers related to remote sensing data and combined with regional attributes supplemented by multi-source geographic data. The third knowledge recommendation module is used to obtain a query template that matches the at least one keyword, fill the first target node into the query template, and generate a query statement. The fourth knowledge recommendation module is used to obtain contextual data related to the query statement from the aerospace remote sensing map, and input the contextual data and the query question as prompt information into the generative big model to obtain the knowledge answer output by the generative big model.
[0013] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the program to implement the knowledge recommendation method based on aerospace remote sensing maps as described above.
[0014] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the knowledge recommendation method based on aerospace remote sensing maps as described above.
[0015] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the knowledge recommendation method based on aerospace remote sensing maps as described above.
[0016] This invention provides a knowledge recommendation method, apparatus, device, and medium based on aerospace remote sensing maps. It rewrites and breaks down query questions using a question template, extracting keywords to accurately pinpoint the core of the user's question. Secondly, the construction of the aerospace remote sensing map integrates entities, relationships, attributes from remote sensing papers, and multi-source geographic data, allowing previously overlapping, fragmented, and redundant knowledge to be presented in a structured manner, facilitating the rapid and accurate identification of the first target node matching the keywords. Finally, the first target node is filled into the query template to generate a query statement, further clarifying the query direction. The contextual data related to the query statement obtained from the aerospace remote sensing map provides rich background knowledge for the generative large-scale model, enabling it to output accurate answers that conform to the characteristics of remote sensing knowledge, thus achieving efficient recommendation of remote sensing knowledge. Attached Figure Description
[0017] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.
[0018] Figure 1 This is a flowchart illustrating the knowledge recommendation method based on aerospace remote sensing maps provided by the present invention.
[0019] Figure 2 This is a schematic diagram of the relationship of the ontological framework provided by the present invention.
[0020] Figure 3 This is a schematic diagram of the knowledge recommendation device based on aerospace remote sensing maps provided by the present invention.
[0021] Figure 4 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation
[0022] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.
[0023] The knowledge recommendation method based on aerospace remote sensing maps in this invention embodiment, such as... Figure 1 As shown, it includes steps 110, 120, 130 and 140.
[0024] Step 110: In response to the query question input by the user, the query question is rewritten and split according to the question template, and at least one keyword is extracted from the rewritten and split query question.
[0025] It should be understood that user-input queries are often variable and ambiguous, and directly processing these queries may result in inaccurate matching of nodes in the knowledge graph. Therefore, it is necessary to first rewrite and break down the user-input queries so that they can be mapped to one or more predefined, more structured question templates.
[0026] In this step, the user-input query is converted into a more standardized expression. Then, the rewritten query is further broken down into finer-grained questions that conform to predefined question templates for separate processing. Finally, keywords are identified from the rewritten and split questions.
[0027] In this embodiment, five keyword types are defined for extracting keywords from query questions: Task: The main objective or problem of the research; Algorithm model or software tool: A specific method or tool used to solve a problem; Data: Data resources used in the research.
[0028] Region + Task: Research tasks within a specific region; Region + Data: Data resources within a specific region.
[0029] Here, the predefined question templates are used to standardize users' query questions, making them easier to map to nodes and edges in the knowledge graph.
[0030] The problems addressed in this embodiment revolve around keywords such as task, data, algorithm model, software tool, and region. Therefore, two types of problem templates are defined: The first type: Problem templates that are not limited to a specific area, such as the implementation plan of a certain task, the purpose and input data of a certain algorithm model or software tool, or the source and purpose of certain data; The second type: the implementation plan for a certain task in a limited area, and the source and purpose of certain data.
[0031] Step 120: Determine the first target node that matches the at least one keyword in the aerospace remote sensing map; the aerospace remote sensing map is a knowledge graph constructed based on entities, relationships and attributes extracted from papers related to remote sensing data, combined with supplementary attributes of regional entities extracted from multi-source geographic data, and constructed according to a preset ontology framework.
[0032] After extracting the keywords, the same or similar target nodes are found in the aerospace remote sensing map.
[0033] Here, the predefined ontology framework refers to a formal framework used to define and standardize concepts, entities, relationships, and attributes within a domain when constructing a knowledge graph. It provides a standardized template for the construction of knowledge graphs, ensuring that the organization and storage of knowledge are consistent and logical.
[0034] In one example, the default ontology framework is designed as follows: Identify the paper as the primary core entity type, and create associated authors and associated regions for the paper. The task is identified as the second core entity type. A paper-related task is created, and a task-related process, a process-related data processing tool, and data processing tool-related parameters are created. The data processing tool includes at least one of an algorithm model or a software tool.
[0035] First, "paper" is the first core entity type; A "paper" is associated with an "author," and the work of the same author is usually systematic and interconnected. The term "paper" is associated with "region," indicating the scope of the research. Descriptions of the research scope vary widely, ranging from global to any specific local area. Regions are connected by terms like "belongs to" or "contains." A paper may use one or more administrative regions as its experimental area. An experimental area may not be a standard administrative region (e.g., the five central areas of city A), but it may still have a "contains" relationship with a specific administrative region. For regions outside of country A, the entity name attribute value retains its original language. Furthermore, the experimental area description section of a paper typically describes the region's area, geographical location, natural conditions (climate, soil, etc.), and social conditions (economy, population, etc.). These characteristics are crucial for the transfer of the proposed solution. To avoid errors, it is necessary to distinguish whether such "features" apply to the entire experimental area, a part of the experimental area, or the administrative region to which the experimental area belongs. Failure to accurately identify the scope of such features can easily lead to errors. Secondly, "task" is the second core entity type. A "task" is the research topic of a "paper," and multiple papers typically address the same "task." A “task” associated with a “process” refers to a series of steps to achieve that “task”. The “process” can be further subdivided, and the closer the process is to the underlying level, the more likely it is to be shared by multiple “tasks”. "Process" is associated with "algorithm model" or "software tool". "Algorithm model" includes exponential methods, machine learning models, deep learning models, wildfire spread models, and power generation calculation formulas, etc. "Software tool" refers to geographic information system software or software for specific purposes. The "algorithm model" and "software tools" are associated with "parameters," "remote sensing images," "bands," and "general data." "Parameters" are the input parameters of the algorithm model; "remote sensing images" refer to Google imagery, high-resolution imagery series, Sentinel imagery, UAV imagery, LiDAR maps, etc.; "bands" refer to red bands, blue bands, near-infrared bands, etc.; and "general data" refers to population data, building vector data, road vector data, solar radiation data, GDP data, DEM data, DSM data, etc. These data have attributes such as spatial resolution and coverage.
[0036] For ease of understanding, please refer to Figure 2 As shown, Figure 2 This embodiment provides a relational diagram of an ontology framework, which illustrates the main entity types and their relationships within the ontology framework.
[0037] It should be understood that entity types are the basic units in a knowledge graph, used to represent different concepts or objects. For the storage of entity types, each row is a list containing two elements: the first element represents the specified entity type, and the second element is a sublist containing examples of that entity type. Relationships represent connections between entities. For the storage of relationships, each row is a list containing at least three elements: the first and third elements are the specified entity types, the second element is the possible relationships between these two types of entities, and the remainder are the relationship's attributes. Attributes are specific characteristics of an entity. For the storage of attributes, each row is a list containing two elements: the first element is the attribute, and the second element is the entity types that may possess that attribute.
[0038] This step begins by collecting relevant academic papers related to remote sensing data, which serve as the primary data source for constructing the knowledge graph. Next, key entities, relationships between entities, and attributes of entities or relationships are extracted from these papers. Furthermore, for entities of the "region" type, in addition to the regional information from the papers, it can be supplemented and improved from multi-source geographic data. Combining this multi-source geographic data with the regional information extracted from the papers creates a more comprehensive attribute system for "region" type entities. Finally, the extracted entities, relationships, and attributes are organized according to a pre-defined ontology framework to form a structured aerospace remote sensing atlas. This atlas can be stored in a graph database for easy subsequent querying and analysis.
[0039] Step 130: Obtain the query template matching the at least one keyword, fill the first target node into the query template, and generate a query statement. In this embodiment, different query templates are designed for different keyword types in order to achieve accurate retrieval of graph databases.
[0040] In this step, the appropriate query template is selected based on the identified keyword type. The identified keyword values (such as task name, region name, etc.) are then used as parameters to fill the query template to generate the query statement.
[0041] Step 140: Obtain contextual data related to the query statement from the aerospace remote sensing map, and input the contextual data and the query question as prompt information into the generative big data model to obtain the knowledge answer output by the generative big data model.
[0042] It should be understood that the main function of a generative large model is to generate new, coherent text content based on input prompts. The generative large model used in this embodiment is consistent with that in the prior art, and there is no limitation thereon.
[0043] In this step, the query is first executed, and relevant data retrieved from the aerospace remote sensing maps in the graph database is used as contextual data. This data is retrieved from the aerospace remote sensing maps in the graph database based on the keywords and query template entered by the user. Next, the retrieved contextual data is merged with the user's original query to form a complete prompt. Finally, the merged prompt is input into the generative big data model, which uses its language generation capabilities to generate a reasonable knowledge answer based on the prompt.
[0044] The knowledge recommendation method based on aerospace remote sensing maps provided in this invention rewrites and breaks down query questions using question templates, extracting keywords to accurately pinpoint the core of the user's question. Secondly, the construction of the aerospace remote sensing map integrates entities, relationships, attributes from remote sensing papers, and multi-source geographic data, allowing previously overlapping, fragmented, and redundant knowledge to be presented in a structured manner, facilitating the rapid and accurate identification of the first target node matching the keywords. Finally, the first target node is filled into the query template to generate a query statement, further clarifying the query direction. The contextual data related to the query statement obtained from the aerospace remote sensing map provides rich background knowledge for the generative large-scale model, enabling it to output accurate answers that conform to the characteristics of remote sensing knowledge, thus achieving efficient recommendation of remote sensing knowledge.
[0045] It should be noted that each implementation method of this application can be freely combined, rearranged, or executed individually, and does not need to rely on or depend on a fixed execution order.
[0046] In some embodiments, the aerospace remote sensing map is constructed in the following manner: Acquire multiple papers related to remote sensing data and standardize the text format of the papers; Each paper with a uniform text format and the first prompt information are input into the large language model to obtain the paper summary corresponding to each paper output by the large language model; the first prompt information includes the preset ontology framework and the remote sensing application methodology logic chain. The summary of each paper and the second prompt information are input into the large language model to obtain the entities, relations and attributes in each paper output by the large language model; the second prompt information includes the preset ontology framework and structured knowledge extraction prompt information. For regional entities within an entity, supplementary attributes of the regional entity are extracted by combining multi-source geographic data; Based on the entities, relationships, attributes, and supplementary attributes of the entities in each paper, an aerospace remote sensing map is constructed.
[0047] It should be understood that a large language model is a model built based on deep learning technology, primarily used for understanding and generating natural language text. The large language model used in this embodiment is consistent with existing technologies and is not limited thereto.
[0048] In this embodiment, papers related to remote sensing data are collected as an external knowledge base. These papers are then converted into a concise and lightweight unified text format to eliminate formatting differences between multiple texts. Furthermore, to improve efficiency, unnecessary sections such as literature reviews, acknowledgments, and references can be removed.
[0049] Since academic papers are typically lengthy and densely packed with knowledge, this embodiment first designs prompts based on the logical chain of remote sensing application methodologies. Additionally, to preserve as much detail as possible, a pre-defined ontology framework is also input into the large language model as a prompt. The large language model generates a paper summary for each paper according to the remote sensing application methodological logical chain and ontology framework, including the title, authors, research background, research topic, research scope and characteristics, specific process and algorithm model, software tools and input data, and research conclusions.
[0050] Here, the remote sensing application methodology logic chain refers to the logical steps from defining the research task to its specific implementation in remote sensing application research. This logic chain describes how to start from a research objective, gradually break it down into specific implementation steps, select appropriate methods and tools, and the required data resources, ultimately completing the research task. In one example, the remote sensing application methodology logic chain is "task - process - algorithm model / software tool - data".
[0051] After extracting the paper summary of each paper based on the first prompt information, the system continues to extract prompt information and preset ontology framework based on the designed structured knowledge extraction, and extracts the entities, relationships and attributes of each paper from the paper summary.
[0052] Here, structured knowledge extraction prompts refer to a series of guiding messages designed to instruct the model to generate output that conforms to specific formats and requirements when using large language models for text processing. These messages help the model understand the specific requirements of the task and generate structured output. In one example, structured knowledge extraction prompts include, but are not limited to: Overall task: Define the main tasks that the model needs to accomplish; Specific requirements: Describe the specific requirements of the task in detail, including the types of entities, relationships, and attributes to be extracted; Output format: Specifies the output format generated by the model to ensure the structure and consistency of the output information; Example: Provide specific examples to help the model understand the specific requirements of the task and the output format.
[0053] Furthermore, as a fundamental geographic entity, the acquisition of attribute information for a "region" extends beyond the regional introduction in the academic literature of a paper. The conclusion of the paper may also include other attributes of the "region," such as calculation results of certain indicators or findings and recommendations. These are the most direct knowledge generated by the paper and should be mined as attributes of the region. Similarly, it is important to distinguish whether these attributes apply to the entire experimental area or a specific region. In addition, various external auxiliary geographic data can be used to supplement and improve the information. Administrative division data can provide basic geographic information such as precise spatial extent, area data, and inclusion relationships between administrative divisions at each level. Furthermore, natural environmental elements and socioeconomic indicators based on raster data can be linked to corresponding regional units through spatial aggregation methods, thereby constructing a comprehensive and accurate multi-dimensional attribute feature system for the "region" entity. This multi-source geographic data fusion method can improve the completeness and accuracy of the description of regional entities.
[0054] Therefore, in this embodiment, after extracting the entities, relationships and attributes of each paper from the paper summary based on the second prompt information, it will also combine multi-source geographic data to extract the supplementary attributes of regional entities and add these supplementary attributes to the attributes of regional entities.
[0055] Specifically, all extracted entities, relationships, attributes, and supplementary attributes of regional entities are constructed into an aerospace remote sensing map and stored in a graph database.
[0056] The knowledge recommendation method based on aerospace remote sensing atlas provided in this invention collects papers related to remote sensing data as an external knowledge base; then, it calls a large language model, uses prompt word technology, and generates a paper summary for each paper according to a prescribed ontology framework, extracts entities, relationships and attributes from it, and supplements regional attributes from multi-source geographic data, so that the originally overlapping, scattered and redundant knowledge can be presented in a structured way, and constructs an aerospace remote sensing atlas.
[0057] In some embodiments, the construction of the aerospace remote sensing atlas based on the entities, relationships, and attributes in each paper includes: For each entity in the paper, create entity-related labels and names; Based on entity-related labels and names, determine whether the current graph database contains the same second target node; If the second target node does not exist in the current graph database, create a new node corresponding to the entity; If the second target node exists in the current graph database, determine whether the second target node has an attribute corresponding to the entity; If the second target node does not have an attribute corresponding to the entity, assign the attribute corresponding to the entity to the second target node; If the second target node has attributes corresponding to the entity, the attribute values are merged. For each relation in the paper, determine the head node and tail node corresponding to the relation, establish the edge between the head node and the tail node, and assign the attribute corresponding to the relation to the edge between the head node and the tail node.
[0058] In aerospace remote sensing mapping, nodes are the basic building blocks of a graph, used to represent entities. Each node typically has two key attributes: a label and a name. The label indicates the entity type to which the node belongs, and the name indicates the node's specific content or name. Under the same label, the name is used to distinguish different nodes. For example, the titles of different papers, or the names of different authors.
[0059] When creating a node, it's necessary to determine if a node with the same label and name already exists in the graph database. Specifically, for each entity, first, create the entity's associated label and name. Next, check if a second target node with the same label and name already exists in the graph database. If not, create a new node directly. If a second target node with the same label and name already exists, check if the second target node has the attribute corresponding to the current entity. If some attributes are missing, assign the missing attributes to the second target node. If the attribute already exists, merge the new attribute value with the old attribute value and store the merged value as a list.
[0060] When creating an edge, the head node and tail node are read first to confirm whether the head node and tail node corresponding to the current relationship are already in the graph database. If the head node and tail node corresponding to the current relationship are already in the graph database, the edge between the two is created and the edge attributes are also entered into the database.
[0061] Furthermore, each "paper" is a core node, containing specific research content, methods, and results. The "task" is the research topic of the "paper," also a core node, representing a specific problem to be solved or a research direction. Using "papers" and "tasks" as core nodes, each "paper" generates a subgraph. This subgraph contains all information related to that "paper," such as authors, region, methods, and data. A "task" node may connect to multiple "paper" nodes, meaning multiple "papers" may research the same "task," forming a multi-faceted exploration of that "task." Each "paper" may contain multiple "process" nodes, representing the research methods and steps described in the "paper." When multiple "papers" share the same "task" node, that "task" node will connect to "process" nodes from multiple "papers." This may lead to confusion among "process" nodes, making it difficult to distinguish which "processes" belong to which "paper."
[0062] Based on this, this embodiment also includes two key attributes: source and sequence number. The source attribute identifies which paper the node belongs to. This clarifies the affiliation of each node and avoids confusion. The sequence number identifies the order of the node within the paper. For example, for the "process" node, the source attribute clarifies which paper it originates from, while the sequence number clarifies its order within the paper, facilitating understanding the research methods and steps outlined in the paper.
[0063] In some embodiments, determining the first target node matching the at least one keyword in the aerospace remote sensing map includes: For each keyword, based on the embedding similarity between the first embedding vector of the keyword and the second embedding vector of each first node in the aerospace remote sensing map, a first target node matching the keyword is determined; the first node is a node with the same entity type as the keyword. Specifically, when the entity type corresponding to the keyword is a task, the first target node for keyword matching is determined based on the embedding similarity between the first embedding vector of the keyword and the second embedding vector of each second node in the aerospace remote sensing map; the second node includes task nodes and process nodes.
[0064] It should be understood that this embodiment defines five keyword types for extracting keywords from query questions: Task: The main objective or problem of the research; Algorithm model or software tool: A specific method or tool used to solve a problem; Data: Data resources used in the research.
[0065] Region + Task: Research tasks within a specific region; Region + Data: Data resources within a specific region.
[0066] Due to the diversity of terminology used in academic papers, phrases with the same meaning often have multiple expressions, which may lead to the identified keywords not perfectly matching the nodes in the graph database. Therefore, this embodiment uses keyword matching based on embedding similarity.
[0067] Specifically, the first embedding vector of the identified keywords is calculated, and a second embedding vector is pre-generated for each node in the graph database. Then, by calculating the embedding similarity between the embedding vectors, the first target node that is semantically closest to the identified keywords is found.
[0068] Furthermore, when the entity type corresponding to the keyword is a task, since "process" nodes can be regarded as a low-level decomposition of "task" nodes, they are the specific steps to implement the "task". Therefore, not only are "task" nodes that directly match the "task" searched, but also "process" nodes related to the "task" are considered.
[0069] The knowledge recommendation method based on aerospace remote sensing maps provided in this invention accurately identifies the first target node matching the keyword by embedding similarity. Furthermore, when the entity type corresponding to the keyword is a task, it not only searches for "task" nodes that directly match the "task," but also considers "process" nodes related to the "task," thus improving the comprehensiveness of the search results.
[0070] In some embodiments, obtaining the query template matching the at least one keyword includes: If the at least one keyword is a single keyword and there is no parent keyword for the at least one keyword, obtain the first query template matching the at least one keyword; If the at least one keyword is a single keyword and the at least one keyword has a parent keyword, obtain a second query template matching the at least one keyword and a third query template matching the parent keyword of the at least one keyword.
[0071] Here, a single keyword refers to a query task that uses only one independent keyword, which is not used in combination with other keywords, nor does it rely on other keywords to determine its specific meaning and query scope. It is an independent and clearly defined query object used to directly locate related content in the database. Examples include: task, algorithm model, software tool, and data.
[0072] In a knowledge graph, the parent keyword refers to the broader category to which the current keyword belongs. It's typically used to describe the attribution relationship of the current keyword, helping to more accurately locate and understand its context. For example, keywords like "algorithm model" and "software tool" need to be derived from higher-level processes and tasks; therefore, the parent keywords for these keywords would be "process" and "task." Similarly, keywords like "data" need to be derived from higher-level algorithm models and software tools; therefore, the parent keywords for these keywords would be "algorithm model" and "software tool."
[0073] In one example, given task keywords, a first query template can be constructed to query the process, data, algorithm model, and software tools required to implement the task. For instance, the query steps designed in the first query template are as follows: Find the task node named "M"; Identify all process nodes directly contained in this task; Recursively identify all subprocesses within these processes; Merge all processes and sub-processes, remove duplicates, and then proceed with the processing. For each process, identify the tools (algorithm model or software tool) required. For each tool, identify the data it requires as input; The returned results include: task name, process name, tool name, tool type (algorithm model or software tool), and data used by the tool.
[0074] For keywords like "algorithm model" and "software tool," a second query template can be constructed to search for descriptions of the algorithm model and software tool's uses. For example, the query steps designed in the second query template are as follows: Find the tool node named "N algorithm" (type: algorithm model or software tool). Returns directly to the "Usage Description" property of the tool node.
[0075] For keywords related to algorithm models and software tools, a third query template is also constructed to query the processes and tasks to which they belong. For example, the query steps designed in the third query template are as follows: Find the tool node named "X"; Identify all process nodes that use this tool; Identify the task nodes to which these processes belong; The returned results include: process name and task name.
[0076] For data keywords, a second query template can be constructed to query attributes such as the source or coverage of the data. For example, the query steps designed in the second query template are as follows: Find the data node named "S"; It directly returns the "data source" attribute of the data node.
[0077] For data keywords, a third-party query template can be built to determine which tools generate or use the data. For example, the query steps designed in the third-party query template are as follows: Find the data node named "S"; Find all tool nodes (algorithm models or software tools) that can output this data. The returned results include: tool name and tool type.
[0078] The knowledge recommendation method based on aerospace remote sensing maps provided in this invention improves the accuracy and comprehensiveness of query results by determining whether a single keyword has a parent keyword and flexibly obtaining the corresponding query template for that keyword.
[0079] In some embodiments, obtaining the query template matching the at least one keyword includes: When at least one keyword is a first keyword pair, a fourth query template, a fifth query template, and a sixth query template matching the at least one keyword pair are obtained; the first keyword pair includes a target region and a task; the fourth query template is used to query papers that match the target region and the task; the fifth query template is used to query solutions in the papers; and the sixth query template is used to find solutions in papers from regions similar to the target region when no papers belonging to the target region exist. When at least one keyword is a second keyword pair, obtain the seventh, eighth, and ninth query templates matching the at least one keyword pair; the second keyword pair includes a target region and data, the seventh query template is used to query numerical data in the target region, the eighth query template is used to query whether the spatial range corresponding to the spatial data includes the target region, and the ninth query template is used to query whether there is an inclusion relationship between regions.
[0080] Here, a keyword pair refers to a query unit composed of two related keywords in a query task, used to more accurately locate and retrieve information. In this embodiment, keyword pairs are divided into two types: keyword pairs between the target area and the task, and keyword pairs between the target area and the data.
[0081] The keyword pairs for target area and task indicate that a solution for carrying out a certain task in the specified target area is to be provided. Therefore, this embodiment designs a fourth, fifth, and sixth query template.
[0082] The fourth query template is used to search for existing research papers on a specific task within a target region. For example, the query steps designed in the fourth query template are as follows: Find paper nodes whose research scope is "region T" and whose research topic is "P".
[0083] The returned results include: paper title, research region, and research topic.
[0084] The fifth query template is used to search for solutions in papers, specifically the processes, data, algorithm models, and software tools used to implement the task. This fifth query template is the same as the first query template mentioned above, and will not be described in detail here.
[0085] The sixth query template is used for similar region substitution; that is, when there are no papers directly researching this task in the target region, it recommends research solutions from papers in regions similar to the target region. For example, the query steps designed in the fourth query template are as follows: Find the task node named "M"; Find all the paper nodes that research this task; Identify the research area nodes in these papers; The returned results include: research topic, paper title, and research area.
[0086] Based on this, the research areas retrieved using the fourth query template can be used to recommend research solutions for this task from papers in regions similar to the target region.
[0087] Here, regional similarity includes, but is not limited to: similar area, similar or identical administrative level, similar geographical location, and similar environment.
[0088] The keyword pair for target area and data indicates that specific data from the designated target area is required. It should be understood that data is categorized into numerical data and spatial data. Numerical data refers to values of indicators such as building roof area or photovoltaic power generation potential, while spatial data refers to data with spatial coverage, such as remote sensing imagery or solar radiation.
[0089] Based on this, different query templates are designed for different types of data in this embodiment. Specifically, for numerical data, the seventh query template is used, which is used to query numerical data in the target area. For example, the query steps designed in the seventh query template are as follows: Find the region node named "T region"; Returns the name of the region, whether the "G" attribute exists, and the value of that attribute.
[0090] For spatial data, use the eighth query template. The eighth query template is used to check whether the spatial range corresponding to the spatial data includes the target region. For example, the query steps designed in the seventh query template are as follows: Find the data node named "S".
[0091] Returns the name and coverage attributes of the data.
[0092] Furthermore, this embodiment also includes an eighth query template for querying whether an inclusion relationship exists between regions. For example, the query steps designed in the eighth query template are as follows: Find if an inclusion path exists from "region T" to "region Y"; The returned results include: whether an inclusion relationship exists, and the hierarchy depth.
[0093] The knowledge recommendation method based on aerospace remote sensing maps provided in this embodiment of the invention, for keyword pairs, achieves multi-level association queries on keyword pairs through the multi-level query template designed above, thereby improving the accuracy of query results.
[0094] The knowledge recommendation device based on aerospace remote sensing maps provided by the present invention is described below. The knowledge recommendation device based on aerospace remote sensing maps described below and the knowledge recommendation method based on aerospace remote sensing maps described above can be referred to in correspondence.
[0095] The knowledge recommendation device based on aerospace remote sensing maps in this invention embodiment, such as... Figure 3 As shown, it includes the following modules: The first knowledge recommendation module 310 is used to respond to the query question input by the user, rewrite and split the query question according to the question template, and extract at least one keyword from the rewritten and split query question. The second knowledge recommendation module 320 is used to determine the first target node that matches the at least one keyword in the aerospace remote sensing map; the aerospace remote sensing map is a knowledge graph constructed according to a preset ontology framework, based on entities, relationships and attributes extracted from papers related to remote sensing data, combined with supplementary attributes of regional entities extracted from multi-source geographic data. The third knowledge recommendation module 330 is used to obtain a query template that matches the at least one keyword, fill the first target node into the query template, and generate a query statement; The fourth knowledge recommendation module 340 is used to obtain contextual data related to the query statement from the aerospace remote sensing map, and input the contextual data and the query question as prompt information into the generative big model to obtain the knowledge answer output by the generative big model.
[0096] This embodiment of the knowledge recommendation device based on aerospace remote sensing maps rewrites and breaks down query questions using question templates, extracting keywords to accurately pinpoint the core of the user's question. Secondly, the construction of the aerospace remote sensing map integrates entities, relationships, attributes from remote sensing papers, and multi-source geographic data, allowing previously overlapping, fragmented, and redundant knowledge to be presented in a structured manner, facilitating the rapid and accurate identification of the first target node matching the keywords. Finally, the first target node is filled into the query template to generate a query statement, further clarifying the query direction. The contextual data related to the query statement obtained from the aerospace remote sensing map provides rich background knowledge for the generative large-scale model, enabling it to output accurate answers that conform to the characteristics of remote sensing knowledge, thus achieving efficient recommendation of remote sensing knowledge.
[0097] Figure 4 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 4As shown, the electronic device may include: a processor 410, a communications interface 420, a memory 430, and a communication bus 440, wherein the processor 410, the communications interface 420, and the memory 430 communicate with each other via the communication bus 440. The processor 410 can call logical instructions in the memory 430 to execute a knowledge recommendation method based on aerospace remote sensing maps, the method including: In response to a user's input query question, the query question is rewritten and split according to the question template, and at least one keyword is extracted from the rewritten and split query question; In the aerospace remote sensing map, a first target node matching the at least one keyword is determined; the aerospace remote sensing map is a knowledge graph constructed based on entities, relationships and attributes extracted from papers related to remote sensing data, combined with supplementary attributes of regional entities extracted from multi-source geographic data, and constructed according to a preset ontology framework. Obtain a query template that matches the at least one keyword, fill the first target node into the query template, and generate a query statement; Contextual data related to the query statement is obtained from the aerospace remote sensing map, and the contextual data and the query question are input as prompt information into the generative big model to obtain the knowledge answer output by the generative big model.
[0098] Furthermore, the logical instructions in the aforementioned memory 430 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, etc., each of which can store program code.
[0099] On the other hand, the present invention also provides a computer program product, the computer program product comprising a computer program that can be stored on a non-transitory computer-readable storage medium, wherein when the computer program is executed by a processor, the computer is capable of executing the knowledge recommendation method based on aerospace remote sensing maps provided by each of the above methods, the method comprising: In response to a user's input query question, the query question is rewritten and split according to the question template, and at least one keyword is extracted from the rewritten and split query question; In the aerospace remote sensing map, a first target node matching the at least one keyword is determined; the aerospace remote sensing map is a knowledge graph constructed based on entities, relationships and attributes extracted from papers related to remote sensing data, combined with supplementary attributes of regional entities extracted from multi-source geographic data, and constructed according to a preset ontology framework. Obtain a query template that matches the at least one keyword, fill the first target node into the query template, and generate a query statement; Contextual data related to the query statement is obtained from the aerospace remote sensing map, and the contextual data and the query question are input as prompt information into the generative big model to obtain the knowledge answer output by the generative big model.
[0100] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to perform the knowledge recommendation method based on aerospace remote sensing maps provided by each of the above methods, the method comprising: In response to a user's input query question, the query question is rewritten and split according to the question template, and at least one keyword is extracted from the rewritten and split query question; In the aerospace remote sensing map, a first target node matching the at least one keyword is determined; the aerospace remote sensing map is a knowledge graph constructed based on entities, relationships and attributes extracted from papers related to remote sensing data, combined with supplementary attributes of regional entities extracted from multi-source geographic data, and constructed according to a preset ontology framework. Obtain a query template that matches the at least one keyword, fill the first target node into the query template, and generate a query statement; Contextual data related to the query statement is obtained from the aerospace remote sensing map, and the contextual data and the query question are input as prompt information into the generative big model to obtain the knowledge answer output by the generative big model.
[0101] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.
[0102] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.
[0103] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in each of the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of each embodiment of the present invention.
Claims
1. A knowledge recommendation method based on space remote sensing atlas, characterized in that, include: In response to a user's input query question, the query question is rewritten and split according to the question template, and at least one keyword is extracted from the rewritten and split query question; In the aerospace remote sensing map, identify a first target node that matches at least one keyword; The aforementioned aerospace remote sensing atlas is a knowledge graph constructed based on entities, relationships, and attributes extracted from papers related to remote sensing data, combined with supplementary attributes of regional entities extracted from multi-source geographic data, and constructed according to a preset ontology framework. The pre-defined ontology framework refers to a formal framework used to define and standardize concepts, entities, relationships, and attributes within a domain when constructing a knowledge graph. The pre-defined ontology framework is designed as follows: Papers are identified as the first core entity type, and associated authors and regions are created. Tasks are identified as the second core entity type, and associated tasks are created, along with associated processes, data processing tools, and parameters. The data processing tools include at least one of algorithmic models or software tools. Obtain a query template that matches the at least one keyword, fill the first target node into the query template, and generate a query statement; Contextual data related to the query statement is obtained from the aerospace remote sensing map, and the contextual data and the query question are input as prompt information into the generative big model to obtain the knowledge answer output by the generative big model; The aerospace remote sensing map was constructed in the following way: Acquire multiple papers related to remote sensing data and standardize the text format of the papers; Each paper with a uniform text format and the first prompt information are input into the large language model to obtain the paper summary corresponding to each paper output by the large language model; the first prompt information includes the preset ontology framework and the remote sensing application methodology logic chain. The summary of each paper and the second prompt information are input into the large language model to obtain the entities, relations and attributes in each paper output by the large language model; the second prompt information includes the preset ontology framework and structured knowledge extraction prompt information. For regional entities within an entity, supplementary attributes of the regional entity are extracted by combining multi-source geographic data; Based on the entities, relationships, attributes, and supplementary attributes of the entities in each paper, an aerospace remote sensing map is constructed. 2.The method of claim 1, wherein, The step of determining the first target node in the aerospace remote sensing map that matches the at least one keyword includes: For each keyword, based on the embedding similarity between the first embedding vector of the keyword and the second embedding vector of each first node in the aerospace remote sensing map, a first target node matching the keyword is determined; the first node is a node with the same entity type as the keyword. Specifically, when the entity type corresponding to the keyword is a task, the first target node for keyword matching is determined based on the embedding similarity between the first embedding vector of the keyword and the second embedding vector of each second node in the aerospace remote sensing map; the second node includes task nodes and process nodes. 3.The method of claim 1, wherein, Obtaining a query template that matches at least one keyword includes: If the at least one keyword is a single keyword and there is no parent keyword for the at least one keyword, obtain the first query template matching the at least one keyword; If the at least one keyword is a single keyword and the at least one keyword has a parent keyword, obtain a second query template matching the at least one keyword and a third query template matching the parent keyword of the at least one keyword. 4.The method of claim 1, wherein, Obtaining a query template that matches at least one keyword includes: When at least one keyword is a first keyword pair, a fourth query template, a fifth query template, and a sixth query template matching the at least one keyword pair are obtained; the first keyword pair includes a target region and a task; the fourth query template is used to query papers that match the target region and the task; the fifth query template is used to query solutions in the papers; and the sixth query template is used to find solutions in papers from regions similar to the target region when no papers belonging to the target region exist. When at least one keyword is a second keyword pair, obtain the seventh, eighth, and ninth query templates matching the at least one keyword pair; the second keyword pair includes a target region and data, the seventh query template is used to query numerical data in the target region, the eighth query template is used to query whether the spatial range corresponding to the spatial data includes the target region, and the ninth query template is used to query whether there is an inclusion relationship between regions.
5. The knowledge recommendation method based on aerospace remote sensing maps according to claim 1, characterized in that, The method further includes: For each entity in the paper, create entity-related labels and names; Based on entity-related labels and names, determine whether the current graph database contains the same second target node; If the second target node does not exist in the current graph database, create a new node corresponding to the entity; If the second target node exists in the current graph database, determine whether the second target node has an attribute corresponding to the entity; If the second target node does not have an attribute corresponding to the entity, assign the attribute corresponding to the entity to the second target node; If the second target node has attributes corresponding to the entity, the attribute values are merged. For each relation in the paper, determine the head node and tail node corresponding to the relation, establish the edge between the head node and the tail node, and assign the attribute corresponding to the relation to the edge between the head node and the tail node.
6. A knowledge recommendation device based on aerospace remote sensing maps, characterized in that, include: The first knowledge recommendation module is used to respond to the query questions input by the user, rewrite and split the query questions according to the question template, and extract at least one keyword from the rewritten and split query questions; The second knowledge recommendation module is used to determine a first target node in the aerospace remote sensing map that matches the at least one keyword; The aforementioned aerospace remote sensing atlas is a knowledge graph constructed based on entities, relationships, and attributes extracted from papers related to remote sensing data, combined with supplementary attributes of regional entities extracted from multi-source geographic data, and constructed according to a preset ontology framework. The pre-defined ontology framework refers to a formal framework used to define and standardize concepts, entities, relationships, and attributes within a domain when constructing a knowledge graph. The pre-defined ontology framework is designed as follows: Papers are identified as the first core entity type, and associated authors and regions are created. Tasks are identified as the second core entity type, and associated tasks are created, along with associated processes, data processing tools, and parameters. The data processing tools include at least one of algorithmic models or software tools. The third knowledge recommendation module is used to obtain a query template that matches the at least one keyword, fill the first target node into the query template, and generate a query statement. The fourth knowledge recommendation module is used to obtain contextual data related to the query statement from the aerospace remote sensing map, and input the contextual data and the query question as prompt information into the generative big model to obtain the knowledge answer output by the generative big model; The aerospace remote sensing map was constructed in the following way: Acquire multiple papers related to remote sensing data and standardize the text format of the papers; Each paper with a uniform text format and the first prompt information are input into the large language model to obtain the paper summary corresponding to each paper output by the large language model; the first prompt information includes the preset ontology framework and the remote sensing application methodology logic chain. The summary of each paper and the second prompt information are input into the large language model to obtain the entities, relations and attributes in each paper output by the large language model; the second prompt information includes the preset ontology framework and structured knowledge extraction prompt information. For regional entities within an entity, supplementary attributes of the regional entity are extracted by combining multi-source geographic data; Based on the entities, relationships, attributes, and supplementary attributes of the entities in each paper, an aerospace remote sensing map is constructed.
7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the knowledge recommendation method based on aerospace remote sensing maps as described in any one of claims 1 to 5.
8. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the knowledge recommendation method based on aerospace remote sensing maps as described in any one of claims 1 to 5.