Water conservancy knowledge base management method and device based on large language model and medium

Abstract

The invention discloses a water conservancy knowledge base management method and device based on a large language model and a medium, and relates to the technical field of water conservancy management. The method comprises the following steps: collecting multi-source water conservancy data, and preprocessing the multi-source water conservancy data to obtain standard water conservancy data; analyzing the standard water conservancy data based on a preset target large language model, and converting the unstructured text into a structured text; integrating the structured text, the structured table and the image data to obtain a target water conservancy data knowledge base; and when a data query instruction is received, based on query information contained in the data query instruction, performing similarity matching in the target water conservancy data knowledge base according to a preset query rule to generate a water conservancy plan. By means of the method, intelligent water conservancy plan matching and automatic query are achieved, and the water conservancy knowledge base management efficiency is improved.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of water conservancy management, and in particular to a water conservancy knowledge base management method based on a large language model, equipment and a medium. BACKGROUND

[0002] In today's era, the water conservancy industry is accelerating the digital transformation, but the water conservancy knowledge management still has the following problems: The data generated by the water conservancy industry is diverse, including structured monitoring data (such as real-time monitoring values of water level, flow, etc.), unstructured text data (such as historical documents, work reports, etc.), data with time series characteristics (such as long-term hydrological observation records), and water system topology in the form of graph data. These different types of data have great differences in format, semantics and processing methods. Due to the lack of effective integration mechanism and technical means, the traditional water conservancy data processing system is difficult to uniformly process and analyze these heterogeneous data, making it difficult to fully exploit and utilize the relevance and overall value of the data.

[0003] In addition, the current water conservancy knowledge management tool mostly relies on simple keyword matching technology, lacking semantic understanding ability, which makes it difficult to accurately understand the real intention of the user when querying knowledge, and it is more difficult to realize the correlation analysis of multi-source data. For example, it is difficult to intelligently match the complex relationship between rainfall characteristics and historical flood disasters, so as to provide comprehensive, accurate and in-depth reference for water conservancy decision-making.

[0004] Therefore, for the water conservancy knowledge management tool, how to realize the automatic arrangement of data and provide an intelligent query scheme has become a technical problem to be solved. SUMMARY

[0005] The embodiments of the application provide a water conservancy knowledge base management method based on a large language model, equipment and a medium, to solve the technical problem of how to realize the automatic arrangement of data and provide an intelligent query scheme for the water conservancy knowledge management tool.

[0006] In the first aspect, the embodiments of the application provide a water conservancy knowledge base management method based on a large language model, which comprises: collecting multi-source water conservancy data, and pre-processing multi-dimensional water conservancy data to obtain standard water conservancy data; analyzing the standard water conservancy data based on a preset target large language model, and converting unstructured text into structured text; integrating the structured text, structured table and image data to obtain a target water conservancy data knowledge base; when receiving a data query instruction, performing similarity matching in the target water conservancy data knowledge base according to the query information contained in the data query instruction according to a preset query rule, to generate a water conservancy plan.

[0007] In an implementation form of the present application, a large language model is constructed, specifically comprising: collecting a water conservancy text dataset, and defining natural language parsing rules according to the water conservancy text dataset; wherein the water conservancy text dataset includes hydrological entities, disaster events, and disposal measures; the natural language parsing rules include entity extraction templates, attribute mapping relationships, and triple construction specifications; training the large language model based on the natural language parsing rule set until a preset recognition accuracy is reached, to obtain a target large language model.

[0008] In an implementation form of the present application, based on the preset target large language model, unstructured text is converted into structured text, specifically comprising: inputting preprocessed unstructured text data into the target large language model; extracting water conservancy entities, entity attributes, and entity relationships in the unstructured text data through the natural language parsing rule set; generating structured text data in triple format according to the extraction results.

[0009] In an implementation form of the present application, structured text, structured table and image data are integrated to obtain a target water conservancy data knowledge base, specifically comprising: parsing the field metadata of the structured table, and performing same-name matching between the field metadata of the table and the triple entities generated by the structured text to obtain an attribute mapping relationship; identifying spatial coordinate information in the image data, and matching the spatial coordinate information with the triple entities generated by the structured text to obtain a spatial position binding relationship; based on the attribute mapping relationship and the spatial position binding relationship, fusing the standard water conservancy data to obtain a water conservancy knowledge graph; based on the water conservancy knowledge graph, constructing the target water conservancy data knowledge base.

[0010] In an implementation form of the present application, based on the water conservancy knowledge graph, the target water conservancy data knowledge base is constructed, specifically comprising: extracting water conservancy event data with time stamps in the water conservancy knowledge graph, and establishing a space-time dimension index node according to the spatial position binding relationship; importing the space-time index node into a Neo4j graph database to perform relationship topology storage, to obtain the target water conservancy data knowledge base.

[0011] In an implementation form of the present application, based on the query information contained in the data query instruction, similarity matching is performed in the target water conservancy data knowledge base according to a preset query rule, to generate a target water conservancy plan, specifically comprising: performing semantic analysis on the query information to generate a digital feature vector containing a query intent; calculating the cosine similarity of the digital feature vector and each historical vector in the target water conservancy data knowledge base; extracting water conservancy events in the target water conservancy data knowledge base with a similarity score greater than a preset threshold, and integrating them into the target water conservancy plan.

[0012] In an implementation form of the present application, after the target water conservancy plan is generated, the method further comprises: collecting user feedback data on the target water conservancy plan, and generating a feedback score and a treatment effect index; based on the feedback score, adjusting the recommendation weight value of the corresponding water conservancy event in the target water conservancy data knowledge base, and updating the query rule.

[0013] In an implementation form of the present application, multi-source water conservancy data is collected, and multi-dimensional water conservancy data is preprocessed to obtain standard water conservancy data, specifically including: extracting unstructured text and structured table keywords, and performing deduplication to obtain standard unstructured text data; extracting table header information and data content of the structured table, and performing data type conversion to obtain standard structured table data; using an image enhancement algorithm to perform brightness, contrast and resolution processing on image data to obtain standard image data; integrating the standard unstructured text data, the standard structured table data and the standard image data to obtain the standard water conservancy data.

[0014] In a second aspect, the embodiments of the present application also provide a water conservancy knowledge base management device based on a large language model, the device comprising: at least one processor; and a memory in communication connection with the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to: collect multi-source water conservancy data, and preprocess multi-dimensional water conservancy data to obtain standard water conservancy data; based on a preset target large language model, analyze the standard water conservancy data, and convert unstructured text into structured text; integrate the structured text, structured table and image data to obtain a target water conservancy data knowledge base; when a data query instruction is received, based on query information contained in the data query instruction, perform similarity matching in the target water conservancy data knowledge base according to a preset query rule to generate a water conservancy plan.

[0015] In a third aspect, the embodiments of the present application also provide a non-volatile computer storage medium for water conservancy knowledge base management based on a large language model, which stores computer executable instructions, and the computer executable instructions are configured to: collect multi-source water conservancy data, and preprocess multi-dimensional water conservancy data to obtain standard water conservancy data; based on a preset target large language model, analyze the standard water conservancy data, and convert unstructured text into structured text; integrate the structured text, structured table and image data to obtain a target water conservancy data knowledge base; when a data query instruction is received, based on query information contained in the data query instruction, perform similarity matching in the target water conservancy data knowledge base according to a preset query rule to generate a water conservancy plan.

[0016] The embodiments of the present application provide a water conservancy knowledge base management method, device and medium based on a large language model, which have the following beneficial effects: using a large language model to convert unstructured text into structured text, effectively improving the availability and processability of unstructured data, and facilitating better mining of the information value therein; integrating multi-source water conservancy data to construct a target water conservancy data knowledge base, and when receiving a data query instruction, being able to quickly perform similarity matching in the knowledge base based on the query information to generate a water conservancy plan, providing a strong basis for water conservancy decision-making, making the decision more scientific and reasonable, and being able to better respond to various water conservancy issues and challenges, improving the level and efficiency of water conservancy management, and ensuring the rational use of water resources and the safe operation of water conservancy facilities. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings: Figure 1 A flow chart of a water conservancy knowledge base management method based on a large language model provided in an embodiment of the present application; Figure 2 A schematic diagram of the internal structure of a water conservancy knowledge base management device based on a large language model provided in an embodiment of the present application. DETAILED DESCRIPTION

[0018] To make the purpose, technical solutions, and advantages of this application more clear, the technical solutions of this application will be clearly and completely described below in conjunction with the specific embodiments of this application and the corresponding drawings. Obviously, the embodiments described are only part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0019] The embodiments of the present application provide a water conservancy knowledge base management method, device and medium based on a large language model to solve the following technical problems: for water conservancy knowledge management tools, how to realize automatic data organization and provide an intelligent query solution.

[0020] The technical solutions proposed in the embodiments of the present application are described in detail below with reference to the accompanying drawings.

[0021] Figure 1 This is a flow chart of a method for managing a water conservancy knowledge base based on a large language model provided in an embodiment of the present application. Figure 1 As shown, the embodiment of the present application provides a water conservancy knowledge base management method based on a large language model, which specifically includes the following steps: Step 10, collect multi-source water conservancy data, and preprocess multi-dimensional water conservancy data to obtain standard water conservancy data.

[0022] As an optional embodiment, collecting multi-source water conservancy data and preprocessing multi-dimensional water conservancy data to obtain standard water conservancy data can specifically include: step 101, extracting unstructured text and structured table keywords, and removing duplicates to obtain standard unstructured text data; Step 102: Extracting table header information and data content of structured table and performing data type conversion to obtain standard structured table data.

[0023] In this step, the table header information usually represents the meaning or category of the data in each column of the table, and the data content is the specific information stored in the table. Since in the actual water conservancy scene, data may exist in various formats or types, such as text form of numbers, different date formats, etc., the purpose of data type conversion is to convert these data into data types that meet the preset standards, such as converting text type numbers to numerical type, adjusting dates to a unified date format, etc., so that the data in the table is more standardized and unified in type, facilitating subsequent data processing, analysis, storage and interaction with other operations, and ultimately obtaining standard structured table data.

[0024] Step 103: Using image enhancement algorithm to process brightness, contrast and resolution of image data to obtain standard image data.

[0025] In this step, the brightness, contrast and resolution of image data are processed using image enhancement algorithm, the over-bright or over-dark areas in the image are adjusted to make the brightness of the image more balanced, the contrast processing mainly enhances the difference between different regions in the image, making the object outline, details, etc. in the image more prominent, improving the clarity and recognizability of the image, and the resolution processing optimizes the adjustment of image pixels to make the image clearer and sharper based on the original details. After a series of image enhancement processing operations, the obtained image data is converted into standard image data, which can provide more accurate, clear and reliable visual information in subsequent storage, analysis, display and other links, meeting the requirements of image data quality in water conservancy scenarios.

[0026] Step 104: Integrating standard unstructured text data, standard structured table data and standard image data to obtain standard water conservancy data.

[0027] Step 20, based on the preset target large language model, analyzing the standard water conservancy data, and converting the unstructured text into structured text.

[0028] As an optional embodiment, the method further includes: step 201: constructing a large language model, specifically including: step 2011: collecting a water conservancy text dataset and defining natural language parsing rules based on the water conservancy text dataset; wherein the water conservancy text dataset includes hydrological entities, disaster events, and disposal measures; the natural language parsing rules include entity extraction templates, attribute mapping relationships, and triple construction specifications; In this step, a variety of text information covering the water conservancy field is collected to form a water conservancy text dataset. Key information items such as hydrological entities, disaster events, and disposal measures are accurately located and extracted from the water conservancy text, and the patterns and characteristics that may appear in these entities in the text are specified so that they can be identified and extracted from the text; the various attributes of the extracted entities (for example, a water conservancy facility entity may have attributes such as construction time and scale) are corresponded with the specific description content in the text to form an attribute mapping relationship, which can accurately map the description of entity attributes in the text to the corresponding attribute items; the triple construction specification specifies how to organize the extracted entities and their attributes, the relationships between entities, etc. into classic triple forms, so as to construct a structured representation that can clearly express the logical relationship of water conservancy knowledge, thereby facilitating subsequent query, analysis, and other operations on water conservancy text knowledge.

[0029] Step 2012: The large language model is trained based on the natural language parsing rule set until a preset recognition accuracy is reached to obtain a target large language model.

[0030] In this step, the large language model is trained using the constructed natural language parsing rule set. During the training process, water conservancy text data is continuously input into the large language model, and the model's recognition and extraction of key information such as hydrological entities, disaster events, and disposal measures in the text is evaluated based on the rule set. By continuously adjusting the model's parameters and structure, the model's recognition accuracy of key information is gradually improved. When the recognition accuracy of the large language model reaches the preset standard, the training process is considered complete. The large language model obtained at this time is called the target large language model, which has the ability to accurately understand and process water conservancy text data according to the established natural language parsing rules, and can provide a strong language understanding foundation for subsequent water conservancy data analysis, decision support and other tasks.

[0031] Step 202: Input the pre-processed unstructured text data into the target large language model; Step 203: Extract the water conservancy entities, entity attributes and inter-entity relationships in the unstructured text data through the natural language parsing rule set.

[0032] In this step, the pre-processed unstructured text data is input into the target large language model that has completed training. According to the entity extraction templates and other rules in the natural language parsing rule set, various water-related entities such as hydrological elements (including but not limited to water conservancy facilities, geographical areas, etc.), disaster events (such as floods, droughts, etc.), and disposal measures (response strategies for disasters, etc.) are identified in the text. According to the attribute mapping relationship, the water-related entities extracted have various attributes. According to the rules of triple construction specification, the relationships between water-related entities are analyzed, and the association structure between entities is constructed, such as the corresponding relationship between a disaster event and the disposal measures taken, to clearly express the knowledge logic in the water conservancy field.

[0033] Step 204: Generating structured text data in the form of triples from the extraction results.

[0034] In this step, the triple format is a data structuring representation method, usually composed of subject, predicate, and object. It can simplify complex knowledge into basic semantic units. In the water conservancy scenario, the subject refers to a water-related entity, the predicate represents a certain attribute of the entity or a relationship with other entities, and the object is the specific value of the attribute or another entity associated with it. In this way, originally scattered and irregular unstructured text data is converted into structured text data, facilitating subsequent storage, query, and analysis, and providing a foundation for applications such as building a water conservancy data knowledge base.

[0035] Step 30: Integrating structured text, structured table, and image data to obtain the target water conservancy data knowledge base.

[0036] As an optional embodiment, the structured text, structured table, and image data are integrated to obtain the target water conservancy data knowledge base, which can specifically include: step 301: parsing the field metadata of the structured table and performing same-name matching between the field metadata of the table and the triple entities generated by the structured text to obtain the attribute mapping relationship.

[0037] In this step, the names, data types, meanings, and other information of the columns in the table are analyzed and extracted. Field metadata essentially describes the basic information of each column in the table, which can help understand the structure and meaning of the data in the table. The field metadata of the table is matched with the triple entities generated by the structured text. By comparing the field names in the field metadata with the names of the subject or object in the triple, the same or similar names are matched to obtain the logical correspondence rules between the fields in the structured table and the knowledge entity attributes. Through field name matching or synonym alignment, accurate projection of table data to knowledge nodes is achieved.

[0038] Step 302: Identify the spatial coordinate information in the image data and match the spatial coordinate information with the triples entities generated by the structured text to obtain the spatial position binding relationship.

[0039] In this step, spatial coordinate information is extracted from image data through image processing and analysis techniques, which can be two-dimensional or three-dimensional coordinates, used to represent the specific position of an element in the image. By comparing the spatial coordinate information with the subject or object names in the triples, the same or similar names are found for matching. Through coordinate analysis and entity matching, the spatial information is converted into the GIS attribute of the knowledge node, and the spatial correlation rule between the geographic coordinates in the image and the knowledge entity is obtained.

[0040] Step 303: Based on the attribute mapping relationship and the spatial position binding relationship, fuse the standard water conservancy data to obtain the water conservancy knowledge graph.

[0041] In this step, under the unified knowledge framework, the semantics, space, and time dimensions of heterogeneous data are logically connected to form a multi-dimensional interconnected knowledge network, i.e., the water conservancy knowledge graph. In this water conservancy knowledge graph, the meaning expressed by the data can be understood from the semantic dimension, and the key water conservancy entities, attributes, and their relationships can be identified. In the spatial dimension, the geographic location and spatial distribution of water conservancy objects are clarified. In the time dimension, the changes of water conservancy data over time are sorted out. Through this multi-dimensional connection, a three-dimensional water conservancy knowledge network is constructed, which can more comprehensively and accurately reflect the actual situation of the water conservancy field. By analyzing the data and relationships in the knowledge graph, potential rules and trends in the water conservancy field can be discovered, providing scientific basis for water conservancy planning, disaster warning, water resource management, etc., and helping decision-makers make more reasonable and effective decisions.

[0042] Step 304: Based on the water conservancy knowledge graph, construct the target water conservancy data knowledge base.

[0043] As an optional embodiment, based on the water conservancy knowledge graph, the target water conservancy data knowledge base is constructed, which can specifically include: Step 3041: Extract the water conservancy event data with time stamp in the water conservancy knowledge graph, and establish the space-time dimension index node according to the spatial position binding relationship.

[0044] In this step, the timestamp records the specific time of the event occurrence, which is crucial for understanding the sequence, development process and time relationship with other events of water conservancy events. Extracting these water conservancy event data with timestamps helps to obtain the time characteristics and dynamic information of water conservancy events; the index node is equivalent to the label or directory in the knowledge graph, which marks and organizes water conservancy events according to time and space dimensions, and can more efficiently retrieve and query water conservancy events according to time and space conditions, realizing the rapid positioning and access of water conservancy data in the time and space dimensions, and providing stronger functions for the query, analysis and application of water conservancy knowledge.

[0045] Step 3042: Import the spatio-temporal index node into the Neo4j graph database to perform relationship topology storage to obtain the target water conservancy data knowledge base.

[0046] In this step, Neo4j is a graph database specially used for storing and querying graph structure data. The graph database represents data through nodes and relationships, where nodes represent entities (such as water conservancy events, facilities, etc.), and relationships represent the association between nodes (such as the location of event occurrence, the causal relationship between events, etc.). Relationship topology storage refers to storing data in a graph database in a way that can reflect the complex relationships and structures between nodes. In the context of water conservancy knowledge graph, not only the water conservancy events and their spatio-temporal index nodes need to be stored, but also the relationships between these nodes, such as the causal relationship between events, the time sequence relationship, the spatial adjacency relationship, etc. The created spatio-temporal index node is imported into the Neo4j graph database, so that Neo4j can utilize its efficient graph data processing capabilities to store, query and analyze the data in the water conservancy knowledge graph.

[0047] Step 40: Upon receiving the data query instruction, based on the query information contained in the data query instruction, similarity matching is performed in the target water conservancy data knowledge base according to the preset query rule to generate the target water conservancy plan.

[0048] As an optional embodiment, upon receiving the data query instruction, based on the query information contained in the data query instruction, similarity matching is performed in the target water conservancy data knowledge base according to the preset query rule to generate the target water conservancy plan, which can specifically include: step 401: performing semantic analysis on the query information to generate a digital feature vector containing the query intent.

[0049] In this step, when the user inputs the query information, the query information is usually expressed in natural language. In order to enable the computer to understand the query intention of the user, semantic analysis needs to be performed on the query information; first, the continuous text is divided into independent words, the part of speech of each word is recognized, the semantic roles of each component in the sentence are analyzed, the roles of the subject, object and the like in the event are clarified, and after semantic analysis, the text content of the query information is converted into a structured data form; according to the result of semantic analysis, the key features representing the query intention are extracted, each key feature is weighted, the extracted features and the corresponding weights are combined into a vector, i.e. a digital feature vector, which represents the core content and intention of the query information in a high-dimensional space.

[0050] Step 402: Calculate the cosine similarity of the digital feature vector and each historical vector in the target water conservancy data knowledge base.

[0051] In this step, the cosine similarity is a similarity measurement method, which evaluates the similarity of two vectors by calculating the cosine value of the included angle of the two vectors in the vector space. Specifically, the cosine similarity has a value range of -1 to 1, where 1 indicates that the directions of the two vectors are completely the same (i.e. completely similar), 0 indicates that the two vectors are orthogonal (i.e. not similar), and -1 indicates that the directions of the two vectors are completely opposite. In the calculation process, the generated digital feature vector is compared with each historical vector in the knowledge base, the cosine similarity of the two is calculated, and the most similar historical vector to the input feature vector is found, which helps to understand the degree of association between the current query or input and the existing information in the knowledge base, thereby providing a basis for subsequent water conservancy knowledge retrieval, query answering or decision support tasks.

[0052] Step 403: Extract water conservancy events in the target water conservancy data knowledge base with a similarity score greater than a preset threshold, and integrate them into a target water conservancy plan.

[0053] In this step, according to the calculated similarity, water conservancy events in the knowledge base with a similarity score greater than a preset threshold are selected, and key information of similar events such as event background, treatment measures, results, etc. is collected. These information is summarized in chronological order, processing flow and other logical order, and a complete water conservancy plan is generated after the information is arranged, which contains multiple possible treatment schemes, resource allocation suggestions, etc. to provide reference for decision makers and help decision makers deal with current water conservancy problems.

[0054] Step 404: Collect user feedback data on the execution of the target water conservancy plan, and generate feedback scores and treatment effect indicators.

[0055] In this step, after generating the water conservancy plan, feedback of the user on the execution of the plan is collected, including the difficulty of plan execution, the contribution to problem solving, and the advantages and disadvantages in the execution process, etc. According to the execution feedback data, a feedback score and a treatment effect index are generated, wherein the feedback score is a comprehensive evaluation of the execution effect of the whole plan, and the treatment effect index measures the actual effectiveness of the plan in handling specific problems from different dimensions.

[0056] Step 405: Based on the feedback score, adjust the recommended weight value of the corresponding water conservancy event in the target water conservancy data knowledge base, and update the query rule.

[0057] In this step, the recommended weight value determines the probability and priority of the water conservancy event being recommended in the future. A high feedback score may increase the weight value, so that the event-related plan is more likely to be recommended in subsequent similar situations, and a low score may decrease the weight value. According to the adjusted weight value, the rule is updated to more accurately match the user query intention with the water conservancy event in the knowledge base, thereby improving the intelligent level and service quality of the water conservancy knowledge base.

[0058] The above is the method embodiment of the present application. Based on the same inventive concept, the present application also provides a water conservancy knowledge base management device based on a large language model, which has a structure as shown in Figure 2 .

[0059] Figure 2 A water conservancy knowledge base management device based on a large language model provided by an embodiment of the present application has an internal structure as shown in Figure 2 . The device comprises: at least one processor 201; and a memory 202 in communication connection with the at least one processor; The memory 202 stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor 201 to enable the at least one processor 201 to: collect multi-source water conservancy data and pre-process multi-dimensional water conservancy data to obtain standard water conservancy data; analyze the standard water conservancy data based on a preset target large language model to convert unstructured text into structured text; integrate the structured text, structured table and image data to obtain a target water conservancy data knowledge base; when receiving a data query instruction, perform similarity matching in the target water conservancy data knowledge base according to the query information contained in the data query instruction based on a preset query rule to generate a water conservancy plan.

[0060] Some embodiments of the present application provide a water conservancy knowledge base management device based on a large language model. Figure 1The non-volatile computer storage medium based on a large language model for water conservancy knowledge base management stores computer executable instructions, and the computer executable instructions are configured to: collect multi-source water conservancy data, and preprocess multi-dimensional water conservancy data to obtain standard water conservancy data; analyze the standard water conservancy data based on a preset target large language model, and convert unstructured text into structured text; integrate the structured text, structured tables and image data to obtain a target water conservancy data knowledge base; when a data query instruction is received, similarity matching is performed in the target water conservancy data knowledge base according to a preset query rule based on query information contained in the data query instruction, to generate a water conservancy plan.

[0061] Each of the embodiments in the present application is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other. Each embodiment focuses on the difference from other embodiments. In particular, for the Internet of Things device and medium embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the part of the method embodiments.

[0062] The system and medium provided by the embodiments of the present application are one-to-one corresponding to the method, so the system and medium also have similar beneficial technical effects as the method. Since the beneficial technical effects of the method have been described in detail above, the beneficial technical effects of the system and medium will not be described here.

[0063] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer usable program code.

[0064] The present application is described with reference to flowcharts and / or block diagrams according to the method, device (system), and computer program product of the embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of the flows and / or blocks in the flowcharts and / or block diagrams can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device produce a device that implements the functions specified in the flowcharts and / or block diagrams. Figure 1 The device that implements the functions specified in one flow or multiple flows and / or blocks. Figure 1 The device that implements the functions specified in one flow or multiple flows and / or blocks.

[0065] These computer program instructions can also be stored in a computer- readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instructions which implement the Figure 1 function specified in the flow or flows and / or blocks Figure 1 of the block or blocks.

[0066] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the Figure 1 function specified in the flow or flows and / or blocks Figure 1 of the block or blocks.

[0067] In a typical configuration, a computing device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

[0068] The memory can include non-persistent memory and / or volatile memory, such as random access memory (RAM) about which the computer stores information about an operating system, application software, and / or the like. Memory is an example of computer readable media.

[0069] Computer readable media includes permanent and non-permanent, moveable and non- moveable media that can be implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disc read-only memory (CD-ROM), digital versatile discs (DVDs) or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information that is accessible to a computing device. According to the definition provided herein, computer readable media excludes transitory media, such as modulated data signals and carrier waves.

[0070] It should also be noted that the terms "comprising," "including," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can also include other elements not expressly listed or inherent to such process, method, article, or apparatus. An element proceeded by "comprises a... " does not, without more constraints, exclude the existence of additional identical elements in the process, method, article, or apparatus that comprises the element.

[0071] The above description is merely illustrative of the application, and not restrictive. Various modifications and changes can become apparent to those skilled in the art. Incorporating any modification, equivalent substitution, improvement, etc. within the spirit and principle of the application, shall be included in the scope of the claims of the application.

Claims

1. A water conservancy knowledge base management method based on a large language model, characterized in that: The method comprises: Collecting multi-source water conservancy data and pre-processing the multi-dimensional water conservancy data to obtain standard water conservancy data; wherein the multi-source water conservancy data includes unstructured text, structured tables, and image data; Parsing the standard water conservancy data based on a preset target large language model to convert the unstructured text into structured text; Integrating the structured text, the structured table and the image data to obtain a target water conservancy data knowledge base; When a data query instruction is received, similarity matching is performed in the target water conservancy data knowledge base according to preset query rules based on the query information contained in the data query instruction to generate a target water conservancy plan.

2. A water conservancy knowledge base management method based on a large language model according to claim 1, characterized in that: The method further comprises: Building a large language model includes: Collecting a water conservancy text dataset and defining natural language parsing rules based on the water conservancy text dataset; wherein the water conservancy text dataset includes hydrological entities, disaster events, and disposal measures; the natural language parsing rules include entity extraction templates, attribute mapping relationships, and triple construction specifications; The large language model is trained based on the natural language parsing rule set until a preset recognition accuracy rate is reached to obtain a target large language model.

3. A water conservancy knowledge base management method based on a large language model according to claim 2, characterized in that: Parsing the standard water conservancy data based on a preset target large language model and converting the unstructured text into structured text specifically includes: Inputting the preprocessed unstructured text data into the target large language model; Extracting water conservancy entities, entity attributes and relationships between entities in the unstructured text data through the natural language parsing rule set; The extraction results are converted into structured text data in triple format.

4. The water conservancy knowledge base management method based on a large language model according to claim 1 is characterized in that: Integrating the structured text, the structured table, and the image data to obtain a target water conservancy data knowledge base specifically includes: Parsing the field metadata of the structured table, and matching the field metadata of the table with the triple entities generated by the structured text by the same name to obtain an attribute mapping relationship; Identifying spatial coordinate information in the image data, and matching the spatial coordinate information with triple entities generated from the structured text to obtain a spatial position binding relationship; Based on the attribute mapping relationship and the spatial position binding relationship, the standard water conservancy data is integrated to obtain a water conservancy knowledge graph; Based on the water conservancy knowledge graph, the target water conservancy data knowledge base is constructed.

5. A water conservancy knowledge base management method based on a large language model according to claim 4, characterized in that: Based on the water conservancy knowledge graph, the target water conservancy data knowledge base is constructed, specifically including: Extracting water conservancy event data with timestamps from the water conservancy knowledge graph, and establishing spatiotemporal dimension index nodes according to the spatial position binding relationship; The spatiotemporal index nodes are imported into a Neo4j graph database to perform relational topology storage, so as to obtain a target water conservancy data knowledge base.

6. The water conservancy knowledge base management method based on a large language model according to claim 1 is characterized in that: Based on the query information contained in the data query instruction, similarity matching is performed in the target water conservancy data knowledge base according to preset query rules to generate a target water conservancy plan, specifically including: Performing semantic analysis on the query information to generate a digital feature vector containing the query intent; Calculating the cosine similarity between the digital feature vector and each historical vector in the target water conservancy data knowledge base; Water conservancy events with similarity scores greater than a preset threshold value in the target water conservancy data knowledge base are extracted and integrated into a target water conservancy plan.

7. The water conservancy knowledge base management method based on a large language model according to claim 1 is characterized in that: After generating the target water conservancy plan, the method further includes: Collecting user feedback data on the implementation of the target water conservancy plan and generating feedback scores and treatment effect indicators; Based on the feedback score, the recommendation weight value of the corresponding water conservancy event in the target water conservancy data knowledge base is adjusted, and the query rule is updated.

8. The method for managing a water conservancy knowledge base based on a large language model according to claim 1, characterized in that: Collect multi-source water conservancy data and pre-process the multi-dimensional water conservancy data to obtain standard water conservancy data, specifically including: Extracting the unstructured text and the structured table keywords and performing deduplication to obtain standard unstructured text data; Extracting header information and data content of the structured table and performing data type conversion to obtain standard structured table data; Using an image enhancement algorithm, processing the image data in terms of brightness, contrast and resolution to obtain standard image data; The standard unstructured text data, the standard structured table data and the standard image data are integrated to obtain standard water conservancy data.

9. A water conservancy knowledge base management device based on a large language model, characterized in that: The device comprises: at least one processor; and, a memory communicatively coupled to the at least one processor; The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform a method according to any one of claims 1 to 8.

10. A non-volatile computer storage medium for managing a water conservancy knowledge base based on a large language model, storing computer executable instructions, characterized in that: When the computer-executable instructions are executed, a method according to any one of claims 1 to 8 is implemented.

Images