Report number asking method and device based on PLM dimension index self-sensing and multi-layer alignment

Abstract

The invention provides a report number asking method and device based on PLM dimension index self-sensing and multi-layer alignment, and relates to the technical field of intelligent report number asking. The method comprises the steps that when a number asking request of a user is received, element disassembling is carried out on the number asking request based on a large model and a pre-constructed dimension index vector library, and the number asking request of the user is obtained; obtaining dimension index information corresponding to the number asking request; performing index matching according to the dimension index information corresponding to the question number request and an obtained index library, and determining a native index corresponding to the question number request; and calling a data query interface to obtain a corresponding data result based on the native index corresponding to the question number request, and returning a query result to the user. According to the method and the device, automatic standardization from the native index to the structured dimension or the index is realized, and the number asking intention of the user is accurately mapped to the target index, so that the accuracy, the robustness and the service adaptation capability of the number asking of the report are remarkably improved under the condition that a number bin or an index system does not need to be reconstructed.

Description

Technical Field

[0001] This application relates to the field of intelligent report query technology, and in particular to a report query method and apparatus based on PLM dimension index self-sensing and multi-layer alignment. Background Technology

[0002] As enterprises continuously improve their digitalization and refined management, data-driven decision-making has become an important approach to business management. Reporting systems, as core tools for business data analysis and presentation, directly impact the realization of data value through their query capabilities. Existing query solutions for report indicator libraries primarily model the indicators and dimensions within the reporting system, enabling users to obtain the necessary business data through natural language. However, these solutions still have certain limitations in practical applications.

[0003] One existing technology is a knowledge graph-based reporting and data collection solution. This solution constructs a semantic network containing entities and their relationships by structurally modeling elements such as dimensions, metrics, and business objects in a report metric library. When a user inputs a natural language question, the system performs semantic parsing, maps the parsed entities and relationships to corresponding nodes and relationships in the knowledge graph, and retrieves the target metric data through graph querying or reasoning. However, due to the large number of dimensions and metrics in the reporting system, the construction and maintenance of the knowledge graph requires significant manual intervention, resulting in high modeling and implementation costs. Furthermore, with frequent business changes, the entities and relationships in the knowledge graph need continuous adjustment, making the maintenance process complex and prone to inconsistencies or errors, affecting system stability and data collection effectiveness.

[0004] Another type of existing technology is the intelligent query solution based on indicator middleware or data warehouse reconstruction. This solution typically reconstructs the enterprise's existing data warehouse, using the indicator middleware as the core, re-standardizing data models, indicator definitions, and calculation logic, centralizing complex processing logic to a wide surface layer, and training models on this basis to achieve natural language querying. While this type of solution improves the consistency of indicator definitions and query efficiency to some extent, it requires large-scale adjustments to the existing data warehouse system, involving data migration, system transformation, and business process reconstruction. Implementation costs are high, the cycle is long, and it has a certain impact on the enterprise's existing business operations. At the same time, the data warehouse reconstruction process is technically complex, requiring high professional capabilities from the implementation team, and the project risks and uncertainties are significant.

[0005] Therefore, existing reporting and data query solutions generally suffer from problems such as high modeling and maintenance costs, strong dependence on existing system modifications, and insufficient adaptability to business changes. They are difficult to achieve efficient and accurate responses to user queries without changing the existing reporting system and data architecture.

[0006] This section is intended to provide background or context for the embodiments of the invention set forth in the claims. The description herein is not an admission that it is prior art simply because it is included in this section. Summary of the Invention

[0007] To address the problems in the existing technology, this application provides a report query method and apparatus based on PLM dimension indicator self-awareness and multi-layer alignment. This method can solve the problems in existing report systems, such as non-standard indicator naming, missing semantic information, and difficulty in accurately matching native indicators with users' natural language query needs. These issues lead to the reliance on manual rules for indicator positioning in query scenarios, resulting in insufficient accuracy and scalability.

[0008] One aspect of the present invention provides a report query method based on PLM dimensional indicator self-awareness and multi-level alignment, the method comprising: Upon receiving a user's data query request, the data query request is decomposed into elements based on the large model and a pre-built dimensional indicator vector library to obtain the dimensional indicator information corresponding to the data query request. Based on the dimension indicator information corresponding to the question number request and the obtained indicator library, indicator matching is performed to determine the native indicator corresponding to the question number request; Based on the native metrics corresponding to the query request, the data query interface is called to obtain the corresponding data results, and the query results are returned to the user.

[0009] Furthermore, the step of pre-constructing the dimensional indicator vector library includes: The obtained report templates are parsed based on the large model to generate structured dimensional indicator information; A dimensional indicator vector library is generated based on the structured dimensional indicator information and the native indicators of the report template.

[0010] Further, the step of generating a dimensional indicator vector library based on the structured dimensional indicator information and the native indicators of the report template includes: After the structured dimensional indicator information is manually verified and confirmed to be acceptable, the mapping relationship between the report template and the structured dimensional indicator information is determined. Based on the mapping relationship, the original indicators of the report template are aligned to the corresponding structured dimensional indicator information to obtain the aligned dimensional indicator information. The aligned dimensional index information is then cleaned. Write the cleaned dimensional indicator information into the dimensional indicator vector library.

[0011] Furthermore, the question number request is decomposed into elements based on the large model and the pre-built dimensional indicator vector library to obtain the dimensional indicator information corresponding to the question number request, including: The large model is invoked by a preset first prompt word to perform an element decomposition on the question request and obtain the user intent index. Based on the user intent metric, the dimensional metric vector library, and the large model, the dimensional metric information corresponding to the question number request is generated.

[0012] Further, the step of generating the dimensional indicator information corresponding to the question count request based on the user intent indicator, the dimensional indicator vector library, and the large model includes: Based on the user intent metrics, vector matching is performed in the dimension metric vector library; Construct known dimension indicator information based on vector matching results; Based on the known dimensional indicator information, the large model is invoked through a preset second prompt word to perform secondary element decomposition on the question number request, thereby obtaining the dimensional indicator information corresponding to the question number request.

[0013] Further, the step of matching metrics based on the dimension metric information corresponding to the question count request and the obtained metric library to determine the native metric corresponding to the question count request includes: Based on the dimension indicator information corresponding to the question number request, a match is performed in the indicator library to obtain candidate native indicators; Based on preset matching rules and business rules, the candidate native metrics are matched a second time to determine the native metrics corresponding to the question number request.

[0014] Furthermore, it also includes: Construct known indicator data based on the native indicators corresponding to the question requests; Based on the known indicator data and the number of questions, the large model performs relevance recommendations on the original indicators and generates corresponding indicator calculation models.

[0015] Another aspect of the present invention provides a report query device based on PLM dimension index self-awareness and multi-level alignment, the device comprising: The element decomposition unit is used to decompose the question request based on the large model and the pre-built dimension indicator vector library when it receives the user's question request, so as to obtain the dimension indicator information corresponding to the question request. The indicator matching unit is used to perform indicator matching based on the dimension indicator information corresponding to the question number request and the obtained indicator library to determine the native indicator corresponding to the question number request. The data query unit is used to call the data query interface to obtain the corresponding data results based on the native indicators corresponding to the query request, and return the query results to the user.

[0016] Furthermore, it also includes: The structured information generation unit is used to parse the acquired report templates based on the large model and generate structured dimensional indicator information. The vector library generation unit is used to generate a dimension indicator vector library based on the structured dimension indicator information and the native indicators of the report template.

[0017] Furthermore, the vector library generation unit includes: The mapping relationship generation module is used to determine the mapping relationship between the report template and the structured dimensional indicator information after the structured dimensional indicator information has been manually verified and confirmed to be valid. The indicator alignment module is used to align the native indicators of the report template to the corresponding structured dimensional indicator information based on the mapping relationship, so as to obtain the aligned dimensional indicator information. The data cleaning module is used to clean the aligned dimensional indicator information. The vector library generation module is used to write the cleaned dimensional indicator information into the dimensional indicator vector library.

[0018] Furthermore, the element decomposition unit includes: The user intent index generation module is used to call the large model through a preset first prompt word, perform an element decomposition on the question request, and obtain the user intent index. The dimensional indicator information generation module is used to generate dimensional indicator information corresponding to the question number request based on the user intent indicator, the dimensional indicator vector library, and the large model.

[0019] Furthermore, the dimensional indicator information generation module includes: The vector matching submodule is used to perform vector matching in the dimension indicator vector library based on the user intent indicator; The information construction submodule is used to construct known dimension indicator information based on vector matching results; The dimension indicator information generation submodule is used to, based on the known dimension indicator information, call the large model through a preset second prompt word to perform secondary element decomposition on the question number request and obtain the dimension indicator information corresponding to the question number request.

[0020] Furthermore, the indicator matching unit includes: The first matching module is used to perform a matching in the indicator library based on the dimension indicator information corresponding to the question number request to obtain candidate native indicators. The second matching module is used to perform secondary matching on the candidate native metrics based on preset matching rules and business rules to determine the native metrics corresponding to the question number request.

[0021] Furthermore, it also includes: The data construction unit is used to construct known indicator data based on the native indicators corresponding to the question requests; The model recommendation unit is used to make relevant recommendations for the original indicators based on the known indicator data and the number of questions, and to generate the corresponding indicator calculation model.

[0022] To achieve the above objectives, according to another aspect of the present invention, a computer device is also provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described report query method based on PLM dimension index self-awareness and multi-layer alignment.

[0023] To achieve the above objectives, according to another aspect of the present invention, a computer-readable storage medium is also provided, on which a computer program / instruction is stored, which, when executed by a processor, implements the steps of the above-described report query method based on PLM dimension index self-awareness and multi-layer alignment.

[0024] To achieve the above objectives, according to another aspect of the present invention, a computer program product is also provided, including a computer program / instruction that, when executed by a processor, implements the steps of the above-described report query method based on PLM dimension index self-awareness and multi-layer alignment.

[0025] The beneficial effects of this invention are as follows: This invention introduces the semantic awareness capability of large models to automatically parse the vast number of reports with diverse structures in business systems. It can autonomously identify the dimension and indicator information contained in the report, and modeling can be completed with only manual confirmation of the results. This significantly reduces the workload of manually sorting out dimensions, indicators and their relationships, and significantly improves the efficiency of indicator governance and report configuration. At the same time, by aligning the native indicators in business systems to a unified structured dimension / indicator system, it achieves standardization and consistency of indicator expression in different systems and report formats, avoiding misunderstandings caused by scattered indicator definitions and inconsistent naming. Furthermore, this invention, through a secondary element decomposition mechanism, further refines the alignment of user intent metrics with structured dimensions / metrics after initially identifying the user's query intent. This enables more accurate parsing of the user's true query needs, thereby improving the accuracy of native metric matching and the reliability of query results. Simultaneously, through deep integration with business systems, this invention can automatically identify and respond to the data and organizational permission rules of business systems during the query process. It controls data according to the different user's permission range, effectively preventing unauthorized access and data leakage risks, and improving the security and availability of the report query system in actual business scenarios. Attached Figure Description

[0026] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. In the drawings: Figure 1 This is a schematic diagram of the first process of the report query method based on PLM dimension index self-perception and multi-layer alignment provided in the embodiment of the present invention; Figure 2 This is a schematic diagram of the second process of the report query method based on PLM dimension index self-perception and multi-layer alignment provided in the embodiment of the present invention; Figure 3 This is a schematic diagram of the third process of the report query method based on PLM dimension index self-sensing and multi-layer alignment provided in the embodiment of the present invention; Figure 4 This is a schematic diagram of the fourth process of the report query method based on PLM dimension index self-perception and multi-layer alignment provided in the embodiment of the present invention; Figure 5 This is a schematic diagram of the fifth process of the report query method based on PLM dimension index self-sensing and multi-layer alignment provided in the embodiment of the present invention; Figure 6 This is a schematic diagram of the sixth process of the report query method based on PLM dimension index self-sensing and multi-layer alignment provided in the embodiment of the present invention; Figure 7 This is a schematic diagram of the seventh process of the report query method based on PLM dimension index self-perception and multi-layer alignment provided in the embodiment of the present invention; Figure 8 This is a schematic block diagram of the first structure of the report query device based on PLM dimension index self-sensing and multi-layer alignment provided in the embodiment of the present invention; Figure 9 This is a schematic block diagram of the second structure of the report query device based on PLM dimension index self-sensing and multi-layer alignment provided in the embodiment of the present invention; Figure 10 This is a schematic block diagram of the third structure of the report query device based on PLM dimension index self-sensing and multi-layer alignment provided in the embodiments of the present invention; Figure 11 This is a schematic block diagram of the fourth structure of the report query device based on PLM dimension index self-sensing and multi-layer alignment provided in the embodiments of the present invention; Figure 12 This is a fifth structural schematic block diagram of the report query device based on PLM dimension index self-sensing and multi-layer alignment provided in the embodiments of the present invention; Figure 13 This is a sixth structural schematic block diagram of the report query device based on PLM dimension index self-sensing and multi-layer alignment provided in the embodiments of the present invention; Figure 14 This is the seventh structural schematic block diagram of the report query device based on PLM dimension index self-sensing and multi-layer alignment provided in the embodiments of the present invention; Figure 15 This is a schematic diagram of the structure of the computer device provided in an embodiment of the present invention; Figure 16 This is a schematic diagram of the overall process of the report query method based on PLM dimension index self-perception and multi-layer alignment provided in the embodiment of the present invention; Figure 17 This is a schematic diagram of the process for generating a dimensional indicator vector library provided in an embodiment of the present invention; Figure 18 This is a schematic diagram of the secondary element decomposition process provided in an embodiment of the present invention; Figure 19 This is a flowchart illustrating the matching of native metrics provided in an embodiment of the present invention. Detailed Implementation

[0027] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0028] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0029] It should be noted that the terms "comprising" and "having" and any variations thereof in the specification, claims and accompanying drawings of this invention are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products or devices.

[0030] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0031] In business systems, the creation and maintenance of multidimensional reports exhibit significant diversity. Specifically, the same business metric may be created multiple times in different reports under the same name but with different codes, while the same data retrieval formula is used during actual data collection. Furthermore, the naming conventions for report metrics are highly arbitrary, with some metric names failing to accurately reflect their true business meaning, or even having their semantics hidden or missing. For example, "Original Value of Fixed Assets for Public Institutions at the End of the Period" might be simply named "Original Value of Fixed Assets" in a report, failing to clearly express key information such as "Public Institutions" and "End of Period," making it impossible to accurately locate the corresponding business metric in data retrieval scenarios. These issues make it difficult for existing business systems to meet the requirements for accurate and reliable data retrieval in natural language data retrieval and intelligent analysis scenarios.

[0032] To address the common problems existing in the aforementioned business systems and to better achieve query performance, this invention proposes a report query method and apparatus based on pre-trained language model (PLM) dimensionality self-awareness and multi-layer alignment, aiming to more efficiently and accurately meet the query requirements in complex business scenarios. This solution is implemented through the following technical measures: 1. Unified parsing method: Based on the report template information, the semantic awareness capability of the large model is introduced to automatically identify and parse the dimension and indicator information contained in the template; the system further processes the perceived dimension / indicator information to generate standardized dimension indicators, thereby achieving a unified semantic representation under different reports and different indicator expression methods.

[0033] 2. Create a business knowledge base: Transform complex business logic embedded in the business system into rules that the model can understand and call, and store them accordingly. For example, in the scenario of calculating loan delinquency rates, the definition criteria for delinquency days and the calculation formula for delinquency rates can be clearly defined in the rule base. This allows the model to refer to the business knowledge base when parsing query requirements and accurately generate target indicators consistent with business standards.

[0034] 3. Establish an indicator library: Systematically clean and manage the indicator names and related information in the business system, and formulate corresponding handling strategies for different types of data quality problems. Specifically, for data redundancy issues, duplicate indicators are removed using data deduplication techniques; for data inconsistency issues, unified data standards are established to standardize indicator coding and data formats; and for insufficient data timeliness, the indicator library's data is kept timely and effective through scheduled or incremental update mechanisms.

[0035] 4. Access Control Mechanism: By adding access filtering interfaces to the business system, fine-grained control over user query permissions is achieved. This not only determines whether a user has permission to access a specific indicator but also further narrows down the range of data tables involved in the indicator query. For example, in a reporting system with multiple data table types such as individual household tables, summary tables, and analysis tables, frontline staff may have query permissions for individual household tables, while higher-level departments may not. When higher-level departments query relevant indicators, they only need to search from the summary and analysis tables, thereby effectively reducing the risk of data leakage and improving query efficiency while meeting business needs.

[0036] like Figure 16 As shown, this invention employs a technical framework based on PLM-based dimensional indicator self-awareness and multi-layer alignment. By introducing a large model to build dimensional / indicator self-awareness capabilities, the server can automatically identify and extract the dimensional and indicator elements contained in the report template information, generating structured dimensional / indicator information. Subsequently, the server further analyzes and processes the perceived dimensional / indicator information to form a unified and standardized dimensional indicator representation.

[0037] During the query process, the server combines the user's input question content with pre-built business knowledge, and performs secondary element decomposition on the user's query request based on a large model. The semantic alignment of the indicators involved in the user's intent is aligned into corresponding structured dimensions / indicator information. After completing the semantic alignment, the server further matches and infers with the original indicators in the business based on the aligned dimensions / indicators to determine the target indicators that match the user's query intent.

[0038] After determining the target native metrics, the server connects with the business logic, calls the data query interface to obtain the corresponding native metric data, processes and renders the query results, and presents them to the user in the form of reports or visualizations on the interface, thus realizing a complete processing flow from natural language query to data result display.

[0039] The following describes the specific implementation process of the report query method based on PLM dimension index self-awareness and multi-layer alignment provided in this application embodiment, taking the server as the execution subject as an example.

[0040] Figure 1 This is a schematic diagram of the first process of the report query method based on PLM dimension index self-awareness and multi-level alignment provided in an embodiment of the present invention, as shown below. Figure 1 As shown, in one embodiment of the present invention, the report query method based on PLM dimension index self-awareness and multi-level alignment includes: S101: Upon receiving a user's question request, the question request is decomposed into elements based on the large model and a pre-built dimensional indicator vector library to obtain the dimensional indicator information corresponding to the question request; S102: Based on the dimension indicator information corresponding to the question number request and the obtained indicator library, perform indicator matching to determine the native indicator corresponding to the question number request; S103: Based on the native metrics corresponding to the query request, call the data query interface to obtain the corresponding data results, and return the query results to the user.

[0041] from Figure 1 As shown in the flowchart, the report query method based on PLM dimension indicator self-awareness and multi-layer alignment provided in this application, when receiving a user's query request, decomposes the query request into elements based on a large model and a pre-built dimension indicator vector library to obtain the dimension indicator information corresponding to the query request; performs indicator matching based on the dimension indicator information corresponding to the query request and the obtained indicator library to determine the native indicator corresponding to the query request; and calls the data query interface to obtain the corresponding data results based on the native indicator corresponding to the query request, and returns the query results to the user. This achieves automatic standardization from native indicators to structured dimensions or indicators, and accurately maps the user's query intent to the target indicator, thereby significantly improving the accuracy, robustness, and business adaptability of report queries without reconstructing the data warehouse or indicator system.

[0042] Each step is explained in detail below.

[0043] S101: Upon receiving a user's question request, the question request is decomposed into elements based on the large model and a pre-built dimensional indicator vector library to obtain the dimensional indicator information corresponding to the question request; Specifically, when the server receives a query request from a user through the client, it first processes the query request. The query request is a report query statement entered by the user in natural language, expressing the user's query needs for business data.

[0044] The server, based on a large model and a pre-built dimensional indicator vector library, performs element decomposition processing on the query request. The server inputs the query request into the large model, which analyzes the semantic content of the query request, identifying the indicator elements, dimensional elements, and their corresponding value information. At the same time, the server combines the semantic features of dimensional indicators stored in the dimensional indicator vector library to provide semantic assistance and supplementation to the analysis results of the large model, thereby obtaining the dimensional indicator information corresponding to the query request.

[0045] Figure 2 This is a schematic diagram of the second process of the report query method based on PLM dimension index self-awareness and multi-level alignment provided in an embodiment of the present invention, as shown below. Figure 2 As shown, in one embodiment of the present invention, the step of pre-constructing the dimensional indicator vector library includes: S201: Based on the large model, the obtained report templates are parsed to generate structured dimensional indicator information; Specifically, the server first retrieves report template information from the business system. The report template describes the structure of the reports in the business system and the metrics, dimension fields, and display rules they contain. Its form may include, but is not limited to, report configuration files, report templates, or report field definition information.

[0046] The server parses and processes the report templates based on a large model. The server inputs the report templates into the large model, which analyzes the template content from a business semantic perspective, identifying the indicator and dimensional semantics contained within the templates. The identification results are then output in a structured format, generating structured dimensional indicator information. This structured dimensional indicator information is used to uniformly express the implicit business meanings in the report templates, avoiding inconsistencies in indicator semantics caused by differences in templates.

[0047] Using the above method, the server can automatically parse a large number of report templates and generate unified and standardized dimensional indicator descriptions without requiring manual configuration.

[0048] S202: Generate a dimension indicator vector library based on the structured dimension indicator information and the native indicators of the report template.

[0049] Specifically, after generating structured dimensional indicator information, the server further combines the corresponding native indicator information in the report template to construct a dimensional indicator vector library. Native indicators are actual indicator fields or calculated items existing in the business system, used to support the statistics and display of report data.

[0050] Based on structured dimensional indicator information and native indicators, the server uniformly organizes the semantic content of the dimensional indicators and generates corresponding semantic feature representations for subsequent semantic analysis and matching processing. Then, the server writes the structured dimensional indicators and their corresponding semantic feature information into a dimensional indicator vector library for storage.

[0051] The dimension indicator vector library constructed in the above manner can centrally store the semantic information of various dimensions and indicators in the business system, providing efficient and unified semantic support for the server when processing user query requests, thereby improving the accuracy and consistency of query request element decomposition.

[0052] Figure 3 This is a schematic diagram of the third process of the report query method based on PLM dimension index self-awareness and multi-layer alignment provided in the embodiments of the present invention, as follows: Figure 3 As shown, in one embodiment of the present invention, S202 includes: S301: After the structured dimensional indicator information is manually verified and confirmed to be acceptable, the mapping relationship between the report template and the structured dimensional indicator information is determined. Specifically, the server provides the structured dimensional indicator information obtained from parsing the report template to humans for verification and confirmation. Human verification is used to determine whether the structured dimensional indicator information conforms to the actual business semantics and indicator definition requirements, avoiding errors in the understanding of dimensions or indicators due to model recognition bias.

[0053] After manual verification confirms the structured dimension indicator information is acceptable, the server determines the mapping relationship between the report template and the structured dimension indicator information. The mapping relationship describes the correspondence between the native indicators, fields, or display items in the report template and the structured dimension indicators, thus providing a clear association basis for subsequent indicator semantic processing.

[0054] S302: Based on the mapping relationship, align the original indicators of the report template to the corresponding structured dimensional indicator information to obtain the aligned dimensional indicator information; Specifically, after determining the mapping relationship, the server aligns the native metrics in the report templates to the corresponding structured dimension metrics information based on the mapping relationship. The alignment process is used to uniformly map native metrics with inconsistent formats and naming in different report templates into a standardized structured dimension metric expression.

[0055] Through this alignment process, the server can eliminate the differences in indicator expression between different business systems or different tables, so that the native indicators have consistent dimensions and indicator structure at the semantic level, laying the foundation for subsequent semantic-based analysis and processing.

[0056] S303: Perform data cleaning on the aligned dimensional indicator information; Specifically, after aligning the native metrics with the structured dimensional metrics, the server performs data cleaning on the aligned dimensional metric information. Data cleaning removes irrelevant characters, redundant descriptions, or abnormal formatting information, and standardizes the dimensional metric information to ensure consistency and accuracy in semantic expression.

[0057] Through data cleaning, the server can further improve the quality of dimensional indicator information and reduce the impact of noisy data on subsequent semantic processing.

[0058] S304: Write the cleaned dimensional indicator information into the dimensional indicator vector library.

[0059] Specifically, after data cleaning, the server vectorizes the cleaned dimensionality metrics and stores the generated semantic feature data in a dimensionality metric vector library. Vectorization converts the dimensionality metrics into a vector representation that can be used for semantic similarity calculation.

[0060] Through the above methods, the server completed the construction of the dimension indicator vector library, enabling the dimension and indicator information in the business system to be stored and managed in a unified semantic feature form, thereby providing a reliable data foundation for the semantic parsing and element decomposition of subsequent query requests.

[0061] In one embodiment, such as Figure 17 As shown, to achieve unified alignment of native metrics in the business system to structured dimensions / metrics, a self-aware dimension / metric processing flow based on a large model is constructed, which specifically includes the following steps: First, the server builds a dimension / metric perceiver and provides the report template information from the business system to the large model. The report template describes the metric fields, dimension fields, and their display structure contained in the report. The large model performs semantic analysis on the template from the perspective of business experts, automatically identifies the dimension and metric information present in the template, and outputs the identification results in a structured form, thereby generating standardized dimension / metric description results.

[0062] Subsequently, the server provides the dimension / metric information obtained from the large model to humans for confirmation, in order to determine whether the dimension / metric identification results conform to the actual business semantics and business requirements. After human confirmation, the server sets the corresponding dimensions / metrics in the table template and records the mapping relationship between the report template and the standardized dimensions / metrics, which is used to describe the correspondence between the table template fields and the standardized dimensions / metrics.

[0063] After setting up the mapping relationship, the server aligns the native indicators in the business system according to the mapping relationship between the report template and the standardized dimensions / indicators, mapping the native indicators to the corresponding structured dimension / indicator expressions. For example, for the native indicator "Beginning Balance of Current Assets", the server aligns it to a structured dimension / indicator, where the dimension is "Assets: Current Assets" and the indicator is "Beginning Balance", thereby achieving a standardized expression of the native indicator at the semantic level.

[0064] After aligning the original metrics, the server performs data cleaning on the aligned dimension / metric data. Data cleaning includes, but is not limited to: removing special characters such as numbers and punctuation marks, and removing supplementary explanatory information within parentheses, such as cleaning "of which: raw materials" to "raw materials" to reduce the impact of noise on semantic understanding.

[0065] Finally, the server writes the cleaned dimension / metric data into a vector library for storage. The vector library is used to store the semantic feature information corresponding to the dimensions / metrics to support semantic similarity-based matching and alignment processing during subsequent data collection.

[0066] The method described in this embodiment can reduce the cost of manual intervention while achieving semantic self-awareness, structured expression, and unified alignment of native indicators of the business system, providing a reliable semantic foundation for subsequent natural language-based reporting.

[0067] Figure 4 This is a schematic diagram of the fourth process of the report query method based on PLM dimension index self-awareness and multi-layer alignment provided in the embodiment of the present invention, as shown below. Figure 4 As shown, in one embodiment of the present invention, S101 includes: S401: The large model is invoked by the preset first prompt word to perform an element decomposition on the question request and obtain the user intent index; Specifically, the server invokes the large model using a preset first prompt word to perform an element breakdown of the user's input query request. The first prompt word guides the large model to parse the query request from a business semantic perspective, identify the indicator elements and dimensional elements contained in the query request, and output the parsing results in a structured manner.

[0068] Through element decomposition, the server can obtain user intent metrics that characterize the core objectives of user queries. These user intent metrics reflect the main business metrics or statistical objects that users are interested in during their query requests, providing a basic semantic basis for the generation of subsequent dimensional metric information.

[0069] S402: Generate the dimensional indicator information corresponding to the question number request based on the user intent indicator, the dimensional indicator vector library, and the large model.

[0070] Specifically, after obtaining the user intent metrics, the server generates the dimensional metric information corresponding to the question number requests based on the user intent metrics, the dimensional metric vector library, and the large model.

[0071] The server uses user intent metrics as semantic input and combines them with the semantic features of dimensions and metrics stored in the dimension metric vector library to assist and constrain the semantic understanding results of the large model. This enables the large model to further analyze the number of requests within the existing dimension metric semantic space, thereby generating dimension metric information that matches the actual metric system of the business system.

[0072] By employing the above methods, the server can fully leverage the semantic understanding capabilities of large models while introducing a dimensional indicator vector library to guide and standardize the semantic parsing process. This avoids generating dimensional indicator expressions that are unrelated to or inconsistent with the business system, thereby improving the accuracy and stability of question request element decomposition.

[0073] Figure 5 This is a schematic diagram of the fifth process of the report query method based on PLM dimension index self-awareness and multi-layer alignment provided in the embodiment of the present invention, as shown below. Figure 5 As shown, in one embodiment of the present invention, S402 includes: S501: Perform vector matching in the dimension indicator vector library based on the user intent indicator; Specifically, the server performs vector matching processing in the dimensional indicator vector library based on the user intent indicator. Vector matching is used to calculate the similarity between the user intent indicator and the semantic features of each dimensional indicator in the dimensional indicator vector library, thereby obtaining candidate information of dimensions or indicators that are semantically similar to the user intent indicator.

[0074] Through vector matching, the server can identify multiple dimensions or metric expressions that are semantically similar to user intent metrics, providing candidate semantic space support for subsequent more refined semantic analysis of user query requests.

[0075] S502: Construct known dimension indicator information based on vector matching results; Specifically, after obtaining the vector matching results, the server constructs known dimension indicator information based on the vector matching results. The known dimension indicator information is used to describe the dimensions, indicators, and their semantic range that are explicit or inferable in the user's query request, thereby providing known constraints for subsequent semantic parsing.

[0076] S503: Based on the known dimension indicator information, the large model is invoked through a preset second prompt word to perform secondary element decomposition on the question number request and obtain the dimension indicator information corresponding to the question number request.

[0077] Specifically, after constructing the known dimension indicator information, the server, based on the known dimension indicator information, calls the large model again through the preset second prompt word to perform a secondary element decomposition on the question number request.

[0078] The second prompt word guides the large model to further analyze the question count request under the constraint of known dimensional indicator information, thereby enabling more granular identification of the dimensional elements, indicator elements and their value relationships hidden in the question count request, and generating the dimensional indicator information corresponding to the question count request.

[0079] Through the above-mentioned secondary element decomposition process, the server can correct and refine the semantic understanding of the user's query request based on the results of the primary element decomposition, avoiding parsing deviations caused by semantic ambiguity or incomplete expression, thereby improving the accuracy and reliability of dimensional indicator information generation.

[0080] In one embodiment, such as Figure 18 As shown, to improve the accuracy of understanding user natural language query requests and avoid inaccurate metric positioning due to semantic ambiguity or incomplete business expression, the server adopts a secondary element decomposition approach to align user intent metrics into structured dimensions / metrics. The specific processing flow is as follows: First, the server receives a query request from the user, such as: "What is the total equipment assets of Henan Province in 2024 (in ten thousand yuan)?" The server invokes a large model using pre-defined prompts to perform an initial element breakdown of the user's question, extracting structured dimension / indicator information from the user's natural language. The result of this initial element breakdown might be, for example: { "Indicator Element": ["Total Assets"], "Filtering conditions": [ ["Year", "=", "2024"], ["Region", "=", "Henan Province"], ["Asset Type", "in", "Equipment"] ], "Statistical dimension": [], "Sort": [], "Number of records returned": 1, "Dimensional hierarchy": [] }; The filtering conditions in the first element decomposition result are used to characterize the dimensional information involved in the user query. The format is [dimensional name, relational operator, dimension value], and it can be used as a preliminary expression of user intent indicators.

[0081] Subsequently, based on the dimension values ​​extracted from the first element decomposition result, the server performs vector matching operations in the dimension indicator vector library to obtain standardized dimension / indicator names that are semantically similar to the dimension value. For example, for the dimension value "equipment", the vector matching result can include multiple semantically similar candidate dimension values ​​such as "equipment", "mechanical equipment", "specialized equipment", "general equipment", and "medical equipment", thereby narrowing down the range of dimension indicators to be matched for the original expression.

[0082] After obtaining the vector matching results, the server constructs known dimension / indicator information based on the vector matching results, and then calls the large model again using preset large model prompts to perform secondary element decomposition on the user question. Under the constraints of the known dimension / indicator information, the secondary element decomposition further analyzes the semantics of the dimensions and indicators in the user question to reduce ambiguity and improve the accuracy of semantic recognition.

[0083] For example, the server constructs the following known information as input constraints for a large model: User question: What was the total equipment asset value in Henan Province in 2024 (in ten thousand yuan)? #### Sample data for known dimension fields; Asset categories: Equipment, machinery, special equipment, general equipment, medical equipment; #### Known Indicators (the content in parentheses is a field description); Total assets.

[0084] Under the constraints of the aforementioned known information, the server obtains more accurate structured parsing results through secondary element decomposition, for example: { Query fields: ["Total Assets|Indicators", "Asset Categories|Dimensions"], "Filter criteria": [ "Dimension | Asset Category = Equipment", "Dimension | Region = Henan Province", "Dimension|Year=2024" ], Query Requirements: ["Statistics | Asset Categories | Total Assets"] }; The server parses the results of the secondary element decomposition and aligns the obtained dimension and indicator semantics with the standardized dimensions / indicators in the system, thereby completing the alignment process of user intent indicators with structured dimensions / indicators.

[0085] Through the method of this embodiment, the server can perform more granular and accurate semantic understanding of user query requests by combining vector matching and secondary element decomposition mechanisms on the basis of primary element decomposition. This effectively improves the accuracy of user intent recognition and the reliability of subsequent indicator matching, providing stable and accurate semantic support for report queries in complex business scenarios.

[0086] S102: Based on the dimension indicator information corresponding to the question number request and the obtained indicator library, perform indicator matching to determine the native indicator corresponding to the question number request; Specifically, after obtaining the dimension metric information corresponding to the question count request, the server performs metric matching processing based on the dimension metric information and the obtained metric library to determine the native metric corresponding to the question count request.

[0087] The indicator library stores native indicators from the business system and their corresponding structured dimensional indicator data, and records the mapping relationship between native indicators and structured dimensional indicators. The server uses dimensional indicator information as a matching condition to search and compare native indicators in the indicator library, thereby filtering out the native indicators corresponding to the question requests that match the dimensional indicator information.

[0088] Through this metric matching process, the server can establish a correspondence between the user's natural language query request and the actual native metrics available in the business system. This avoids misunderstandings caused by differences in metric naming, table format, or system differences, thereby ensuring that the native metrics used for subsequent data queries match the user's true query intent. Figure 1 To.

[0089] Figure 6 This is a schematic diagram of the sixth process of the report query method based on PLM dimension index self-awareness and multi-level alignment provided in the embodiment of the present invention, as shown below. Figure 6 As shown, in one embodiment of the present invention, S102 includes: S601: Based on the dimension indicator information corresponding to the question number request, perform a match in the indicator library to obtain candidate native indicators; Specifically, based on the dimensional metric information corresponding to the query request, the server performs a matching operation in the metric library to obtain candidate native metrics. The metric library stores the native metrics in the business system and their corresponding structured dimensional metric data, and records the mapping relationship between native metrics and structured dimensional metrics.

[0090] During a matching process, the server uses dimensional indicator information as a matching condition and compares it with the structured dimensional indicators stored in the indicator library. This allows it to filter out native indicators that are consistent with or have a correlation with the dimensional indicator information in terms of dimension and indicator structure, thus forming a set of candidate native indicators.

[0091] Through the above matching process, the server can initially narrow down the range of native indicators in the indicator library, providing a foundation for subsequent refined screening.

[0092] S602: Based on preset matching rules and business rules, perform secondary matching on the candidate native metrics to determine the native metrics corresponding to the question number request.

[0093] Specifically, after obtaining the set of candidate native metrics, the server performs a secondary matching process on the set of candidate native metrics based on preset matching rules and business rules to determine the native metrics corresponding to the question number request.

[0094] Among them, the matching rules are used to judge the consistency between candidate native metrics and dimensional metric information, so as to further confirm the degree of matching of candidate native metrics in semantic structure; the business rules are used to filter candidate native metrics at the business level to ensure that the final determined native metrics meet the actual usage requirements of the business system.

[0095] Through secondary matching, the server can determine the native metric that best matches the query request from the set of candidate native metrics, thereby providing accurate and reliable metric basis for subsequent data queries.

[0096] Figure 7 This is a schematic diagram of the seventh process of the report query method based on PLM dimension index self-awareness and multi-layer alignment provided in the embodiment of the present invention, as shown below. Figure 7 As shown, in one embodiment of the present invention, the report query method based on PLM dimension index self-awareness and multi-level alignment further includes: S701: Construct known indicator data based on the native indicators corresponding to the question number requests; Specifically, after the server completes the screening / determination of native metrics, it organizes the set of candidate native metrics into "known metric data" and organizes and caches it in a structured manner. The known metric data includes at least the metric name (or metric identifier) ​​of each candidate native metric.

[0097] S702: Based on the known indicator data and the question request, the original indicators are recommended for relevance through the large model, and a corresponding indicator calculation model is generated.

[0098] Specifically, the server constructs prompt content based on user question requests and known indicator data, and submits the user questions and the list of native indicators to the large model. The large model infers and ranks the relevance between the native indicators and the user's intent, recommending the most relevant target native indicators. At the same time, the server instructs the large model to generate an indicator calculation model corresponding to the recommended native indicators in the output results. The indicator calculation model includes at least: indicator mapping relationship and calculation expression / rule, and can further output the meaning of the expression for result interpretation and display, thereby realizing the question processing capability of "recommending target indicators + automatically generating calculation models".

[0099] In one embodiment, such as Figure 19 As shown, after aligning user intent metrics to structured dimensions / metrics, the server further matches and filters the native metrics of the business system in the metric library based on the aligned structured dimensions / metrics to determine the final target native metrics used for question response. The specific processing procedure is as follows: First, the server uses the aligned structured dimensions / metrics as filtering criteria to query the metric library, retrieving a set of native metrics that meet the dimension / metric criteria. The metric library stores native metrics from the business system and their corresponding structured dimension / metric information, providing the basic data source for native metric matching.

[0100] Subsequently, the server performs a consistency check on the retrieved set of native metrics based on preset native metric matching rules. These rules include: when the aligned structured dimension / metric is completely identical to the structured dimension / metric corresponding to a native metric, the server prioritizes acquiring that native metric; when the aligned structured dimension / metric does not contain the required dimensions or metric names for the native metric, the server fills in the missing dimensions or metric names with default values; if the filled-in values ​​completely match the structured dimension / metric corresponding to the native metric, then that native metric is prioritized; when a native metric lacks necessary dimensional information and cannot completely match the aligned structured dimension / metric, the server determines that the native metric cannot be matched and does not include it in the candidate results.

[0101] After completing the consistency matching based on structured dimensions / metrics, the server further processes the candidate native metrics through the business rule processing engine. First, the server performs time range filtering, selecting a list of native metrics whose effective and expiration times meet the criteria based on the time elements identified in the user query requests. The attribute information of native metrics includes metric name, report task, report scheme, report name, effective time (corresponding report scheme), organizational type, and period type. The server sorts the native metrics in reverse chronological order according to their effective time, prioritizing native metrics closest to the query time to determine the report task, organizational type, and period type to which the metric belongs.

[0102] Based on this, the server further processes organizational elements. According to the organizational elements identified in the query requests and the organizational types corresponding to the native metrics, the server returns the Top-k candidate organizations via vector retrieval. Subsequently, the server queries the organizational permission interface for organizations that the user has access to, and selects the first organization that the user has permission to from the Top-k candidate organizations as the target organization. The remaining organizations can be retained as subsequent switching conditions.

[0103] Next, based on the determined target organization, the task information to which the native metric belongs, and the period type information, the server performs permission filtering on the native metrics. Specifically, the server queries the permission configuration of the native metrics, and takes the intersection of the native metrics that have passed permission verification with the aforementioned aligned list of native metrics to obtain the list of native metrics that the user has access to in the current query scenario.

[0104] After obtaining the list of native metrics that has passed permission verification, the server constructs known metric information based on the list and uses a large model to recommend the most relevant native metrics based on the user's query, determining the target native metric that best matches the user's query intent and generating the corresponding calculation formula. For example, if the user's query is "What is the total equipment asset value in Henan Province in 2024 (in ten thousand yuan)?", the server constructs the following known metric data: #### Known indicator data; Indicator names: Total equipment assets, Total equipment assets of public institutions, Total equipment assets of administrative units.

[0105] Based on the known indicator data mentioned above, the server uses a large model to recommend the most relevant target indicators and their calculation expressions, for example: { "Indicators": { "Name": { "name1": "Total Equipment Assets" }, "expression": { "name2": "Math.round((name1 / 10000),2)" }, "Expression meaning": { "name2": "Total Equipment Assets (Ten Thousand Yuan)" } } } Through the method described in this embodiment, the server can accurately locate native indicators that match the user's intent based on structured dimension / indicator alignment, combined with indicator library matching rules, business rule processing, and large model inference capabilities, and automatically generate the corresponding calculation model, thereby providing high-accuracy and high-reliability technical support for intelligent data querying in complex business scenarios.

[0106] S103: Based on the native metrics corresponding to the query request, call the data query interface to obtain the corresponding data results, and return the query results to the user.

[0107] Specifically, after determining the native metric corresponding to the query request, the server calls the data query interface based on the native metric to obtain the corresponding data results from the business system. The data query interface is used to query or calculate business data according to the report task, data table, or calculation logic corresponding to the native metric to obtain data results that satisfy the query request.

[0108] After retrieving the data results, the server performs necessary encapsulation and processing on the query results, and then returns the processed results to the user client to complete a report query response. In this way, users can query complex business report data using natural language, without needing to understand the underlying indicator structure or report configuration details.

[0109] In one embodiment, after determining the target native metric and generating the corresponding calculation model, the server further queries, calculates, processes, and displays the metric data to achieve the final response to the user's query request. The specific processing flow is as follows: (a) Indicator data query Based on the organizational structure, period, and aligned native metrics identified in the user's query request, the server invokes the data query interface (API) provided by the reporting system to retrieve the corresponding native metric result dataset from the business system. This result dataset contains the basic data corresponding to the target native metric and is used for subsequent metric calculations and result processing.

[0110] (II) Calculation and processing of indicator formulas

[0111] After obtaining the raw indicator result dataset, the server performs transformation and calculation processing on the raw indicator data based on the generated indicator calculation expressions. The specific calculation process includes the following steps: 1) The original metric data cached during the server's data parsing process; 2) The server parses the indicator calculation expression and summarizes the indicator names that need to be replaced in the expression; 3) The server replaces the aggregated indicator names with the corresponding original indicator data values, resulting in the replaced data expression; 4) The server executes the data expression by calling the ScriptEngine.eval(expr).toString() method, obtains the calculation result, and retrieves the business definition information corresponding to the expression from its meaning. The server stores the expression meaning as the key and the calculation result as the value, thereby forming indicator result data that can be displayed.

[0112] (III) Data Encapsulation and Chart Format Conversion

[0113] After the indicator calculation is completed, the server performs unified encapsulation processing on the result data according to the display requirements. The encapsulation content includes at least: organizational information, period information, indicator name, and the meaning of the indicator calculation expression, to ensure that the result data has complete business semantics.

[0114] The server then converts the packaged result data into the data format required for chart display, including the fields shown in Table 1 below: Table 1

[0115] (iv) Interface rendering and result display

[0116] After converting the chart data format, the server transmits the chart data to the front-end interface for rendering and display. Depending on the type of metric data, the server guides the front-end to use different display formats, including but not limited to: For indicator-type data, a tabular format is used, such as "Total Overdue Balance at the End of the Period"; For trend-based data, a line chart format is used to display it, such as "Monthly Trend of Overdue Rate"; For percentage-based data, use pie charts or donut charts to display the data, such as "Distribution of Overdue Amounts"; For ranking data, a bar chart is used to display it, such as "Branch Office Defect Rate Ranking".

[0117] Through the method of this embodiment, the server can automatically query, flexibly calculate and diversely display indicator data after completing indicator matching and calculation model generation, thereby providing users with intuitive, accurate and business-semantic query results, improving the usability and user experience of the report query system in actual business scenarios.

[0118] During the implementation of this solution, an authentication service module was added to the management system and integrated with the business system to achieve unified permission authentication and verification. This ensures that subsequent indicator queries and data processing comply with the access control requirements of the business system. Users configure and annotate reports, clearly identifying the dimensional and indicator information contained in the reports to provide basic data support for the system's subsequent understanding and processing of report semantics. An indicator library was created, and indicator data from the business system was synchronized to the indicator library to achieve centralized management of native indicators and their related information. The system supports automatic or manual triggering of indicator data cleaning and processing, and after cleaning, the processed indicator data is automatically written to the vector library for subsequent semantic-based analysis, matching, and data processing.

[0119] This application provides a report query method based on PLM dimensional indicator self-awareness and multi-layer alignment. Upon receiving a user's query request, the method decomposes the query request into elements based on a large model and a pre-built dimensional indicator vector library to obtain the corresponding dimensional indicator information. Then, based on the dimensional indicator information and the obtained indicator library, indicator matching is performed to determine the native indicator corresponding to the query request. Based on the native indicator, a data query interface is called to obtain the corresponding data results, which are then returned to the user. This achieves automatic standardization from native indicators to structured dimensions or indicators, and accurately maps the user's query intent to the target indicator. Therefore, it significantly improves the accuracy, robustness, and business adaptability of report queries without requiring the reconstruction of the data warehouse or indicator system.

[0120] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0121] Based on the same inventive concept, embodiments of the present invention also provide a report query device based on PLM dimension indicator self-awareness and multi-level alignment, which can be used to implement the report query method based on PLM dimension indicator self-awareness and multi-level alignment described in the above embodiments, as described in the following embodiments. Since the principle of the report query device based on PLM dimension indicator self-awareness and multi-level alignment is similar to that of the report query method based on PLM dimension indicator self-awareness and multi-level alignment, embodiments of the report query device based on PLM dimension indicator self-awareness and multi-level alignment can refer to embodiments of the report query method based on PLM dimension indicator self-awareness and multi-level alignment, and repeated details will not be elaborated further. As used below, the terms "unit" or "module" can refer to a combination of software and / or hardware that performs a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0122] Figure 8 This is a schematic block diagram of the first structure of the report query device based on PLM dimension index self-sensing and multi-layer alignment provided in an embodiment of the present invention, as shown below. Figure 8 As shown, in one embodiment of the present invention, the report query device based on PLM dimension index self-awareness and multi-layer alignment includes: The element decomposition unit 801 is used to decompose the question request based on the large model and the pre-built dimension index vector library when it receives the user's question request, so as to obtain the dimension index information corresponding to the question request. The indicator matching unit 802 is used to perform indicator matching based on the dimension indicator information corresponding to the question number request and the obtained indicator library to determine the native indicator corresponding to the question number request. The data query unit 803 is used to call the data query interface to obtain the corresponding data results based on the native indicators corresponding to the query request, and return the query results to the user.

[0123] Figure 9 This is a schematic block diagram of the second structure of the report query device based on PLM dimension index self-sensing and multi-layer alignment provided in the embodiments of the present invention. Figure 8 Based on the embodiments, further, such as Figure 9 As shown, in one embodiment of the present invention, the report query device based on PLM dimension index self-awareness and multi-layer alignment of the present invention further includes: The structured information generation unit 901 is used to parse the obtained report templates based on the large model and generate structured dimensional indicator information. The vector library generation unit 902 is used to generate a dimension indicator vector library based on the structured dimension indicator information and the native indicators of the report template.

[0124] Figure 10 This is a schematic block diagram of the third structure of the report query device based on PLM dimension index self-sensing and multi-layer alignment provided in the embodiments of the present invention. Figure 9 Based on the embodiments, further, such as Figure 10 As shown, in one embodiment of the present invention, the vector library generation unit 902 includes: The mapping relationship generation module 1001 is used to determine the mapping relationship between the report template and the structured dimension indicator information after the structured dimension indicator information has been manually verified and confirmed to be valid. The indicator alignment module 1002 is used to align the native indicators of the report template to the corresponding structured dimensional indicator information based on the mapping relationship, so as to obtain the aligned dimensional indicator information. Data cleaning module 1003 is used to clean the aligned dimensional indicator information; The vector library generation module 1004 is used to write the cleaned dimensional indicator information into the dimensional indicator vector library.

[0125] Figure 11 This is a schematic block diagram of the fourth structure of the report query device based on PLM dimension index self-sensing and multi-layer alignment provided in the embodiments of the present invention. Figure 8 Based on the embodiments, further, such as Figure 11 As shown, in one embodiment of the present invention, the element decomposition unit 801 includes: User intent index generation module 1101 is used to call the large model through a preset first prompt word, perform an element decomposition on the question request, and obtain user intent index; The dimension indicator information generation module 1102 is used to generate the dimension indicator information corresponding to the question number request based on the user intent indicator, the dimension indicator vector library and the large model.

[0126] Figure 12 This is a fifth structural schematic block diagram of the report query device based on PLM dimension index self-sensing and multi-layer alignment provided in the embodiments of the present invention. Figure 11 Based on the embodiments, further, such as Figure 12 As shown, in one embodiment of the present invention, the dimensional indicator information generation module 1102 includes: The vector matching submodule 1201 is used to perform vector matching in the dimension indicator vector library based on the user intent indicator; The information construction submodule 1202 is used to construct known dimension indicator information based on the vector matching results; The dimension indicator information generation submodule 1203 is used to call the large model based on the known dimension indicator information through a preset second prompt word to perform secondary element decomposition on the question number request and obtain the dimension indicator information corresponding to the question number request.

[0127] Figure 13 This is a sixth structural schematic block diagram of the report query device based on PLM dimension index self-sensing and multi-layer alignment provided in the embodiments of the present invention. Figure 8 Based on the embodiments, further, such as Figure 13 As shown, in one embodiment of the present invention, the index matching unit 802 includes: The first matching module 1301 is used to perform a matching in the indicator library based on the dimension indicator information corresponding to the question number request to obtain candidate native indicators. The second matching module 1302 is used to perform secondary matching on the candidate native indicators based on preset matching rules and business rules to determine the native indicator corresponding to the question number request.

[0128] Figure 14 This is the seventh structural schematic block diagram of the report query device based on PLM dimension index self-sensing and multi-layer alignment provided in the embodiments of the present invention. Figure 8 Based on the embodiments, further, such as Figure 14 As shown, in one embodiment of the present invention, the report query device based on PLM dimension index self-awareness and multi-layer alignment of the present invention further includes: Data construction unit 1401 is used to construct known indicator data based on the native indicators corresponding to the question number request; The model recommendation unit 1402 is used to make relevance recommendations for the original indicators based on the known indicator data and the number of questions, and to generate the corresponding indicator calculation model.

[0129] This application provides a method and apparatus for reporting data collection based on PLM dimensional indicator self-awareness and multi-layer alignment. Upon receiving a user's data collection request, the method decomposes the request into elements based on a large model and a pre-built dimensional indicator vector library to obtain the corresponding dimensional indicator information. Based on the dimensional indicator information and the obtained indicator library, indicator matching is performed to determine the native indicator corresponding to the data collection request. Based on the native indicator, a data query interface is called to obtain the corresponding data results, which are then returned to the user. This achieves automatic standardization from native indicators to structured dimensions or indicators, and accurately maps the user's data collection intent to the target indicator. Therefore, it significantly improves the accuracy, robustness, and business adaptability of reporting data collection without requiring the reconstruction of the data warehouse or indicator system.

[0130] Figure 15 This is a schematic diagram of the structure of the computer device provided in an embodiment of the present invention, such as... Figure 15 As shown, the electronic device may include: a processor 1501, a communication interface 1502, a memory 1503, and a communication bus 1504, wherein the processor 1501, the communication interface 1502, and the memory 1503 communicate with each other through the communication bus 1504. The processor 1501 can call logical instructions in the memory 1503 to execute the following method: upon receiving a user's query request, the processor decomposes the query request into elements based on a large model and a pre-built dimensional indicator vector library to obtain the dimensional indicator information corresponding to the query request; based on the dimensional indicator information corresponding to the query request and the obtained indicator library, the processor performs indicator matching to determine the native indicator corresponding to the query request; based on the native indicator corresponding to the query request, the processor calls a data query interface to obtain the corresponding data results and returns the query results to the user.

[0131] Furthermore, the logical instructions in the aforementioned memory 1503 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, a top-drive control center server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. 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.

[0132] This embodiment discloses a computer program product, which includes a computer program stored on a computer-readable storage medium. The computer program includes program instructions, and when the program instructions are executed by the computer, the computer can execute the methods provided in the above-described method embodiments, such as: upon receiving a user's query request, decomposing the query request into elements based on a large model and a pre-built dimensional indicator vector library to obtain the dimensional indicator information corresponding to the query request; performing indicator matching based on the dimensional indicator information corresponding to the query request and the obtained indicator library to determine the native indicator corresponding to the query request; and calling a data query interface based on the native indicator corresponding to the query request to obtain the corresponding data results and returning the query results to the user.

[0133] This embodiment provides a computer-readable storage medium storing a computer program that causes the computer to execute the methods provided in the above-described method embodiments. For example, upon receiving a user's query request, the computer decomposes the query request into elements based on a large model and a pre-built dimensional indicator vector library to obtain dimensional indicator information corresponding to the query request; it performs indicator matching based on the dimensional indicator information corresponding to the query request and the obtained indicator library to determine the native indicator corresponding to the query request; and based on the native indicator corresponding to the query request, it calls a data query interface to obtain the corresponding data results and returns the query results to the user.

[0134] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0135] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0136] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0137] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0138] In the description of this specification, the references to terms such as "an embodiment," "a specific embodiment," "some embodiments," "for example," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the 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.

[0139] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of this application. It should be understood that the above descriptions are merely specific embodiments of this application and are not intended to limit the scope of protection of this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A report query method based on PLM dimensional indicator self-awareness and multi-level alignment, characterized in that, include: Upon receiving a user's data query request, the data query request is decomposed into elements based on the large model and a pre-built dimensional indicator vector library to obtain the dimensional indicator information corresponding to the data query request. Based on the dimension indicator information corresponding to the question number request and the obtained indicator library, indicator matching is performed to determine the native indicator corresponding to the question number request; Based on the native metrics corresponding to the query request, the data query interface is called to obtain the corresponding data results, and the query results are returned to the user. The step of matching metrics based on the dimension metric information corresponding to the question count request and the obtained metric library to determine the native metric corresponding to the question count request includes: Based on the dimension indicator information corresponding to the question number request, a match is performed in the indicator library to obtain candidate native indicators; Based on preset matching rules and business rules, the candidate native metrics are matched a second time to determine the native metrics corresponding to the question number request.

2. The report query method based on PLM dimensional indicator self-awareness and multi-layer alignment according to claim 1, characterized in that, The steps for pre-constructing the dimensional indicator vector library include: The obtained report templates are parsed based on the large model to generate structured dimensional indicator information; A dimensional indicator vector library is generated based on the structured dimensional indicator information and the native indicators of the report template.

3. The report query method based on PLM dimensional indicator self-awareness and multi-layer alignment according to claim 2, characterized in that, The step of generating a dimension indicator vector library based on the structured dimension indicator information and the native indicators of the report template includes: After the structured dimensional indicator information is manually verified and confirmed to be acceptable, the mapping relationship between the report template and the structured dimensional indicator information is determined. Based on the mapping relationship, the original indicators of the report template are aligned to the corresponding structured dimensional indicator information to obtain the aligned dimensional indicator information. The aligned dimensional index information is then cleaned. Write the cleaned dimensional indicator information into the dimensional indicator vector library.

4. The report query method based on PLM dimensional indicator self-awareness and multi-layer alignment according to claim 1, characterized in that, The method involves decomposing the question count request based on a large model and a pre-built dimensional indicator vector library to obtain the dimensional indicator information corresponding to the question count request, including: The large model is invoked by a preset first prompt word to perform an element decomposition on the question request and obtain the user intent index. Based on the user intent metric, the dimensional metric vector library, and the large model, the dimensional metric information corresponding to the question number request is generated.

5. The report query method based on PLM dimensional indicator self-awareness and multi-layer alignment according to claim 4, characterized in that, The step of generating the dimensional indicator information corresponding to the question count request based on the user intent indicator, the dimensional indicator vector library, and the large model includes: Based on the user intent metrics, vector matching is performed in the dimension metric vector library; Construct known dimension indicator information based on vector matching results; Based on the known dimensional indicator information, the large model is invoked through a preset second prompt word to perform secondary element decomposition on the question number request, thereby obtaining the dimensional indicator information corresponding to the question number request.

6. The report query method based on PLM dimensional indicator self-awareness and multi-layer alignment according to claim 1, characterized in that, Also includes: Construct known indicator data based on the native indicators corresponding to the question requests; Based on the known indicator data and the number of questions, the large model performs relevance recommendations on the original indicators and generates corresponding indicator calculation models.

7. A report query device based on PLM dimensional indicator self-sensing and multi-layer alignment, characterized in that, include: The element decomposition unit is used to decompose the question request based on the large model and the pre-built dimension indicator vector library when it receives the user's question request, so as to obtain the dimension indicator information corresponding to the question request. The indicator matching unit is used to perform indicator matching based on the dimension indicator information corresponding to the question number request and the obtained indicator library to determine the native indicator corresponding to the question number request. The data query unit is used to call the data query interface to obtain the corresponding data results based on the native indicators corresponding to the query request, and return the query results to the user.

8. A computer 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, it implements the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program / instructions stored thereon, characterized in that, When the computer program / instructions are executed by the processor, they implement the steps of the method according to any one of claims 1 to 6.

10. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instructions are executed by the processor, they implement the steps of the method according to any one of claims 1 to 6.

Images