A multi-level collaborative urban physical examination knowledge reasoning method and related equipment

Abstract

The present invention provides a multi-level collaborative urban physical examination knowledge reasoning method and related equipment, which obtain a knowledge base and question request constructed by an urban physical examination knowledge graph; input the question request into an entity-level reasoning module for entity and attribute recognition to obtain an initial evidence triplet, and perform knowledge mining on the initial evidence triplet according to the knowledge base to obtain an evidence triplet; input the question request and the evidence triplet into a function-level reasoning module for recognition to obtain the urban physical examination analysis function required in the question request, and call a spatiotemporal analysis tool according to the urban physical examination analysis function to perform spatiotemporal analysis to obtain an evaluation analysis result; input the question request, evidence triplet, knowledge base and evaluation analysis result into a decision-level reasoning module to perform knowledge reasoning on the question request to obtain a reasoning result; the problems of insufficient depth of knowledge mining, low tool arrangement efficiency and insufficient precision of knowledge reasoning in the urban physical examination knowledge reasoning process are solved.

Description

Technical Field

[0001] The present invention relates to the field of knowledge reasoning technology, and in particular to a multi-level collaborative urban physical examination knowledge reasoning method and related equipment. Background Art

[0002] The core task of urban health examination knowledge reasoning is to use knowledge engineering and reasoning technology to conduct knowledge mining, tool arrangement and knowledge reasoning on urban health examination spatiotemporal data, so as to extract high-level knowledge information, improve analysis and evaluation efficiency, and promote decision-making implementation. Under the knowledge-guided reasoning mechanism, knowledge is extracted from urban health examination spatiotemporal data to support improvement measures and development strategies for urban development; therefore, effective urban health examination knowledge reasoning has become one of the keys to transforming urban health examination knowledge into decision-making.

[0003] In recent years, in order to transform the results of urban physical examinations into specific and executable decision-making plans, it has become an inevitable trend for urban physical examination knowledge reasoning to shift to deep knowledge mining, automated tool arrangement and refined decision execution. Deep knowledge mining means deeply analyzing the core of spatiotemporal data, capturing implicit patterns and associations, and providing data empirical support for decision-making; automated tool arrangement means promoting the automation of decision-making processes through efficient collaboration of various tools, automated processing and analysis of spatiotemporal data; refined task execution emphasizes micro-control to ensure that each step achieves the expected goals and improves the implementation of decisions.

[0004] However, existing research and past experience are unable to meet the needs of urban health check decision support. There are many challenges. The reasons are mainly reflected in the following three aspects: (1) Insufficient knowledge mining. Traditional data reasoning processing relies on keyword matching, while knowledge graph technology achieves more complex relational pattern reasoning by building explicit associations and reasoning relationships between data. However, current knowledge graph reasoning is often limited to data associations within the same level, lacks attention to cross-level reasoning, fails to reveal the deep connections and patterns behind cross-level urban health check spatiotemporal data, and cannot directly adapt to the complex patterns of urban health check application scenarios. For example, although a single environmental indicator query can quickly return results, it lacks in-depth analysis of the causes, impact range, and interaction between the indicators and applicable scenarios, which limits the comprehensive understanding and effective response to urban problems. (2) Inefficient tool arrangement. There is still a lack of automated in-depth analysis of complex urban problems in terms of quantitative analysis and tool application for urban health check. At present, there are many tools and models involved in urban health check, and the connection between them often relies on manual serial or parallel operations, lacking automated arrangement and integration. (3) Extensive knowledge reasoning. In the past, the decision-making process was too general, often relying on manual analysis and experience judgment, lacking in-depth analysis and process processing of decision-making tasks, and lacking flexible and clear execution paths and evaluation standards for instructions, which led to the extensive and inefficient transformation of decision support applications. For example, the urban physical examination goal setting is not clear, the task decomposition is not refined, and the instruction execution path is not flexible, which affects the precise execution of decisions and limits the transformation of urban physical examination knowledge into decisions. Therefore, how to promote the coordinated urban physical examination knowledge reasoning of deepening knowledge mining, automated tool arrangement and improving knowledge reasoning to comprehensively improve the execution efficiency of urban physical examination decisions has become the key to current research. Summary of the invention

[0005] The present invention provides a multi-level collaborative urban physical examination knowledge reasoning method and related equipment, which aims to solve the problems of insufficient depth of knowledge mining, low tool arrangement efficiency and insufficient precision of knowledge reasoning in the urban physical examination knowledge reasoning process.

[0006] In order to achieve the above object, the present invention provides a multi-level collaborative urban physical examination knowledge reasoning method, comprising:

[0007] Step 1, obtaining a knowledge base constructed by a city physical examination knowledge graph and question requests related to data in the knowledge base, wherein the city physical examination knowledge graph includes city physical examination spatiotemporal data;

[0008] Step 2: Input the question request into the entity-level reasoning module in the trained large language model to perform entity and attribute recognition to obtain an initial evidence triple, and perform knowledge mining on the initial evidence triple according to the knowledge base to obtain an evidence triple;

[0009] Step 3: Input the question request and the evidence triples into the function-level reasoning module in the trained large language model for recognition, obtain the city physical examination analysis function required in the question request, and call the spatiotemporal analysis tool to perform spatiotemporal analysis based on the city physical examination analysis function to obtain the evaluation analysis result;

[0010] Step 4: Input the question request, evidence triples, knowledge base, and evaluation analysis results into the decision-level reasoning module in the trained large language model, perform knowledge reasoning on the question request through the evidence triples, knowledge base, and evaluation analysis results, and obtain the reasoning result.

[0011] Further, the question request is a question request predefined according to the content in the knowledge base or a question request generated by a large language model according to the predefined question request and knowledge base exploration, and the question request is related to the data in the knowledge base.

[0012] Furthermore, before step 2, it also includes:

[0013] Build a large language model including entity-level reasoning module, function-level reasoning module, and decision-level reasoning module;

[0014] The large language model is trained using the knowledge base and the question request set used for training to obtain a trained large language model.

[0015] More specifically, step 2 includes:

[0016] Input the question request into the entity-level reasoning module of the trained large language model;

[0017] The entity-level reasoning module performs entity recognition on the question request to obtain data entities, knowledge point entities, and scene entities, and returns the initial evidence triplet;

[0018] In the entity-level reasoning module, the knowledge graph entities are generated through the relationship paths between entities in the initial evidence triples, and the attribute feature information corresponding to the knowledge graph entities is extracted;

[0019] The initial evidence triples are mined based on the urban physical examination knowledge graph in the knowledge base to obtain evidence triples.

[0020] More specifically, step 3 includes:

[0021] Input the question request and evidence triples into the function-level reasoning module in the trained large language model;

[0022] In the function-level reasoning module, the question request is identified to obtain the city physical examination analysis function required in the question request. The city physical examination analysis function includes indicator analysis, spatiotemporal prediction, and policy evidence.

[0023] In the function-level reasoning module, the analysis task is decomposed according to the city physical examination analysis function to obtain multiple subtasks, and a tool interface module is assigned to each subtask;

[0024] After reading the functions of all tool interface modules, the function-level reasoning module plans the tool chain to interactively analyze the problem request and obtain the evaluation and analysis results.

[0025] Further, the expression for evaluating the analysis result is:

[0026]

[0027]

[0028] in, Indicates the evaluation and analysis results. Represents a processing function that generates a corresponding output response based on the input information. Represents the result of functional analysis and reasoning, Represents the tool interface module, Indicates a question request The city physical examination analysis function required in Indicates a question request, represents the sequential steps of the toolchain requested for the issue, represents the planning toolchain, represents the feature-level inference function, Represents a collection of question requests.

[0029] More specifically, step 4 includes:

[0030] Given the city physical examination goals and urban planning, determine the execution decision steps;

[0031] Input the question request, evidence triples, knowledge base, and evaluation analysis results into the decision-level reasoning module in the trained large language model;

[0032] In the decision-level reasoning module, the problem request is transformed into multiple decision subtasks through evidence triples, knowledge base and evaluation analysis results;

[0033] In the decision-level reasoning module, the execution path is generated by selecting the required decision steps in the execution decision steps according to each decision subtask;

[0034] In the decision-level reasoning module, each decision subtask is executed based on the execution path to obtain multiple decision analysis results, and all decision analysis results are integrated to obtain the reasoning result.

[0035] The present invention also provides a multi-level collaborative urban physical examination knowledge reasoning device, comprising:

[0036] An acquisition module is used to acquire a knowledge base constructed by a city physical examination knowledge graph and question requests related to data in the knowledge base, wherein the city physical examination knowledge graph includes city physical examination spatiotemporal data;

[0037] The knowledge mining module is used to input the question request into the entity-level reasoning module in the trained large language model to perform entity and attribute recognition to obtain the initial evidence triples, and perform knowledge mining on the initial evidence triples according to the knowledge base to obtain evidence triples;

[0038] The spatiotemporal analysis module is used to input the question request and evidence triples into the function-level reasoning module in the trained large language model for recognition, obtain the city physical examination analysis function required in the question request, and call the spatiotemporal analysis tool to perform spatiotemporal analysis based on the city physical examination analysis function to obtain the evaluation analysis results;

[0039] The knowledge reasoning module is used to input the question request, evidence triples, knowledge base and evaluation analysis results into the decision-level reasoning module in the trained large language model, and perform knowledge reasoning on the question request through the evidence triples, knowledge base and evaluation analysis results to obtain the reasoning result.

[0040] The present invention also provides a terminal device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements a multi-level collaborative urban physical examination knowledge reasoning method when executing the computer program.

[0041] The present invention also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, a multi-level collaborative urban physical examination knowledge reasoning method is implemented.

[0042] The above scheme of the present invention has the following beneficial effects:

[0043] The present invention obtains a knowledge base constructed by a city physical examination knowledge graph and a question request related to the data in the knowledge base; inputs the question request into the entity-level reasoning module in the trained large language model to perform entity and attribute recognition to obtain an initial evidence triple, performs knowledge mining on the initial evidence triple according to the knowledge base to obtain an evidence triple; inputs the question request and the evidence triple into the function-level reasoning module in the trained large language model to perform recognition to obtain the city physical examination analysis function required in the question request, and calls the spatiotemporal analysis tool to perform spatiotemporal analysis according to the city physical examination analysis function to obtain an evaluation analysis result; inputs the question request, evidence triple into the function-level reasoning module in the trained large language model to perform recognition to obtain the city physical examination analysis function required in the question request, and calls the spatiotemporal analysis tool to perform spatiotemporal analysis according to the city physical examination analysis function to obtain an evaluation analysis result; According to the triples, knowledge base and evaluation analysis results, they are all input into the decision-level reasoning module in the trained large language model, and the question request is subjected to knowledge reasoning through the evidence triples, knowledge base and evaluation analysis results to obtain the reasoning result. Compared with the prior art, the present invention performs deep mining through the entity-level reasoning module in the large language model, calls the tool through the function-level reasoning module, and provides precise guidance through the decision-level reasoning module, thereby providing a systematic solution for urban physical examination knowledge reasoning, and solves the problems of insufficient depth of knowledge mining, low tool arrangement efficiency and insufficient precision of knowledge reasoning in the process of urban physical examination knowledge reasoning.

[0044] Other beneficial effects of the present invention will be described in detail in the subsequent specific implementation section. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 It is a schematic diagram of a flow chart of an embodiment of the present invention;

[0046] Figure 2 Schematic diagram of the structure of the urban physical examination knowledge reasoning device in an embodiment of the present invention;

[0047] Figure 3 Schematic diagram of the structure of a terminal device in an embodiment of the present invention. DETAILED DESCRIPTION

[0048] In order to make the technical problems, technical solutions and advantages to be solved by the present invention clearer, the following will be described in detail with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0049] In the description of the present invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc., indicating the orientation or positional relationship, are based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", and "third" are used for descriptive purposes only, and cannot be understood as indicating or implying relative importance.

[0050] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installation", "connection" and "connection" should be understood in a broad sense, for example, it can be a locking connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0051] In addition, the technical features involved in the different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0052] In view of the existing problems, the present invention provides a multi-level collaborative urban physical examination knowledge reasoning method and related equipment.

[0053] like Figure 1 As shown, an embodiment of the present invention provides a multi-level collaborative urban physical examination knowledge reasoning method, including:

[0054] Step 1, obtaining a knowledge base constructed by a city physical examination knowledge graph and question requests related to data in the knowledge base, wherein the city physical examination knowledge graph includes city physical examination spatiotemporal data;

[0055] Step 2: Input the question request into the entity-level reasoning module in the trained large language model to perform entity and attribute recognition to obtain an initial evidence triple, and perform knowledge mining on the initial evidence triple according to the knowledge base to obtain an evidence triple;

[0056] Step 3: Input the question request and the evidence triples into the function-level reasoning module in the trained large language model for recognition, obtain the city physical examination analysis function required in the question request, and call the spatiotemporal analysis tool to perform spatiotemporal analysis based on the city physical examination analysis function to obtain the evaluation analysis result;

[0057] Step 4: Input the question request, evidence triples, knowledge base, and evaluation analysis results into the decision-level reasoning module in the trained large language model, perform knowledge reasoning on the question request through the evidence triples, knowledge base, and evaluation analysis results, and obtain the reasoning result.

[0058] In an embodiment of the present invention, the spatiotemporal data of urban health examinations are derived from multiple channels, which include but are not limited to a city statistical yearbook, project documents of urban planning and construction, environmental data reports, etc. These spatiotemporal data cover a longer period of time in the time dimension, such as records of city-related data in the past ten years or even longer. Therefore, urban health examination indicators can be analyzed in time series, such as the city's gross domestic product growth trend, changes in economic development speed, etc. In the spatial dimension, the data covers different areas of a city, from the municipal level to the district level, which helps to analyze the differences and distribution patterns of urban health examination indicators in different regions, such as the environmental quality conditions in different regions, differences in construction levels, etc.

[0059] Specific indicators of urban health examination can be the annual growth rate of regional GDP, air quality index (AQI), etc. By analyzing these indicators, we can understand the development status and existing problems of the city.

[0060] Specifically, the question request is a question request predefined according to the content in the knowledge base or a question request generated by a large language model according to the predefined question request and knowledge base exploration, and the question request is related to the data in the knowledge base.

[0061] Specifically, before step 2, it also includes:

[0062] Build a large language model including entity-level reasoning module, function-level reasoning module, and decision-level reasoning module;

[0063] The large language model is trained using the knowledge base and the question request set used for training to obtain a trained large language model.

[0064] In the embodiment of the present invention, the entity-level reasoning module, the function-level reasoning module, and the decision-level reasoning module are all soft modules and are all mounted on one or more hardware structures with processing functions in the large language model.

[0065] In the embodiment of the present invention, the entities, attributes and relationships in the urban physical examination knowledge graph are systematically organized at the input stage of the entity-level reasoning module in the large language model, and this information is converted into text form. These texts are vectorized and encoded as input to the large language model, so that the model can understand the hierarchical characteristics of the scenes, knowledge points and data in the urban physical examination. During the processing, the large language model conducts an in-depth analysis of the input vectorized encoded text, and learns the specific language patterns and knowledge structures of urban physical examinations through fine-tuning, aiming to enable the large language model to better understand the professional context and knowledge structure in the field of urban physical examinations. By learning the entity relationships in the knowledge graph, the large language model can perform reasoning and decision support more accurately. This process not only involves the classification and identification of entities, such as urban themes, knowledge points, indicators, etc., but also the extraction of detailed attribute information of entities, such as ID name, spatial range, time range and data source, as well as explicit relationships between entities, such as support, measurement, calculation, expression and subordination; in the output stage, by deeply mining the relationship between "scenario-knowledge point-data", high-level knowledge is extracted from spatiotemporal data of different levels and dimensions, including hierarchical structures and association rules.

[0066] In response to question requests, the large language model and the city physical examination knowledge graph maximize the probability distribution To generate a response, where represents the parameters of the large language model, It represents the city physical examination knowledge graph, and then due to the length limit of the input context of the large language model, it is impossible to use the entire city physical examination knowledge graph to construct prompts, which undoubtedly increases the complexity of the reasoning process. For this reason, an embodiment of the present invention proposes a deep chain knowledge reasoning method, which draws on the chain knowledge reasoning, and inputs the question request into the entity-level reasoning module in the trained large language model for entity and attribute recognition to obtain the initial evidence triples, and performs knowledge mining on the initial evidence triples according to the knowledge base to obtain evidence triples.

[0067] Specifically, the calculation expression for maximizing the probability distribution is:

[0068]

[0069]

[0070] in, Request a collection for an issue A question in the request, The response answer generated by the model, Represents the knowledge graph of urban physical examination, which is an important source of knowledge in the reasoning process. is a function representing the entity-level reasoning method, Represents the knowledge graph of city physical examination The subgraph generated in contains entities related to the problem and their relationship paths. Used to describe specific relationships between entities, such as support, measurement, etc.

[0071] In the embodiment of the present invention, the question request is input into the entity-level reasoning module in the trained large language model to perform entity and attribute recognition to obtain an initial evidence triple, and knowledge mining is performed on the initial evidence triple according to the knowledge base to obtain the evidence triple, which specifically includes:

[0072] Input the question request into the entity-level reasoning module of the trained large language model;

[0073] The entity-level reasoning module performs entity recognition on the question request to obtain data entities, knowledge point entities, and scene entities, and returns the initial evidence triplet;

[0074] In the entity-level reasoning module, the knowledge graph entities are generated through the relationship paths between entities in the initial evidence triples, and the attribute feature information corresponding to the knowledge graph entities is extracted;

[0075] The initial evidence triples are mined based on the urban physical examination knowledge graph in the knowledge base to obtain evidence triples.

[0076] Specifically, the embodiment of the present invention uses "Are there any areas in a certain city that have particularly outstanding performance in terms of annual growth rate of regional GDP in 2022?" as a question request;

[0077] The entity-level reasoning module performs entity recognition on the question request to obtain data entities, knowledge point entities, and scene entities, and returns the initial evidence triples as follows:

[0078] (Annual growth rate of regional GDP, corresponding regions, industrial areas)

[0079] (Annual growth rate of regional GDP, corresponding regions, trade zones)

[0080] (Annual growth rate of regional GDP, corresponding regions, emerging industrial zones)

[0081] (Annual growth rate of regional GDP, corresponding regions, business districts);

[0082] In the entity-level reasoning module, the knowledge graph entity is generated through the relationship path between entities in the initial evidence triples, and the attribute feature information corresponding to the knowledge graph entity is extracted as follows:

[0083] {ID: Statistical chart, space: a certain city area, time: 2022, source: Statistical Yearbook};

[0084] According to the urban physical examination knowledge graph in the knowledge base, the initial evidence triples are mined and the evidence triples are obtained as follows:

[0085] (Regional GDP statistics chart, chart, data)

[0086] (Data, support, economic vitality scenario of a certain urban area)

[0087] (Economic vitality scenario of a certain urban area, with measurement, annual growth rate of regional GDP)

[0088] (Scenario of economic vitality in a certain urban area, with expression, knowledge points on scientific and technological innovation and regional economic vitality).

[0089] Specifically, step 3 includes:

[0090] Input the question request and evidence triples into the function-level reasoning module in the trained large language model;

[0091] In the function-level reasoning module, the question request is identified to obtain the city physical examination analysis function required in the question request. The city physical examination analysis function includes indicator analysis, spatiotemporal prediction, and policy evidence.

[0092] In the function-level reasoning module, the analysis task is decomposed according to the city physical examination analysis function to obtain multiple subtasks, and a tool interface module is assigned to each subtask;

[0093] After reading the functions of all tool interface modules, the function-level reasoning module plans the tool chain to interactively analyze the problem request and obtain the evaluation and analysis results.

[0094] Specifically, in the input stage of the function-level reasoning module, the functions required for the city physical examination are clarified starting from the question request, such as indicator analysis, spatiotemporal prediction, policy evidence, etc. The indicator analysis includes the annual growth rate of regional GDP, high-tech enterprise data, the number of national science and technology enterprise incubators, the amount of entrepreneurship subsidies, air quality index, etc. The request for spatiotemporal prediction includes time string, administrative district string, prediction string, etc. The request for policy evidence includes: government documents issued by the state, province, city, and county, government work reports, departmental regulations, etc.; then the question request and the corresponding evidence triplet are input into the function-level reasoning module to prompt the function-level reasoning module to design a complete tool interface module for the request. Each of these modules corresponds to an interface and is stored in the dependent tool interface library for the large language model to use first in each iteration to guide the selection of appropriate analysis tools and design dependencies between tools. In the face of requests that cannot be solved through existing interfaces, the function-level reasoning module will design new interfaces to meet the needs; after reading the functions of all tool interface modules, the function-level reasoning module plans the tool chain to interactively analyze the question request and obtain the evaluation analysis results; formally, the expression of the evaluation analysis results is:

[0095]

[0096]

[0097] in, Indicates the evaluation and analysis results. Represents a processing function, which is used to generate corresponding output responses based on input information. The results of functional analysis and reasoning are the answers to questions, analysis of relevant data, evaluation reports and other information needed for further decision-making. Represents the tool interface module, Indicates a question request The city physical examination analysis function required in Indicates a question request, represents the sequential steps of the toolchain requested for the issue, represents the planning toolchain, represents the feature-level inference function, Represents a collection of question requests.

[0098] It should be noted that the tool interface module mentioned in the embodiment of the present invention can be regarded as a code module composed of a name, parameters, function description and implementation code, which can perform specific tasks such as data retrieval, calculation and visualization.

[0099] In the processing of the function-level reasoning module, considering that the tools involved in different requests are significantly different, not all tool interface modules need to be loaded every time. Therefore, the embodiment of the present invention subdivides complex analysis tasks into manageable subtasks based on the input functional requirements, and assigns corresponding tool interface modules to each subtask; after reading all retrieved interface descriptions, the function-level reasoning module plans a tool chain to interactively analyze user requests; the tool chain interactive analysis is output in JSON format: ToolInterface Call: {step1 = {"arg": "","Interface":"", "output": "","desc":""}, step2 = {..},..}; In the output stage of the function-level reasoning module, the evaluation and analysis results are integrated and converted to form intuitive charts and evaluation reports. These charts and reports show the current status and changing trends of urban physical examination indicators, and also provide detailed descriptions of policy basis.

[0100] It should be noted that the tool chain includes spatiotemporal data acquisition tools, such as Table_Extract for extracting specific spatiotemporal data tables from data cubes, overall trend forecasting and analysis tools, such as Overall_Trend for analyzing the overall trend presented by a single-column pd.DataFrame (i.e., the passed-in Series parameter) composed of single-series indicator values ​​indexed by year, trend feature analysis tools, such as Trend_Features for calculating the trend feature points described by the indicator value pd.DataFrame indexed by year under a specific administrative division, indicator value periodicity and anomaly analysis tools, such as Cycle_Detect for detecting the periodicity of given time series data, policy analysis tools, such as Policy_Analysis for conducting in-depth analysis of urban physical examination-related policies by using year or administrative division as index, and plotting tools, such as Plot_Data for generating statistical charts based on the input single-column indicator value pandas.DataFrame (data_to_plot).

[0101] It should be noted that tool chain interaction analysis refers to multiple tool interface modules called in a certain order to form a chain, parallel or complex loop structure; in the chain mode, the output of one tool is directly used as the input of the next tool. In the parallel mode, multiple tools run simultaneously, and their outputs are aggregated to the next stage. According to the logical relationship between tools, including preconditions, post-dependencies, data flow order, etc., the appropriate analysis tool is automatically selected and called.

[0102] For example, the embodiment of the present invention takes "Are there any areas in a city that have particularly outstanding performance in the annual growth rate of regional GDP in 2022?" as the question request; the evaluation and analysis result obtained is "Based on the data in 2022, the trade zone performed most outstandingly in the annual growth rate of regional GDP, reaching 4.20%, followed by the industrial zone at 2.40%, the emerging industry zone at 1.50%, and the commercial zone at 1.40%".

[0103] Specifically, step 4 includes:

[0104] Given the city physical examination goals and urban planning, determine the execution decision steps;

[0105] Input the question request, evidence triples, knowledge base, and evaluation analysis results into the decision-level reasoning module in the trained large language model;

[0106] In the decision-level reasoning module, the problem request is transformed into multiple decision subtasks through evidence triples, knowledge base and evaluation analysis results;

[0107] In the decision-level reasoning module, the execution path is generated by selecting the required decision steps in the execution decision steps according to each decision subtask;

[0108] In the decision-level reasoning module, each decision subtask is executed based on the execution path to obtain multiple decision analysis results, and all decision analysis results are integrated to obtain the reasoning result.

[0109] The best option is to take innovation vitality and ecological livability as examples, and the city physical examination goals include:

[0110] 1. Innovative vitality-innovative development

[0111] 2. Innovation Vitality - Urban Vitality

[0112] 3. Innovation Vitality-Economic Development

[0113] 4. Ecological Livability - Green and Low-Carbon Development

[0114] 5. Ecological Livability - Ecological Environmental Protection

[0115] 6. Ecological livability - beautiful living space.

[0116] Most preferably, the inference result is expressed as:

[0117]

[0118]

[0119] in, Represents the inference result, including a comprehensive decision analysis of the problem, Represents the decision library for a given problem The decision database stores a large amount of decision-making knowledge, expert experience and decision-making rules related to urban physical examination. These nodes are various decision-making elements, such as urban development goals, relevant policies and regulations, historical decision-making cases, evaluation criteria, etc. Related nodes provide rich reference information for generating decision answers. For a given problem The execution decision steps plan the The sequential process of decision analysis. Function (usually a generator LM) used to generate responses. Represents the decision-level reasoning function, and converts the decision library ,question and execute the decision-making steps As input, the generator LM is used to generate a set of questions The decision model of the knowledge reasoning process obtained after processing , integrating decision logic and information of the decision-making process.

[0120] Specifically, the decision-making steps include:

[0121] 1. Goal Analysis: The decision-level reasoning module first analyzes the decision goals, understands the intentions behind them and the target outcomes corresponding to the city physical examination theme; formally, this process is defined as ,in It is the target corresponding to the theme of city physical examination. is a given query question, Represents the question-topic answer pair associated with the city physical examination knowledge graph;

[0122] 2. Connotation and indicators: The decision-level reasoning module determines the connotation and key indicators related to the goal; formally, this process is defined as ,in is the currently generated city physical examination target, It is a given query problem, which is executed by conditional judgment ,like Query indicators, then Indicates that the corresponding connotation and index need to be retrieved; if If no index is queried, Indicates that no retrieval is required;

[0123] 3. Indicator quantification / data analysis: The decision-level reasoning module quantifies and analyzes these indicators and processes the relevant data; formally, this process is defined as ,in It is the result of indicator quantification and data analysis. is the retrieved index (if any), is a given query question, Indicates the data analysis results in the evaluation analysis results;

[0124] 4. Compliance requirements: Based on the data analysis results, the decision-level reasoning module evaluates whether the indicators of the city physical examination meet the preset standards; formally, this process is defined as ,in is the currently generated data analysis result, is a given query problem (query indicator), under the given compliance requirements Is a Boolean value indicating whether the standard is met;

[0125] 5. Practice: The decision-level reasoning module examines the implementation of these indicators in actual operations; formally, this process is defined as ,in A statement indicating the performance of the indicator. is a given query question, is the retrieved index (if any), It is the generated compliance status.

[0126] 6. Action: Based on the problems found in the city physical examination, formulate corresponding urban renewal strategies and action plans; formally, this process is defined as ,in is the currently generated city physical examination target, is the execution status of the currently generated indicator, Represents the policy analysis results in the evaluation analysis results.

[0127] It should be noted that if there is no task to be executed in the above steps, it will not be included in the task path.

[0128] At the output stage of the decision-level reasoning module, the aggregator combines multiple decision analysis results to generate a coherent and comprehensive response to the original query, ensuring that the final response processes the original query in a balanced and coherent manner; formally, ,in It is a combination of Is the query set It is a collection of answers generated by each decision-making step; through the decision-level reasoning module, the standards and requirements of a healthy city can be clarified, the gap between the current and the target can be quantitatively analyzed, and the effectiveness of the implementation of existing measures can be evaluated.

[0129] Specifically, the embodiment of the present invention uses "Are there any areas in a certain city that have particularly outstanding performance in terms of annual growth rate of regional GDP in 2022?" as the question request; the inference results obtained are as follows:

[0130] 1. Target analysis

[0131] Technological innovation in a certain city has promoted the development of advantageous industries and future industries such as high-tech industries, strategic emerging industries, and modern service industries, and increased the quality and efficiency of economic growth.

[0132] 2. Content and indicators

[0133] To promote sustained, inclusive and sustainable economic growth and achieve full and productive employment and decent work. This goal focuses on analyzing major economic activities and economic growth rates, including the annual growth rate of regional GDP, and is assessed using this indicator.

[0134] 3. Quantification of indicators

[0135] The annual growth rate of regional GDP is quantified using statistical methods. Table 1 shows the quantification results and evaluation reference of this indicator:

[0136] Table 1 Quantitative results of indicators and evaluation reference table

[0137]

[0138] 4. Compliance requirements

[0139] The annual growth rate of regional GDP remained above the target.

[0140] 5. Implementation

[0141] Trade Zone: In 2022, the GDP of the Trade Zone increased by A%, ranking first in the city. The Trade Zone is an industrial zone after the emerging industrial zone, and its industrial output value above designated size ranks second in the city.

[0142] Industrial Zone: The GDP of the industrial zone will grow by B% in 2022. The added value of its modern service industry accounts for about 70% of GDP, which is an important driving force of the economy.

[0143] Emerging Industrial Zones: In 2022, the GDP of emerging industrial zones increased by C% year-on-year. The emerging industrial zones formed several 100 billion-level and 50 billion-level industrial clusters. The total investment in 2022 ranked first among the national economic and technological development zones and all districts in the province.

[0144] Commercial area: In 2022, the GDP of the commercial area will increase by D% year-on-year.

[0145] 6. Actions

[0146] Trade Zone: Promote comprehensive cooperation and regional development among surrounding cities

[0147] Executing department: Development and Reform Commission of a city

[0148] Measures: Formulate and implement the "Action Plan for Deepening Comprehensive Cooperation with Surrounding Cities in a Certain City Trade Zone Facing the World", coordinate and promote the implementation of various tasks, establish and improve the work system, issue the district action plan and key task list, and strengthen the dispatch and implementation of tasks. At the same time, implement financial, factor and policy support guarantees, promote the construction of science and technology innovation industry cooperation bases, accelerate the construction of major science and technology innovation platforms, expand the scale of high-tech industries, attract investment and talents, and promote rapid economic growth.

[0149] Commercial District: Promote the agglomeration of financial industries and the construction of legal districts

[0150] Executing department: Local Financial Supervision and Administration Bureau of a certain city

[0151] Measures: Relying on major platforms such as the International Financial City, vigorously attract legal person financial institutions, foreign financial institutions, etc., accelerate the construction of the Central Legal District and high-end districts, promote industrial upgrading and transformation, and improve the quality and efficiency of economic development. At the same time, fully stimulate the vitality of corporate innovation entities, support industry leaders to take the lead in establishing innovation alliances, promote the deep integration of innovation chains and industrial chains, and further enhance the region's innovation capabilities and competitiveness.

[0152] Industrial Zone: Increase Industrial Investment and Industrial Upgrading

[0153] Executing department: Industry and Information Technology Bureau of a certain city

[0154] Measures: Continue to leverage the advantages of the industrial zone, increase investment and support for industry, promote the development and upgrading of industrial clusters, and increase the total industrial output value and added value. At the same time, strengthen the integration with scientific and technological innovation, encourage enterprises to increase R&D investment, improve independent innovation capabilities, and promote the development of the industry towards high-end, intelligent, and green directions.

[0155] Emerging Industrial Zones: Optimizing Industrial Structure and Fixed Asset Investment

[0156] Executing department: A city’s Bureau of Commerce

[0157] Measures: Further optimize the industrial structure, increase the cultivation and support of emerging industries, and promote the transformation and upgrading of traditional industries. At the same time, strengthen the guidance and management of fixed asset investment, improve the efficiency and quality of investment, and promote stable economic growth. In addition, it is necessary to strengthen services and support for enterprises, help them solve practical difficulties, and improve their competitiveness and profitability.

[0158] In order to verify the performance of the entire method, the embodiment of the present invention removes different reasoning levels and their combinations through ablation experiments, and analyzes the effects on decision accuracy and knowledge reasoning. The specific process is as follows:

[0159] First, the impact of removing a single reasoning level. (1) Removing the decision-level reasoning module: When the entity-level and function-level reasoning are kept normal, the decision accuracy is significantly reduced. This shows that the decision-level reasoning module plays an indispensable role in comprehensively considering the analysis results, refining the execution decision, and accurately evaluating the compliance and actual implementation of the city physical examination indicators. Its absence will lead to the inability to give reasonable decision-making suggestions. (2) Removing the function-level reasoning module: When the entity-level reasoning module and the decision-level reasoning module are normal, removing the function-level reasoning module will significantly reduce the decision accuracy, highlighting that the function-level reasoning provides key support for obtaining information such as the compliance requirements, change trends and policy basis of the city physical examination indicators through automated tool orchestration and accurate analysis. The lack of it will seriously affect the quality of decision-making. (3) Removing entity-level reasoning. When the function-level and decision-level reasoning are operating normally, removing the entity-level reasoning has a great impact on the accuracy of knowledge reasoning. Because entity-level reasoning is the key link in extracting entities and their attributes and relationships, it is the knowledge basis for subsequent function-level and decision-level reasoning. The absence of it will make the subsequent reasoning lose an important basis. In addition, the impact of the combination of the two reasoning levels is removed. The results showed that decision accuracy decreased significantly more than the effect of removing a single level of reasoning, which again emphasizes the importance of showing that the three levels of reasoning work together.

[0160] Table 2 Ablation experiment results

[0161]

[0162] As can be seen from the EM value and accuracy data in Table 2 above, when there is no ablation, the EM value reaches 0.89 and the accuracy reaches 0.92, showing good performance. After various ablation settings, these indicators all show different degrees of decline, which further confirms the indispensability of each reasoning level module in the method provided by the embodiment of the invention from a quantitative perspective, and their multi-level synergy ensures the effectiveness of the method.

[0163] The embodiment of the present invention obtains a knowledge base constructed by a city physical examination knowledge graph and a question request related to the data in the knowledge base; inputs the question request into the entity-level reasoning module in the trained large language model to perform entity and attribute recognition to obtain an initial evidence triple, performs knowledge mining on the initial evidence triple according to the knowledge base to obtain an evidence triple; inputs the question request and the evidence triple into the function-level reasoning module in the trained large language model for recognition to obtain the city physical examination analysis function required in the question request, and calls the spatiotemporal analysis tool to perform spatiotemporal analysis according to the city physical examination analysis function to obtain an evaluation analysis result; and inputs the question request, evidence triple into the function-level reasoning module in the trained large language model for recognition to obtain the city physical examination analysis function required in the question request, and calls the spatiotemporal analysis tool to perform spatiotemporal analysis according to the city physical examination analysis function to obtain an evaluation analysis result; The evidence triples, knowledge base and evaluation analysis results are input into the decision-level reasoning module in the trained large language model, and knowledge reasoning is performed on the question request through the evidence triples, knowledge base and evaluation analysis results to obtain the reasoning result; compared with the prior art, the embodiment of the present invention performs deep mining through the entity-level reasoning module in the large language model, calls the tool through the function-level reasoning module, and provides precise guidance through the decision-level reasoning module, thereby providing a systematic solution for urban physical examination knowledge reasoning, and solves the problems of insufficient depth of knowledge mining, low tool arrangement efficiency and insufficient precision of knowledge reasoning in the process of urban physical examination knowledge reasoning.

[0164] Corresponding to the multi-level collaborative urban physical examination knowledge reasoning method described in the above embodiment, Figure 2 As shown, the embodiment of the present invention further provides a multi-level collaborative urban physical examination knowledge reasoning device 100, and the urban physical examination knowledge reasoning device 100 includes:

[0165] An acquisition module 101 is used to acquire a knowledge base constructed by a city physical examination knowledge graph and question requests related to data in the knowledge base, wherein the city physical examination knowledge graph includes city physical examination spatiotemporal data;

[0166] The knowledge mining module 102 is used to input the question request into the entity-level reasoning module in the trained large language model to perform entity and attribute recognition to obtain an initial evidence triple, and perform knowledge mining on the initial evidence triple according to the knowledge base to obtain an evidence triple;

[0167] The spatiotemporal analysis module 103 is used to input the question request and the evidence triple into the function-level reasoning module in the trained large language model for recognition, obtain the city physical examination analysis function required in the question request, and call the spatiotemporal analysis tool to perform spatiotemporal analysis according to the city physical examination analysis function to obtain the evaluation analysis result;

[0168] The knowledge reasoning module 104 is used to input the question request, evidence triples, knowledge base and evaluation analysis results into the decision-level reasoning module in the trained large language model, perform knowledge reasoning on the question request through the evidence triples, knowledge base and evaluation analysis results, and obtain reasoning results.

[0169] It should be noted that the information interaction, execution process, etc. between the above-mentioned devices / units are based on the same concept as the method embodiment of the present application. Their specific functions and technical effects can be found in the method embodiment part and will not be repeated here.

[0170] The technicians in the relevant field can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In practical applications, the above-mentioned function allocation can be completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated in a processing unit, or each unit can exist physically separately, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, which will not be repeated here.

[0171] The embodiment of the present invention also provides a terminal device, such as Figure 3 As shown, the terminal device D10 of this embodiment includes: at least one processor D100 ( Figure 3 Only one processor is shown in the figure), a memory D101, and a computer program D102 stored in the memory D101 and executable on the at least one processor D100, wherein the processor D100 implements the above-mentioned multi-level collaborative urban physical examination knowledge reasoning method when executing the computer program D102.

[0172] The terminal device D10 may be a computing device such as a desktop computer, a notebook, a PDA, a server, a server cluster, a cloud server, etc. The terminal device may include, but is not limited to, a processor D100 and a memory D101. Those skilled in the art will appreciate that Figure 3 This is only an example of the terminal device D10 and does not constitute a limitation on the terminal device D10. The terminal device D10 may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, it may also include input and output devices, network access devices, etc.

[0173] The processor D100 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc.

[0174] In some embodiments, the memory D101 may be an internal storage unit of the terminal device D10, such as a hard disk or memory of the terminal device D10. In other embodiments, the memory D101 may also be an external storage device of the terminal device D10, such as a plug-in hard disk, a smart memory card (SMC, SmartMedia Card), a secure digital (SD, Secure Digital) card, a flash card (Flash Card), etc. equipped on the terminal device D10. Further, the memory D101 may also include both an internal storage unit of the terminal device D10 and an external storage device. The memory D101 is used to store an operating system, an application program, a boot loader (BootLoader), data, and other programs, such as the program code of the computer program, etc. The memory D101 may also be used to temporarily store data that has been output or is to be output.

[0175] It should be noted that the information interaction, execution process, etc. between the above-mentioned devices / units are based on the same concept as the method embodiment of the present application. Their specific functions and technical effects can be found in the method embodiment part and will not be repeated here.

[0176] The technicians in the relevant field can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In practical applications, the above-mentioned function allocation can be completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated in a processing unit, or each unit can exist physically separately, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, which will not be repeated here.

[0177] An embodiment of the present invention further provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, a multi-level collaborative urban physical examination knowledge reasoning method is implemented.

[0178] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present application implements all or part of the processes in the above-mentioned embodiment method, which can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may at least include: any entity or device that can carry the computer program code to the construction device / terminal device, recording medium, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, RandomAccess Memory), electric carrier signal, telecommunication signal and software distribution medium. For example, a USB flash drive, a mobile hard disk, a disk or an optical disk.

[0179] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A multi-level collaborative urban physical examination knowledge reasoning method, characterized in that: include: Step 1, obtaining a knowledge base constructed by a city physical examination knowledge graph and a question request related to the data in the knowledge base, wherein the city physical examination knowledge graph includes city physical examination spatiotemporal data; Step 2, inputting the question request into the entity-level reasoning module in the trained large language model to perform entity and attribute recognition to obtain an initial evidence triple, and performing knowledge mining on the initial evidence triple according to the knowledge base to obtain an evidence triple; Step 3: Input the question request and the evidence triplet into the function-level reasoning module in the trained large language model for identification, obtain the city physical examination analysis function required in the question request, and call the spatiotemporal analysis tool to perform spatiotemporal analysis according to the city physical examination analysis function to obtain the evaluation analysis results, including: Inputting the question request and the evidence triple into a function-level reasoning module in the trained large language model; The problem request is identified in the function-level reasoning module to obtain the city physical examination analysis function required in the problem request, wherein the city physical examination analysis function includes indicator analysis, spatiotemporal prediction, and policy evidence; In the functional level reasoning module, the analysis task is decomposed according to the city physical examination analysis function to obtain multiple subtasks, and a tool interface module is assigned to each subtask; After reading the functions of all tool interface modules, the function-level reasoning module plans the tool chain to interactively analyze the problem request to obtain an evaluation analysis result; The expression of the evaluation analysis result is: in, Indicates the evaluation and analysis results. Represents a processing function that generates a corresponding output response based on the input information. Represents the result of functional analysis and reasoning, Represents the tool interface module, Indicates a question request The city physical examination analysis function required in Indicates a question request, represents the sequential steps of the toolchain requested for the issue, represents the planning toolchain, represents the function-level inference function, Represents a collection of question requests; Step 4: input the question request, the evidence triple, the knowledge base, and the evaluation and analysis result into the decision-level reasoning module in the trained large language model, and perform knowledge reasoning on the question request through the evidence triple, the knowledge base, and the evaluation and analysis result to obtain a reasoning result, including: Given the city physical examination goals and urban planning, determine the execution decision steps; Inputting the question request, the evidence triple, the knowledge base and the evaluation analysis result into a decision-level reasoning module in the trained large language model; In the decision-level reasoning module, the question request is transformed and decomposed into a plurality of decision subtasks through the evidence triples, the knowledge base and the evaluation analysis results; In the decision-level reasoning module, a required decision step is selected from the execution decision steps according to each decision subtask to generate an execution path; In the decision-level reasoning module, each decision subtask is executed based on the execution path to obtain multiple decision analysis results, and all decision analysis results are integrated to obtain a reasoning result.

2. The multi-level collaborative urban physical examination knowledge reasoning method according to claim 1 is characterized in that: The question request is a question request predefined according to the content in the knowledge base or a question request generated by a large language model according to a predefined question request and the knowledge base exploration, and the question request is related to the data in the knowledge base.

3. The multi-level collaborative urban physical examination knowledge reasoning method according to claim 2 is characterized in that: Before step 2, the method further includes: Build a large language model including entity-level reasoning module, function-level reasoning module, and decision-level reasoning module; The large language model is trained using the knowledge base and the question request set for training to obtain a trained large language model.

4. The multi-level collaborative urban physical examination knowledge reasoning method according to claim 3 is characterized in that: The step 2 comprises: Inputting the question request into the entity-level reasoning module in the trained large language model; Performing entity recognition on the question request through the entity-level reasoning module to obtain data entities, knowledge point entities, and scene entities, and returning an initial evidence triplet; In the entity-level reasoning module, a knowledge graph entity is generated through the relationship path between entities in the initial evidence triples, and attribute feature information corresponding to the knowledge graph entity is extracted; Knowledge mining is performed on the initial evidence triples according to the city physical examination knowledge graph in the knowledge base to obtain evidence triples.

5. A multi-level collaborative urban physical examination knowledge reasoning device, characterized in that: include: An acquisition module, used to acquire a knowledge base constructed by a city physical examination knowledge graph and question requests related to data in the knowledge base, wherein the city physical examination knowledge graph includes city physical examination spatiotemporal data; A knowledge mining module is used to input the question request into the entity-level reasoning module in the trained large language model to perform entity and attribute recognition to obtain an initial evidence triple, and perform knowledge mining on the initial evidence triple according to the knowledge base to obtain an evidence triple; The spatiotemporal analysis module is used to input the question request and the evidence triple into the function-level reasoning module in the trained large language model for identification, obtain the city physical examination analysis function required in the question request, and call the spatiotemporal analysis tool to perform spatiotemporal analysis according to the city physical examination analysis function to obtain evaluation analysis results, including: Inputting the question request and the evidence triple into a function-level reasoning module in the trained large language model; The problem request is identified in the function-level reasoning module to obtain the city physical examination analysis function required in the problem request, wherein the city physical examination analysis function includes indicator analysis, spatiotemporal prediction, and policy evidence; In the functional level reasoning module, the analysis task is decomposed according to the city physical examination analysis function to obtain multiple subtasks, and a tool interface module is assigned to each subtask; After reading the functions of all tool interface modules, the function-level reasoning module plans the tool chain to interactively analyze the problem request to obtain an evaluation analysis result; The expression of the evaluation analysis result is: in, Indicates the evaluation and analysis results. Represents a processing function that generates a corresponding output response based on the input information. Represents the result of functional analysis and reasoning, Represents the tool interface module, Indicates a question request The city physical examination analysis function required in Indicates a question request, represents the sequential steps of the toolchain requested for the issue, represents the planning toolchain, represents the function-level inference function, Represents a collection of question requests; A knowledge reasoning module is used to input the question request, the evidence triple, the knowledge base and the evaluation and analysis result into the decision-level reasoning module in the trained large language model, and perform knowledge reasoning on the question request through the evidence triple, the knowledge base and the evaluation and analysis result to obtain a reasoning result, including: Given the city physical examination goals and urban planning, determine the execution decision steps; Inputting the question request, the evidence triple, the knowledge base and the evaluation analysis result into a decision-level reasoning module in the trained large language model; In the decision-level reasoning module, the question request is transformed and decomposed into a plurality of decision subtasks through the evidence triples, the knowledge base and the evaluation analysis results; In the decision-level reasoning module, a required decision step is selected from the execution decision steps according to each decision subtask to generate an execution path; In the decision-level reasoning module, each decision subtask is executed based on the execution path to obtain multiple decision analysis results, and all decision analysis results are integrated to obtain a reasoning result.

6. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the multi-level collaborative urban physical examination knowledge reasoning method as described in any one of claims 1 to 4 is implemented.

7. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the multi-level collaborative urban physical examination knowledge reasoning method as described in any one of claims 1 to 4 is implemented.

Images