Question and answer accuracy improving method and device for water conservancy industry large model

Abstract

The invention provides a water conservancy industry large model question and answer accuracy improving method comprising the following steps: collecting multi-source data, and preprocessing the multi-source data; performing deep mining on the preprocessed data, extracting water conservancy facility information, synchronously identifying various hydrological elements, and constructing a water conservancy knowledge graph; simulating and generating water conservancy data in a flood scene according to the water conservancy knowledge graph, and performing sentence recombination by adopting a text data enhancement technology to change a sentence structure but keep semantics unchanged; performing comparative analysis on historical water conservancy data and updated new data, and in combination with importance weight evaluation of the data, screening out new data with high weight and preferentially bringing the new data into the water conservancy knowledge graph for updating; and performing comprehensive analysis on the retrieved knowledge, the ecological environment standard, the knowledge of the water conservancy knowledge graph learned by the deep learning model and semantic deep fusion by utilizing a retrieval enhancement generation technology, and accurately generating an answer of the water conservancy project construction to the ecological environment influence problem in a large model of the water conservancy industry.

Description

Technical Field

[0001] The present invention relates to the technical field of large language models, and in particular to a method and device for improving the question-answering accuracy of a large model in the water conservancy industry. Background Art

[0002] As the water conservancy industry accelerates its digital transformation, the demand for intelligent decision support systems is extremely urgent. Question-and-answer applications based on large models offer water conservancy professionals a new path to easily access information and make informed decisions. However, these applications currently face numerous challenges.

[0003] Lack of professional knowledge: The general large-scale model has a serious shortage of professional knowledge in the field of water conservancy. It is difficult to accurately grasp the complex principles of hydrological models and the unique and sophisticated operating rules of various water conservancy projects. This makes it difficult for its answers to water conservancy professional scenarios to meet the precise requirements and cannot provide a reliable basis for actual work.

[0004] Physical process simulation bias: Water conservancy systems involve numerous complex physical processes, such as the dynamic evolution of floods within a river basin and the flow of water resources through complex pipe networks. Existing Q&A applications often lack in-depth and accurate modeling when simulating these physical processes, resulting in inference results that deviate significantly from reality. This not only misleads decision-making but also seriously undermines the reliability of decisions, potentially leading to significant errors in the operation of water conservancy projects and water resource allocation.

[0005] Incomplete data coverage: Water conservancy data is particularly scarce in extreme scenarios, such as exceptional floods and rare water conservancy facility failures. This severely limits the model's generalization capabilities when faced with these special problems due to a lack of sufficient training data. This makes it difficult to provide accurate and effective answers, making it unable to meet the decision-making needs of dealing with extreme situations.

[0006] Lagging knowledge updates: The water conservancy industry constantly generates new monitoring data, along with new project cases and advanced technical experience. However, traditional knowledge bases have lengthy update cycles, making it difficult to promptly incorporate this latest information into Q&A applications. This results in answers that are significantly out of touch with the latest industry developments, hindering the provision of timely and targeted assistance to water conservancy professionals.

[0007] Difficulties in multimodal integration: The water conservancy industry encompasses a rich and diverse array of data types, encompassing text-based regulations, numerical monitoring data, and image-based remote sensing imagery. However, current question-and-answer applications struggle to integrate this multimodal data, failing to fully explore the potential connections between these different data types. This makes it difficult for models to fully understand the nature of complex questions, impacting the accuracy and comprehensiveness of their responses.

[0008] In view of this, it is necessary to provide a method and device for improving the accuracy of question answering in large models of the water conservancy industry to solve the above problems. Summary of the Invention

[0009] The purpose of the present invention is to provide a method and device for improving the accuracy of question answering in a large model of the water conservancy industry. The method and device can accurately extract key entity information such as water conservancy facility information and hydrological elements, and combine efficient text data enhancement technology to perform synonym replacement, sentence reorganization and other operations on text data, providing a rich, accurate and diverse knowledge base for the large model, ensuring that its answers are highly professional, accurate and comprehensive.

[0010] To achieve the above objectives, the technical solution proposed in the present invention is implemented as follows: a method for improving the accuracy of question answering in a large model of the water conservancy industry, comprising the following steps: S1: Collect hydrological monitoring data, engineering operation data, remote sensing image data, and regulatory document data as multi-source data, and use data cleaning algorithms and data tools to pre-process the multi-source data; S2. Use deep learning models to deeply mine the pre-processed data and extract water conservancy facility information, simultaneously identify various hydrological elements, and build a water conservancy knowledge graph; S3. Use the hydraulic physics simulation model to simulate and generate hydraulic data under flood scenarios based on the hydraulic knowledge graph. At the same time, use text data enhancement technology to restructure sentences, changing the sentence structure while keeping the semantics unchanged. S4. Through comparative analysis of historical water conservancy data and updated new data, combined with the importance weight assessment of the data, select the new data with high weight and prioritize them for inclusion in the water conservancy knowledge graph update; S5. Use retrieval enhancement generation technology to conduct a comprehensive analysis of the retrieved knowledge, ecological and environmental standards, and the knowledge of the water conservancy knowledge graph learned by the deep learning model, and the semantic deep fusion generated by the water conservancy physical simulation model itself, to accurately generate answers to questions about the impact of water conservancy project construction on the ecological environment in the large model of the water conservancy industry.

[0011] Preferably, the method further includes step S6, using reinforcement learning to dynamically optimize the knowledge update and retrieval process of the water conservancy knowledge graph of the deep learning model, and adjusting the question-answering strategy of the large model of the water conservancy industry.

[0012] Preferably, step S6 also includes step S61, which allows the water conservancy industry big model to automatically adjust the timing, content and parameter settings of the knowledge update of the water conservancy knowledge graph and the retrieval algorithm in the process of continuous trial, feedback and learning by setting a reward mechanism to achieve optimal performance.

[0013] Preferably, in step S62, if the user feedback indicates that the answer of the large model of the water conservancy industry lacks certain key information, the position of the key information in the water conservancy knowledge graph and the reason for the omission during the retrieval process will be analyzed; at the same time, based on the key information fed back, the priority and direction of knowledge updating will be adjusted to continuously improve the accuracy and user satisfaction of the large model question and answer of the water conservancy industry.

[0014] Preferably, in step S63, when the answer generated by the water conservancy industry big model is verified to be erroneous, the water conservancy industry big model system not only records the error information, but also analyzes the cause of the error in real time, and quickly transmits the feedback information to the deep learning model, water conservancy knowledge graph update, and various links of multimodal interaction.

[0015] Preferably, step S5 also includes step S51, which maps data of different modalities such as natural language text, speech, and image into the same semantic space using multimodal interaction by constructing a joint embedding space, so that the large model of the water conservancy industry can accurately identify the common semantic information expressed by data of different modalities through cross-modal semantic alignment technology.

[0016] Preferably, step S5 also includes step S52, dynamically adjusting the multimodal fusion weights in real time, so that the large model of the water conservancy industry effectively utilizes the advantages of different modal data to generate accurate answers.

[0017] An electronic device comprises a memory, a processor and a computer program stored in the memory and executable on the processor. When the processor executes the program, steps of a method for improving the accuracy of question-answering of a large model of the water conservancy industry are implemented.

[0018] A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that when the computer program is executed by a processor, the method for improving the accuracy of question and answering of a large model of the water conservancy industry is implemented.

[0019] A computer program product includes a computer program, characterized in that when the computer program is executed by a processor, it implements the steps of a method for improving the accuracy of question answering of a large model in the water conservancy industry.

[0020] Compared with existing technologies, the beneficial effects are: 1) it can accurately extract key entity information such as water conservancy facilities information and hydrological elements, and combine efficient text data enhancement technology to perform synonym replacement, sentence reorganization and other operations on text data, providing a rich, accurate and diverse knowledge base for the large model, ensuring that its answers are highly professional and comprehensive.

[0021] 2) It can autonomously determine the most effective knowledge update frequency and retrieval strategy through reinforcement learning, adapt to the dynamically changing water conservancy industry environment for dynamic optimization, and then adjust the question-answering strategy of the water conservancy industry large model, realizing the self-repair and continuous optimization of the entire water conservancy industry large model, and continuously improving the accuracy and reliability of the answers.

[0022] Other features and advantages of the present invention will be set forth in the following description, and part will be apparent from the description, or may be understood through practice of the present invention. Features and advantages of the present invention may be realized and obtained through the elements and combinations specifically indicated in the appended claims. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. It should be understood that the following drawings only illustrate certain embodiments of the present invention and should not be considered as limiting the scope. A person of ordinary skill in the art can also derive other relevant drawings based on these drawings without inventive effort. Figure 1 This is a flow chart of the method for improving the accuracy of question answering in a large water conservancy industry model provided by the present invention. DETAILED DESCRIPTION

[0024] In order to make the purpose, technical solution and beneficial technical effects of the present invention more clear, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described in this specification are only for the purpose of explaining the present invention and are not intended to limit the present invention.

[0025] It should be understood that the terms "upper", "lower", "front", "back", "left", "right", "top", "bottom", "inside", "outside", etc., indicating the orientation or position relationship, are based on the orientation or position relationship shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operate in a specific orientation. Therefore, it should not be understood as a limitation on the present invention.

[0026] It should also be noted that, unless otherwise expressly specified or limited, terms such as "installed," "connected," "connect," "fixed," and "set" should be understood broadly. For example, they may refer to fixed or detachable connections, or integration; mechanical or electrical connections; direct or indirect connections through an intermediate medium; and internal communication between two components or interaction between two components. Those skilled in the art will readily understand the specific meanings of these terms in the present invention based on specific circumstances.

[0027] Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the quantity of the technical features being referred to. Thus, a feature specified as "first," "second," or "third" may explicitly or implicitly include one or more of such features. Furthermore, "plurality" and "several" refer to two or more, unless otherwise specifically specified.

[0028] See also Figure 1 The present invention proposes a method for improving the accuracy of question answering in a large model of the water conservancy industry, comprising the following steps: S1. Collect a wide range of hydrological monitoring data (real-time and historical data such as water level, flow, and rainfall), project operation data (covering the operating status and maintenance records of water conservancy facilities), remote sensing image data (information such as watershed topography and water body distribution), and normative document data (national standards, industry standards, and water conservancy policies and regulations) as multi-source data. Since the names of water conservancy projects and place names in multi-source data may contain spelling errors due to human errors or differ due to different naming habits in different regions, it is necessary to use data cleaning algorithms and data tools to pre-process multi-source data (denoising, deduplication, and format unification) to ensure data accuracy and consistency.

[0029] Data cleaning algorithms can adopt the statistical 3σ principle, isolation forest algorithm, hash algorithm, Levenshtein distance algorithm, etc. The statistical 3σ principle can accurately detect such outliers, the isolation forest algorithm can effectively identify abnormal equipment operating parameters, the hash algorithm can quickly locate duplicate data, and the Levenshtein distance algorithm can accurately identify similar errors in text data.

[0030] Data tools can include OpenRefine and Trifacta Wrangler. OpenRefine can be used for data screening, clustering, and batch error correction. Trifacta Wrangler can automatically detect anomalies and missing values ​​through intelligent analysis and provide repair suggestions, thereby comprehensively resolving problems such as data anomalies, duplications, and text errors. This provides reliable data support for subsequent data analysis, model training, and decision-making in the water conservancy industry, and pre-processes multi-source data (denoising, deduplication, format unification, etc.) to ensure data accuracy and consistency.

[0031] S2. Through deep learning models, the pre-processed massive data is deeply mined and information on water conservancy facilities (such as reservoirs, dams, sluice gates, etc.) is accurately extracted. Various hydrological elements (runoff for water resource allocation, evaporation that affects regional water balance, rainfall as the main form of precipitation, etc.) are simultaneously identified. Relationship extraction technology is used to accurately determine the relationship between entities (for example, accurately determine the affiliation of a reservoir with its basin), and construct a comprehensive, detailed and logically clear water conservancy knowledge map.

[0032] Moreover, in this embodiment, the knowledge integrity and accuracy of the water conservancy knowledge graph can be ensured by continuously adopting a knowledge graph completion algorithm (such as a link prediction algorithm based on deep learning) to continuously update and improve the relationships between entities.

[0033] Traditional methods struggle to quickly and accurately extract key entity information from this massive amount of data. For example, manually identifying key information about engineering facilities (such as dams and sluice gates), hydrological elements (such as water levels and flow rates), and regulatory documents in numerous water conservancy project construction reports is time-consuming, labor-intensive, and prone to errors.

[0034] In the complex data processing environment of the water conservancy industry, deep learning models, leveraging advanced entity information recognition technology, provide powerful support for solving numerous challenging problems. For example, deep learning models based on convolutional neural networks (CNNs) and recurrent neural networks (RNNs) can effectively capture characteristic patterns in text data. Natural language processing tools such as NLTK and SpaCy provide rich support for deep learning models, including preprocessing, word segmentation, and part-of-speech tagging. These tools can rapidly traverse vast amounts of data and accurately identify information about various entities. This not only significantly improves data processing efficiency but also ensures the accuracy of extracted information. This provides a solid data foundation for subsequent work such as constructing a water conservancy knowledge graph, analyzing water conservancy business relationships, and supporting decision-making, effectively promoting the digital and intelligent development of the water conservancy industry.

[0035] S3. Use hydraulic physics simulation models (e.g., flood evolution simulation models) to simulate and generate hydraulic data (including flood inundation range, evolution path, peak flow, etc.) under extreme flood scenarios based on hydraulic knowledge graphs. Simultaneously, use text data enhancement technology to reorganize sentences (e.g., perform synonym replacement, replacing "strengthen inspections" with "enhanced patrols"), changing sentence structure while maintaining semantics. This expands the diversity and richness of training data, enabling deep learning models to learn a wider range of language expressions and knowledge scenarios.

[0036] Hydraulic physics simulation models are a key technical tool for solving many complex problems in the water conservancy sector. The planning, construction, and management of water conservancy projects present challenges in intuitively understanding and accurately predicting complex physical processes such as water flow, water resource allocation, and flood evolution. For example, when planning large-scale water conservancy projects, it is crucial to know in advance the flow rate and flow pattern of the reservoir under different operating conditions to ensure the safe and stable operation of the project facilities. Traditional empirical methods struggle to provide accurate answers.

[0037] Hydraulic physics simulation models, such as Mike-Flood and HEC-RAS, are mathematical models constructed based on physical principles such as hydrodynamics and thermodynamics. They simulate various physical phenomena in water conservancy systems through numerical calculations. In terms of flood prevention, hydraulic physics simulation models can simulate the generation, evolution path, and inundation range of floods under different rainfall intensities and durations, helping flood control command departments to formulate scientific and reasonable emergency plans in advance, accurately determine personnel evacuation areas and flood control material deployment plans, and effectively reduce flood disaster losses. In water resources management, hydraulic physics simulation models can simulate the flow and distribution of water resources within the basin under different seasons and different water demands, providing a basis for optimizing water resources scheduling strategies, ensuring the rational supply of water for urban and rural life, industrial and agricultural production, resolving the contradiction between the uneven temporal and spatial distribution of water resources and water demand, and achieving scientific management and sustainable development of water conservancy systems.

[0038] S4. Through comparative analysis of historical water conservancy data and updated new data, combined with the importance weight assessment of the data, new data with high weight are screened out and included in the water conservancy knowledge graph update first.

[0039] For example, for sudden extreme hydrological event data, based on factors such as the rarity of the event and the severity of its impact on the water conservancy system, this type of new data is given a higher weight and is prioritized for inclusion in knowledge base updates to ensure that the deep learning model can quickly adapt to and learn the latest important information.

[0040] When new data is included, not only are new relationships between entities simply added, but the semantic reasoning capabilities of the water conservancy knowledge graph are also used to check and optimize the consistency of the existing knowledge system.

[0041] For example, if a newly issued normative document puts forward new requirements for the operation and management of certain water conservancy facilities, the reasoning of the water conservancy knowledge graph can automatically update the relationship between the water conservancy facilities and the normative document, and jointly update the upstream and downstream knowledge nodes involved, to ensure the integrity and accuracy of the knowledge graph and provide more reliable knowledge support for the deep learning model.

[0042] S5. Utilize Retrieval-augmented Generation (RAG) technology to deeply integrate the retrieved knowledge, ecological and environmental standards, and the knowledge of the water conservancy knowledge graph learned by the deep learning model with the semantics generated by the water conservancy physical simulation model itself for comprehensive analysis, and accurately generate answers to complex questions about the impact of water conservancy project construction on the ecological environment in the large water conservancy industry model.

[0043] For example, when answering complex questions about the impact of water conservancy project construction on the ecological environment, retrieval enhancement generation technology is used to conduct a comprehensive analysis of the retrieved engineering construction knowledge, ecological environment assessment standards, and knowledge learned by deep learning models based on large amounts of data, and the semantic deep fusion generated by the water conservancy physical simulation model itself, accurately generating more comprehensive, in-depth and actual answers, effectively improving the quality and accuracy of the answers.

[0044] Because in traditional multimodal interactions, the semantic fusion of data from different modalities often has deviations, resulting in inaccurate understanding of user intentions by large models in the water conservancy industry.

[0045] Step S5 also includes step S51. Therefore, the present invention constructs a joint embedding space to map data of different modalities such as natural language text, voice, and image into the same semantic space using multimodal interaction, so that the large model of the water conservancy industry can accurately identify the common semantic information expressed by data of different modalities through cross-modal semantic alignment technology.

[0046] For example, for a reservoir water level monitoring image and text describing the reservoir water level situation, the water conservancy physics simulation model can accurately identify the common semantic information expressed by the two through cross-modal semantic alignment technology, such as the key information that the current reservoir water level is near the warning water level, thereby achieving deep fusion and collaborative understanding of multimodal data, and improving the ability of large models in the water conservancy industry to understand complex problems.

[0047] Step S5 also includes step S52, dynamically adjusting the multimodal fusion weights in real time, so that the large model of the water conservancy industry can more effectively utilize the advantages of different modal data and generate more accurate and comprehensive answers.

[0048] For example, when a user inputs a question into the water conservancy industry model, the model first analyzes the question type and the domain knowledge involved, determining the importance of different modal data in answering the question. For example, for questions regarding the appearance of damage to water conservancy facilities (reservoirs, dams, sluice gates, etc.), image modal data will be given a higher weight; whereas, for questions regarding regulatory documents, text modal data will be given a higher weight. By dynamically adjusting multimodal fusion weights in real time, the water conservancy industry model can more effectively leverage the strengths of different modal data to generate more accurate and comprehensive answers.

[0049] S6. Use reinforcement learning to dynamically optimize the knowledge update and retrieval process of the water conservancy knowledge graph of the deep learning model, and adjust the question-answering strategy of the large model of the water conservancy industry.

[0050] Specifically, step S6 also includes step S61, by setting a reward mechanism (such as improved answer accuracy, improved user satisfaction, etc. as positive rewards, and wrong answers or user dissatisfaction as negative rewards), the water conservancy industry large model can automatically adjust the timing, content, and parameter settings of the knowledge update of the water conservancy knowledge graph and the retrieval algorithm in the process of continuous trial, feedback, and learning to achieve optimal performance.

[0051] For example, when faced with frequently changing water conservancy monitoring data, the deep learning model can autonomously determine the most effective knowledge update frequency and retrieval strategy through reinforcement learning, adapt to the dynamically changing water conservancy industry environment for dynamic optimization, and then adjust the question-answering strategy of the large model of the water conservancy industry.

[0052] Step S62: If the user feedback indicates that the answer of the large model of the water conservancy industry is missing some key information, the position of the key information in the water conservancy knowledge graph and the reason for the omission during the retrieval process will be analyzed, so as to optimize the parameters and strategies of the retrieval algorithm in a targeted manner and improve the accuracy of the next retrieval; at the same time, according to the key information fed back, the priority and direction of knowledge updating will be adjusted to continuously improve the accuracy of the large model question and answer of the water conservancy industry and user satisfaction.

[0053] Step S63: When the answer generated by the water conservancy industry big model is verified to be incorrect, the water conservancy industry big model system not only records the error information, but also analyzes the cause of the error in real time, and quickly transmits the feedback information to the deep learning model, water conservancy knowledge graph update and various links of multimodal interaction.

[0054] For example, if verification finds that the wrong answer is caused by the missing of some knowledge in the water conservancy knowledge graph, the system will immediately start the knowledge update process to supplement the relevant knowledge; if the error is caused by improper multimodal data fusion, the multimodal fusion strategy and parameters will be adjusted to achieve self-repair and continuous optimization of the entire water conservancy industry model, and continuously improve the accuracy and reliability of the answers.

[0055] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and runnable on the processor. When the processor executes the program, the steps of a method for improving the accuracy of question and answering of a large model of the water conservancy industry are implemented.

[0056] The present invention also provides a non-transitory computer-readable storage medium on which a computer program is stored, characterized in that when the computer program is executed by a processor, the method for improving the accuracy of question and answering of a large model of the water conservancy industry is implemented.

[0057] The present invention also provides a computer program product, including a computer program, characterized in that when the computer program is executed by a processor, the computer program implements the steps of a method for improving the accuracy of question and answering of a large model of the water conservancy industry.

[0058] Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented using software plus a necessary general-purpose hardware platform, or of course, hardware. Based on this understanding, the essence of the above technical solution, or the portion 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, a magnetic disk, or an optical disk, and includes a number of instructions for causing a computer device (such as a personal computer, server, or network device) to execute the methods described in each embodiment or certain portions of the embodiments.

[0059] In particular, according to some embodiments of the present disclosure, the process described above can be implemented as a computer software program. For example, some embodiments of the present disclosure include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes program code for executing the method shown in the flowchart. In some such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from a storage device, or installed from a ROM. When the computer program is executed by a processing device, the above-mentioned functions defined in the method of some embodiments of the present disclosure are performed.

[0060] It should be noted that the computer-readable medium described in some embodiments of the present disclosure may be a computer-readable signal medium or a computer-readable storage medium or any combination of the two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or component, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In some embodiments of the present disclosure, a computer-readable storage medium may be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0061] In some embodiments of the present disclosure, a computer-readable signal medium may include a mission data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. This propagated mission data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium that can transmit, propagate, or convey a program for use by or in conjunction with an instruction execution system, apparatus, or device. The program code embodied on the computer-readable medium may be transmitted using any suitable medium, including but not limited to wire, optical cable, RF (radio frequency), or any suitable combination thereof.

[0062] In some embodiments, the client and server can communicate using any currently known or later developed network protocol, such as HTTP (HyperText Transfer Protocol), and can be interconnected with any form or medium of digital task data communication (e.g., a communication network). Examples of communication networks include local area networks ("LAN"), wide area networks ("WAN"), internetworks (e.g., the Internet), and peer-to-peer networks (e.g., adhoc peer-to-peer networks), as well as any currently known or later developed networks.

[0063] The computer-readable medium may be included in the electronic device, or may exist independently without being incorporated into the electronic device. The computer-readable medium carries one or more programs. When the one or more programs are executed by the electronic device, the electronic device: in response to detecting a query operation on a production collaboration document in the switch production line management application, determines the network connection status of the switch production line management application; in response to determining that the network connection status of the switch production line management application represents an offline state, replaces the web page entry information corresponding to the production collaboration document with target entry file information, and loads target web page resource information to display the web page of the production collaboration document offline in the switch production line management application, wherein the target entry file information is file information of a pre-downloaded entry file corresponding to the web page of the production collaboration document, and the target web page resource information is locally stored resource information corresponding to the web page; in response to determining that the network connection status of the switch production line management application represents an online state and the web page resource information corresponding to the production collaboration document is not stored locally, downloads the web page resource information of the web page from the production line document server, wherein the web page resource information includes the entry file and resource information; displays the web page of the production collaboration document in the switch production line management application according to the web page resource information, and stores the web page resource information in a local database.

[0064] Computer program code for performing the operations of some embodiments of the present disclosure may be written in one or more programming languages, or a combination thereof, including product-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0065] The present invention is not limited to what is described in the specification and embodiments, and additional advantages and modifications will be readily apparent to those skilled in the art. Therefore, the present invention is not limited to the specific details, representative devices, and illustrative examples shown and described herein without departing from the spirit and scope of the general concept defined by the claims and their equivalents.

Claims

1. A method for improving the accuracy of question answering in a large water conservancy industry model, characterized in that: The steps include: S1: Collect hydrological monitoring data, engineering operation data, remote sensing image data, and regulatory document data as multi-source data, and use data cleaning algorithms and data tools to pre-process the multi-source data; S2. Use deep learning models to deeply mine the pre-processed data and extract water conservancy facility information, simultaneously identify various hydrological elements, and build a water conservancy knowledge graph; S3. Use the hydraulic physics simulation model to simulate and generate hydraulic data under flood scenarios based on the hydraulic knowledge graph. At the same time, use text data enhancement technology to restructure sentences, changing the sentence structure while keeping the semantics unchanged. S4. Through comparative analysis of historical water conservancy data and updated new data, combined with the importance weight assessment of the data, select the new data with high weight and prioritize them for inclusion in the water conservancy knowledge graph update; S5. Use retrieval enhancement generation technology to conduct a comprehensive analysis of the retrieved knowledge, ecological and environmental standards, and the knowledge of the water conservancy knowledge graph learned by the deep learning model, and the semantic deep fusion generated by the water conservancy physical simulation model itself, to accurately generate answers to questions about the impact of water conservancy project construction on the ecological environment in the large model of the water conservancy industry.

2. The method for improving the accuracy of question answering in a large water conservancy industry model according to claim 1, characterized in that: It also includes step S6, using reinforcement learning to dynamically optimize the knowledge update and retrieval process of the water conservancy knowledge graph of the deep learning model, and adjust the question-answering strategy of the large model of the water conservancy industry.

3. The method for improving the accuracy of question answering in a large water conservancy industry model according to claim 2, characterized in that: Step S6 also includes step S61, which sets a reward mechanism to allow the water conservancy industry big model to automatically adjust the timing, content, and parameter settings of the knowledge update of the water conservancy knowledge graph and the retrieval algorithm in the process of continuous trial, feedback, and learning to achieve optimal performance.

4. The method for improving the accuracy of question answering in a large water conservancy industry model according to claim 3 is characterized in that: Step S62: If the user feedback indicates that the answer of the water conservancy industry large model is missing some key information, the position of the key information in the water conservancy knowledge graph and the reason for the omission during the retrieval process will be analyzed; at the same time, based on the feedback key information, the priority and direction of knowledge updating will be adjusted to continuously improve the accuracy and user satisfaction of the water conservancy industry large model question and answer.

5. The method for improving the accuracy of question answering in a large water conservancy industry model according to claim 3 is characterized in that: Step S63: When the answer generated by the water conservancy industry big model is verified to be incorrect, the water conservancy industry big model system not only records the error information, but also analyzes the cause of the error in real time, and quickly transmits the feedback information to the deep learning model, water conservancy knowledge graph update and various links of multimodal interaction.

6. The method for improving the accuracy of question answering in a large water conservancy industry model according to claim 1, characterized in that: Step S5 also includes step S51, which maps data of different modalities such as natural language text, speech, and image into the same semantic space using multimodal interaction by constructing a joint embedding space, so that the large model of the water conservancy industry can accurately identify the common semantic information expressed by data of different modalities through cross-modal semantic alignment technology.

7. The method for improving the accuracy of question answering in a large water conservancy industry model according to claim 6, characterized in that: Step S5 also includes step S52, dynamically adjusting the multimodal fusion weights in real time, so that the large model of the water conservancy industry effectively utilizes the advantages of different modal data and generates accurate answers.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the steps of the method for improving the accuracy of question and answering of the large model of the water conservancy industry are implemented as described in any one of claims 1 to 7.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for improving the accuracy of question answering of a large model of the water conservancy industry as described in any one of claims 1 to 7 are implemented.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method for improving the accuracy of question answering of a large model of the water conservancy industry as described in any one of claims 1 to 7 are implemented.

Images