Dam safety knowledge intelligent question and answer method based on knowledge graph and AI large model

Abstract

The invention relates to a dam safety knowledge intelligent question and answer method based on a knowledge graph and an AI large model. The method is suitable for the technical field of dam safety management. The technical problem to be solved by the invention is to provide a dam safety knowledge intelligent question and answer method based on a knowledge graph and an AI large model. According to the technical scheme, the dam safety knowledge intelligent question and answer method based on the knowledge graph and the AI large model comprises the steps that through a trained dam safety semantic analysis professional large model, retrieval keywords and statement intentions are extracted from questioning natural statements; selecting an element corresponding to the retrieval keyword from each element of the dam safety knowledge graph as a retrieval element; selecting a corresponding knowledge graph Cypher query template based on the statement intention, and generating a Cypher query command in combination with the retrieval element; and executing the Cypher query command, and retrieving an answer for replying the natural question statement from the dam safety knowledge graph.

Description

TECHNICAL FIELD

[0001] The application relates to a dam safety knowledge intelligent question answering method based on a knowledge graph and an AI large model. BACKGROUND

[0002] Dam safety management is a semi-theoretical and semi-experiential work. Senior engineers are proficient in various engineering knowledge and have accumulated rich engineering experience, and can quickly deal with complex engineering problems. However, front-line young engineers often do not have such ability, and there are problems such as insufficient engineering knowledge and lack of experience, and vertical field professional knowledge is usually difficult to query in a conventional search engine. Therefore, it is necessary to establish a fast and accurate dam safety knowledge intelligent question answering method to assist front-line dam management personnel in professional knowledge retrieval.

[0003] Intelligent question answering is a means for providing answers to problems for users by understanding user intentions, querying electronic databases such as texts and websites. The initial question answering system uses pre-set templates and built-in databases, and uses pre-defined rules to analyze problems and provide answers. This method is good for solving simple queries, but it is difficult to handle complex problems.

[0004] Since 2012, when Google proposed the concept of knowledge graph, the research focus of question answering systems has gradually shifted to the question answering method of related knowledge. Knowledge graph makes the question answering result systematic, related and visualized. Instead of returning a linear list of websites, the search request returns a complete knowledge system in a networked manner, greatly improving the quality of the answer returned by the search engine and the efficiency of user query information.

[0005] Although the knowledge graph is good at direct queries based on relevance, it does not have the ability of semantic analysis and is difficult to conduct natural language question answering. SUMMARY

[0006] The technical problem to be solved by the application is to provide a dam safety knowledge intelligent question answering method based on a knowledge graph and an AI large model in view of the above problems.

[0007] The technical solution adopted by the application is: a dam safety knowledge intelligent question answering method based on a knowledge graph and an AI large model, comprising: extracting retrieval keywords and sentence intentions from the natural language question through a trained dam safety semantic analysis professional large model; selecting elements corresponding to the retrieval keywords from each element of the dam safety knowledge graph as retrieval elements; based on the sentence intentions, selecting a corresponding knowledge graph Cypher query template, and combining the retrieval elements to generate a Cypher query command; Execute the Cypher query command to retrieve the answer to the natural language statement used to respond to the dam safety knowledge graph.

[0008] The training of the large-scale semantic analysis model for dam safety includes: Construct training data templates in the form of question-answer pairs, with natural language as input statements and statement intent and search keywords as return statements; According to the training data template, collect the question and answer statements that may be involved in the field of dam safety management, and convert them into the corresponding return statement format according to the training data template to form a training set. The training set is used to perform secondary training on the base model to obtain a trained professional large model for semantic parsing of dam safety.

[0009] The base model uses the Qwen3-8B pre-trained model.

[0010] The process of selecting elements corresponding to the search keywords from the various elements of the dam safety knowledge graph includes: The search keywords are segmented by moving a sliding window of length N to obtain a set of word segments for the search keywords; The elements in the dam safety knowledge graph are divided by moving a sliding window of length N to obtain a set of slice segments of candidate elements; Calculate the string distance between the set of keyword segments for the search and the set of segment segments for each candidate element; The candidate element with the smallest string distance to the search keyword is selected to obtain the element in the dam safety knowledge graph corresponding to the search keyword.

[0011] The intent of the statement includes entity attribute query, entity retrieval based on attribute, entity query with multiple attributes, entity retrieval with multiple attributes, entity one-hop relationship query, and entity two-hop relationship query.

[0012] The construction of the dam safety knowledge graph includes: The entities related to dam safety are mainly divided into basic project information, dam safety management entities, dam safety management behaviors, and typical dam cases; The basic information of the project includes the dam and the river; the main body of dam safety management includes relevant units and personnel responsible for dam safety; the dam safety management activities include all business activities involved in dam safety management; and the typical cases of dams include valuable engineering experience and typical cases accumulated in dam safety management activities.

[0013] The collection of typical dam case data used OCR technology to extract target data.

[0014] A smart question-answering system for dam safety knowledge based on knowledge graphs and AI big data models includes: The sentence extraction module is used to extract search keywords and sentence intent from natural language queries using a trained professional model for semantic parsing of dam safety. The element matching module is used to select elements that correspond to the search keywords from the elements of the dam safety knowledge graph, and use them as search elements. The command generation module is used to select the corresponding knowledge graph Cypher query template based on the statement intent and combine it with the search elements to generate Cypher query commands. The command execution module is used to execute Cypher query commands and retrieve answers from the dam safety knowledge graph to respond to natural language queries.

[0015] A storage medium storing a computer program executable by a processor, wherein the computer program, when executed, implements the steps of the intelligent question-and-answer method for dam safety knowledge based on knowledge graphs and AI large models.

[0016] A smart question-and-answer device for dam safety knowledge has a memory and a processor. The memory stores a computer program that can be executed by the processor. When the computer program is executed, it implements the steps of the smart question-and-answer method for dam safety knowledge based on knowledge graphs and AI large models.

[0017] The beneficial effects of this invention are as follows: This invention extracts retrieval keywords and sentence intent from natural language queries using a professional large-scale model for dam safety semantic parsing, selects retrieval elements corresponding to the retrieval keywords from the dam safety knowledge graph, selects the Cypher query template corresponding to the sentence intent from the knowledge graph, and generates a Cypher query command by combining the retrieval elements to retrieve the answer from the dam safety knowledge graph. This method combines knowledge graphs and AI large-scale models, which can simultaneously solve the shortcomings of natural language parsing of knowledge graphs and knowledge illusion of AI large-scale models, realize intelligent question answering of dam safety knowledge in the form of natural language input, and provide return results in the form of combined text and sub-graphs, solving the problems of poor correlation and inconvenient retrieval in traditional databases.

[0018] Based on a self-constructed standardized training data template, this invention utilizes the LoRA high-efficiency parameter fine-tuning technology of AI large models to perform secondary training on the pre-trained large model, forming a professional large model for dam safety semantic parsing, which realizes automatic parsing of natural language and extraction of key information.

[0019] This invention, based on the actual business of dam safety management, proposes a dam safety knowledge framework that includes four aspects: basic project information, dam safety management entities, dam safety management behaviors, and typical dam cases. It constructs a dam safety knowledge graph and realizes the structured expression of a large amount of professional heterogeneous data. Attached Figure Description

[0020] Figure 1 The flowchart is for an example.

[0021] Figure 2 This is the training dataset used in the example.

[0022] Figure 3 The example shows the change curve of the Loss index during the fine-tuning process. Figure 4 The curves showing the changes in the BLEU-4, ROUGE-1, ROUGE-2, and ROUGE-L indices during the fine-tuning process of the example are shown.

[0023] Figure 5 This is the Cypher query template for the knowledge graph in the example.

[0024] Figure 6 This example demonstrates a search result in the form of a combination of text and graphs. Detailed Implementation

[0025] The embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention. The step numbers in the following embodiments are set only for ease of explanation, and there is no limitation on the order between the steps. The execution order of each step in the embodiments can be adaptively adjusted according to the understanding of those skilled in the art.

[0026] In the description of this invention, "multiple" means two or more. The use of "first" and "second" is for distinguishing technical features only and should not be construed as indicating or implying relative importance, or implicitly indicating the number of indicated technical features, or the order of the indicated technical features. Furthermore, unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art.

[0027] Example 1: As Figure 1 As shown, this embodiment is an intelligent question-answering method for dam safety knowledge based on knowledge graphs and AI large models, specifically including the following steps: S100: Using a trained professional model for semantic parsing of dam safety, search keywords and sentence intent are extracted from natural language queries.

[0028] The training of the large-scale professional model for semantic parsing of dam safety in this embodiment includes the following steps: I. Construct training data templates in the form of question-and-answer pairs, with natural language as input statements and statement intent and search keywords as return statements.

[0029] In this embodiment, the statements are intended to represent the retrieval types in the knowledge graph, including entity attribute query, entity retrieval based on attribute, entity query with multiple attributes, entity retrieval with multiple attributes, entity one-hop relationship query, and entity two-hop relationship query.

[0030] In this example, special symbols are used to represent the intent of the statements. "%" represents an attribute query of an entity, "&" represents a search for an entity based on an attribute, "%%" represents a query for an entity with multiple attributes, "&&" represents a search for an entity with multiple attributes, "@" represents a one-hop query of an entity, and "@@" represents a two-hop query of an entity.

[0031] Search keywords are key information extracted from the input statement in natural language representation, which are used as entities, relations or attributes in subsequent knowledge graph query statements.

[0032] In this embodiment, if the input statement is "What is the height of the Longtan Dam?", the returned statement is "%, Longtan, Dam Height"; if the input statement is "How to deal with reservoir siltation?", the returned statement is "@, Reservoir Siltation, Treatment Measures".

[0033] II. Following the training data template, collect question-and-answer statements that may be relevant to dam safety management, convert them into corresponding return statement formats according to the training data template, and randomly divide them into training, validation, and test sets in an 8:1:1 ratio. A total of 81,005 training data entries were collected. Figure 2 As shown.

[0034] III. The training set is used to perform secondary training on the base model to obtain a trained professional large model for dam safety semantic parsing.

[0035] This embodiment selects the Qwen3-8B pre-trained model released by Tongyi Qianwen as the base model. The LoRA efficient parameter fine-tuning framework and training data are used to perform secondary training on the base model. During training, the Loss, BLEU-4, ROUGE-1, ROUGE-2, and ROUGE-L indices are recorded. Figure 3 , Figure 4 As shown.

[0036] The Loss metric is the average of the loss values ​​calculated by the AI ​​large model on each training batch, used to measure the difference between the model's predictions on the training data and the actual labels; the BLEU-4 metric is used to measure the accuracy of text generation, while the ROUGE-1, ROUGE-2, and ROUGE-L metric are used to measure the recall of text generation.

[0037] When the loss index decreases to a stable level, and the values ​​of BLEU-4, ROUGE-1, ROUGE-2, and ROUGE-L all reach above 85%, training can be stopped, forming a professional large-scale model for dam safety semantic parsing that can transform natural language queries into sentence intent and retrieve keyword representations.

[0038] S200. Select the elements corresponding to the search keywords from the elements of the dam safety knowledge graph and use them as search elements.

[0039] In this embodiment, a dam safety knowledge system framework is formulated, relevant data on dam safety management are collected and organized, and a dam safety knowledge graph is constructed.

[0040] In this example, based on the actual business of dam safety management, the entities related to dam safety are mainly divided into four aspects: basic project information, dam safety management entities, dam safety management behaviors, and typical dam cases.

[0041] The basic information of the project includes two entity categories: dams and rivers. Dams are the core of the knowledge system, while rivers serve as the connecting carriers between dams, linking the upstream and downstream relationships of each dam. The engineering characteristics of dams are stored as entity attributes, including dam name, installed capacity, dam height, reservoir capacity, and dam type. The attributes of rivers include river name, river basin, and river code.

[0042] The main entities responsible for dam safety management refer to relevant units and personnel involved in dam safety management. Relevant units encompass all entities involved throughout the entire lifecycle of the dam, from design and construction to operation and management. These entities primarily include design units, construction units, supervising units, and operating units. Dam safety management personnel refer to individuals working within the aforementioned units who participate in various aspects of dam safety management. These entities primarily include those responsible for monitoring and surveillance, emergency response, information reporting, and overall dam safety.

[0043] Dam safety management activities refer to all business activities involved in dam safety management. The main categories of entities include safety registration, periodic inspections, safety monitoring, safety surveillance, emergency management, and information reporting.

[0044] Typical dam cases are valuable engineering experiences and typical cases accumulated in dam safety management, including dam hidden dangers and dam emergency events, involving knowledge and data of the entire process from the causes of the event, its manifestations, to the handling measures and experience summaries.

[0045] In this embodiment, based on the storage format of dam safety management-related data, the knowledge collection method can be divided into two categories. The first category includes basic project information, dam safety management entities, and dam safety management behavior data. This data is stored in the form of databases, tables, lists, etc., and can be retrieved and organized in batches. The second category is typical dam case data. This data is stored in text form in professional reports, academic papers, and related websites, and is mainly extracted using OCR technology combined with manual methods.

[0046] In this example, referring to the dam safety knowledge system framework, the collected data is uniformly organized into triples of (head entity, relation, tail entity) or (entity, attribute, attribute value), such as (Lijiaxia Dam, dam height, 155m) and (overpass, hidden dangers, reservoir siltation). The Neo4j database is used to associate and store the above-mentioned triple data, forming a dam safety knowledge graph with a large amount of industry data.

[0047] In this embodiment, elements corresponding to the search keywords are selected from the elements of the dam safety knowledge graph, specifically including: (1) Search keywords K According to the length N The sliding window is used to move and segment the data, resulting in a set of word segments representing the search keywords. ; (2) Arrange all elements (entities, relations, or attributes) in the dam safety knowledge graph according to their length. N The sliding window is used to move and segment the data, resulting in a set of slices of candidate elements. .

[0048] (3) Calculate the set of keyword slices according to the following formula. Set of slices for each candidate element string distance : (1) in, Each represents a set of sliced ​​words. , The number of word segments, Represents a set of sliced ​​words. , The number of identical word segments in the string. String distance. The smaller the value, the more similar the search keyword is to the candidate element.

[0049] (4) Select the string distance from the search keyword. The smallest candidate element is used to obtain the corresponding retrieval element in the dam safety knowledge graph.

[0050] S300. Based on the statement intent, select the corresponding knowledge graph Cypher query template and generate a Cypher query command by combining the search elements.

[0051] This embodiment, based on defined knowledge graph retrieval types (entity attribute query, entity retrieval by attribute, entity query with multiple attributes, entity retrieval with multiple attributes, entity one-hop relationship query, entity two-hop relationship query), pre-designs fixed-format knowledge graph Cypher query templates to form a query module library, such as... Figure 5 As shown.

[0052] Based on the semantic analysis of the dam safety professional big data model, the corresponding Cypher query statement template is selected from the preset query module library, and the retrieval elements of the knowledge graph are populated to generate the Cypher query command.

[0053] S400, Execute the Cypher query command to retrieve the answers to the natural language questions from the dam safety knowledge graph and display them in a text-graph combination format, such as... Figure 6 As shown.

[0054] Example 2: An intelligent question-and-answer system for dam safety knowledge based on knowledge graphs and AI large models, comprising: The sentence extraction module is used to extract search keywords and sentence intent from natural language queries using a trained professional model for semantic parsing of dam safety. The element matching module is used to select elements that correspond to the search keywords from the elements of the dam safety knowledge graph, and use them as search elements. The command generation module is used to select the corresponding knowledge graph Cypher query template based on the statement intent and combine it with the search elements to generate Cypher query commands. The command execution module is used to execute Cypher query commands and retrieve answers from the dam safety knowledge graph to respond to natural language queries.

[0055] Example 3: This example is a storage medium that stores a computer program that can be executed by a processor. When the computer program is executed, it implements the steps of the intelligent question-and-answer method for dam safety knowledge based on knowledge graphs and AI big models described in Example 1.

[0056] Example 4: This example is an intelligent question-and-answer device for dam safety knowledge, which has a memory and a processor. The memory stores a computer program that can be executed by the processor. When the computer program is executed, it implements the steps of the intelligent question-and-answer method for dam safety knowledge based on knowledge graph and AI big model described in Example 1.

[0057] Furthermore, although the invention has been described in the context of functional modules, it should be understood that, unless otherwise stated, one or more of the aforementioned functions and / or features may be integrated into a single physical device and / or software module, or one or more functions and / or features may be implemented in a separate physical device or software module. It is also understood that a detailed discussion of the actual implementation of each module is unnecessary for understanding the invention. Rather, given the properties, functions, and internal relationships of the various functional modules in the apparatus disclosed herein, the actual implementation of the module will be understood within the scope of conventional skill of an engineer. Therefore, those skilled in the art can implement the invention as set forth in the claims using ordinary techniques without excessive experimentation. It is also understood that the specific concepts disclosed are merely illustrative and not intended to limit the scope of the invention, which is determined by the full scope of the appended claims and their equivalents.

[0058] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0059] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-including system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0060] More specific examples (a non-exhaustive list) of computer-readable media include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the aforementioned program can be printed, because the aforementioned program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0061] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0062] In the foregoing description of this specification, references to terms such as "one embodiment," "another embodiment," or "some embodiments" indicate that a specific feature, structure, material, or characteristic described in connection with an embodiment or example is included in at least one embodiment or example of the present invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0063] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

[0064] The above is a detailed description of the preferred embodiments of the present invention. However, the present invention is not limited to the above embodiments. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of the present invention. All such equivalent modifications or substitutions are included within the scope defined by the claims of this application.

Claims

1. A method for intelligent question answering of dam safety knowledge based on knowledge graphs and AI large models, characterized in that, include: By using a trained professional model for semantic parsing of dam safety, search keywords and sentence intent are extracted from natural language queries. Select elements from the dam safety knowledge graph that correspond to the search keywords as search elements; Based on the intent of the statement, select the corresponding knowledge graph Cypher query template and combine it with the search elements to generate a Cypher query command; Execute the Cypher query command to retrieve the answer to the natural language statement used to respond to the dam safety knowledge graph.

2. The intelligent question-answering method for dam safety knowledge based on knowledge graphs and AI large models according to claim 1, characterized in that, The training of the large-scale semantic analysis model for dam safety includes: Construct training data templates in the form of question-answer pairs, with natural language as input statements and statement intent and search keywords as return statements; According to the training data template, collect the question and answer statements that may be involved in the field of dam safety management, and convert them into the corresponding return statement format according to the training data template to form a training set. The training set is used to perform secondary training on the base model to obtain a trained professional large model for semantic parsing of dam safety.

3. The intelligent question-answering method for dam safety knowledge based on knowledge graphs and AI large models according to claim 2, characterized in that, The base model uses the Qwen3-8B pre-trained model.

4. The intelligent question-answering method for dam safety knowledge based on knowledge graphs and AI large models according to claim 1, characterized in that, The process of selecting elements corresponding to the search keywords from the various elements of the dam safety knowledge graph includes: The search keywords are segmented by moving a sliding window of length N to obtain a set of word segments for the search keywords; The elements in the dam safety knowledge graph are divided by moving a sliding window of length N to obtain a set of slice segments of candidate elements; Calculate the string distance between the set of keyword segments for the search and the set of segment segments for each candidate element; The candidate element with the smallest string distance to the search keyword is selected to obtain the element in the dam safety knowledge graph corresponding to the search keyword.

5. The intelligent question-answering method for dam safety knowledge based on knowledge graphs and AI large models according to claim 1, characterized in that, The intent of the statement includes entity attribute query, entity retrieval based on attribute, entity query with multiple attributes, entity retrieval with multiple attributes, entity one-hop relationship query, and entity two-hop relationship query.

6. The intelligent question-answering method for dam safety knowledge based on knowledge graphs and AI large models according to claim 1, characterized in that, The construction of the dam safety knowledge graph includes: The entities related to dam safety are mainly divided into basic project information, dam safety management entities, dam safety management behaviors, and typical dam cases; The basic information of the project includes the dam and the river; the main body of dam safety management includes relevant units and personnel responsible for dam safety; the dam safety management activities include all business activities involved in dam safety management; and the typical cases of dams include valuable engineering experience and typical cases accumulated in dam safety management activities.

7. The intelligent question-answering method for dam safety knowledge based on knowledge graphs and AI large models according to claim 6, characterized in that, The collection of typical dam case data used OCR technology to extract target data.

8. A smart question-and-answer system for dam safety knowledge based on knowledge graphs and AI large-scale models, characterized in that, include: The sentence extraction module is used to extract search keywords and sentence intent from natural language queries using a trained professional model for semantic parsing of dam safety. The element matching module is used to select elements that correspond to the search keywords from the elements of the dam safety knowledge graph, and use them as search elements. The command generation module is used to select the corresponding knowledge graph Cypher query template based on the statement intent and combine it with the search elements to generate Cypher query commands. The command execution module is used to execute Cypher query commands and retrieve answers from the dam safety knowledge graph to respond to natural language queries.

9. A storage medium having a computer program stored thereon that can be executed by a processor, characterized in that, When the computer program is executed, it implements the steps of the intelligent question-and-answer method for dam safety knowledge based on knowledge graphs and AI big models as described in any one of claims 1 to 7.

10. A smart question-and-answer device for dam safety knowledge, comprising a memory and a processor, wherein the memory stores a computer program executable by the processor, characterized in that, When the computer program is executed, it implements the steps of the intelligent question-and-answer method for dam safety knowledge based on knowledge graphs and AI big models as described in any one of claims 1 to 7.

Images