Quota term content standardization conversion method based on semantic link and retrieval enhancement

Abstract

The invention discloses a quota term content standardization conversion method based on semantic link and retrieval enhancement, and relates to the technical field of standardization conversion, the method comprises the following steps: determining a structured field of a quota term based on multi-source data of power grid quota business; performing semantic understanding on a to-be-processed quota clause text through the trained large language model, mapping a corresponding field in the quota clause text into a structured field, and outputting the structured field; performing grammar and business verification on the structured field output by the large language model, and obtaining an object candidate field from the verified structured field; performing standard entity alignment in a pre-constructed power grid entity dictionary and a pre-constructed knowledge graph through a semantic link module according to the object candidate fields; inputting the structured field aligned by the standard entity into a retrieval enhancement generation module used for realizing structured quota record joint generation; the whole process is well documented and is convenient to audit and maintain.

Description

Technical Field

[0001] This invention relates to the field of standardization conversion technology, and more specifically, to a method for standardizing quota terms based on semantic linking and enhanced retrieval. Background Technology

[0002] As the scale and operation mode of the power grid become increasingly complex, constraints such as cross-regional section limits and line transmission capacity are scattered in unstructured texts such as control procedures, operation rules and meeting minutes. The expressions are inconsistent and the referential relationships are vague. They still mainly rely on manual understanding and maintenance, which makes it difficult to support large-scale automatic verification and systematic management.

[0003] Existing technologies have begun to utilize knowledge graphs to structurally model power grid dispatching procedures and fault handling knowledge for intelligent question answering and decision support. However, these technologies primarily focus on extracting and querying entity relationships at the "procedure clause - equipment / event" level. Meanwhile, Retrieval Enhanced Generation (RAG) technology has been used for question answering and text standardization in general knowledge base scenarios. However, there is still a lack of dedicated solutions for power grid quota-based business that can simultaneously process semantic references of equipment, lines, and systems and output standardized quota-based structured results that can be directly used for dispatching calculations. Summary of the Invention

[0004] The purpose of this invention is to provide a method for standardizing and transforming quota terms based on semantic linking and enhanced retrieval, so as to solve the problems existing in the above-mentioned background technology.

[0005] The above-mentioned technical objective of the present invention is achieved through the following technical solution:

[0006] Firstly, this application provides a method for standardizing and transforming quota terms based on semantic linking and retrieval enhancement, including the following specific steps:

[0007] Based on multi-source data from power grid quota business, determine the structured fields of quota terms;

[0008] The trained large language model performs semantic understanding on the quota clause text to be processed, maps the corresponding fields in the quota clause text to structured fields and outputs them;

[0009] Perform grammatical and business validation on the structured fields output by the large language model, and obtain candidate fields for objects from the validated structured fields;

[0010] The semantic linking module aligns standard entities in a pre-built power grid entity dictionary and knowledge graph based on candidate object fields.

[0011] The structured fields, aligned with standard entities, are input into the retrieval enhancement generation module used to jointly generate structured quota records.

[0012] Based on the above technical solution, the present invention can be further improved as follows.

[0013] Furthermore, the aforementioned structured fields include object candidate fields, quota type fields, quota value fields, quota unit fields, operating condition fields, and applicable time periods.

[0014] Furthermore, the alignment of the aforementioned standard entities is specifically as follows:

[0015] The entity linking scoring function is used to calculate the comprehensive score of the entity in each candidate field of the power grid entity dictionary and knowledge graph, and the entity with the best comprehensive score is determined as the target entity.

[0016] Furthermore, the aforementioned entity link scoring function is as follows: ; In the formula, For comprehensive scoring, This is a string similarity score normalized based on edit distance or n-gram similarity. This is an entity type consistency indicator function, which takes the value 1 when the entity type hint matches the candidate entity type, and 0 otherwise. The compatibility score is calculated based on the topological relationships of the knowledge graph. These are pre-set weights.

[0017] Furthermore, the aforementioned entity link scoring function is as follows: ; In the formula, For comprehensive scoring, To reflect the relative position and distance between the pronoun and the candidate antecedent entity in the clause structure, This indicates a function that indicates consistency between the reference type and the entity type. To match scores, , , These represent the pre-set weights.

[0018] Secondly, this application provides a standardization and transformation system for quota terms based on semantic linking and retrieval enhancement, applicable to the standardization and transformation method for quota terms based on semantic linking and retrieval enhancement in any of the first aspects, including:

[0019] The field determination module is used to determine the structured fields of the quota terms based on multi-source data from power grid quota business.

[0020] The field mapping module is used to perform semantic understanding on the quota clause text to be processed using a trained large language model, and to map the corresponding fields in the quota clause text to structured fields and output them.

[0021] The field acquisition module is used to perform grammatical and business validation on the structured fields output by the large language model, and to obtain candidate fields for objects from the validated structured fields.

[0022] The entity alignment module is used to align standard entities in a pre-built power grid entity dictionary and knowledge graph based on object candidate fields through the semantic linking module.

[0023] The retrieval enhancement module is used to input the structured fields, which have been aligned with standard entities, into the retrieval enhancement generation module, which is used to jointly generate structured quota records.

[0024] Furthermore, the aforementioned structured fields include object candidate fields, quota type fields, quota value fields, quota unit fields, operating condition fields, and applicable time periods.

[0025] Furthermore, the alignment of the aforementioned standard entities is specifically as follows:

[0026] The entity linking scoring function is used to calculate the comprehensive score of the entity in each candidate field of the power grid entity dictionary and knowledge graph, and the entity with the best comprehensive score is determined as the target entity.

[0027] Thirdly, this application provides an electronic device, including: at least one processor, at least one memory, and a data bus;

[0028] In this system, the processor and memory communicate with each other via a data bus; the memory stores program instructions that can be executed by the processor, and the processor calls the program instructions to execute the quota terms content standardization conversion method based on semantic linking and retrieval enhancement, as described in any of the first aspects.

[0029] Fourthly, this application provides a non-transitory computer-readable storage medium that stores computer instructions that cause a computer to execute the quota terms content standardization conversion method based on semantic linking and retrieval enhancement as described in any of the first aspects.

[0030] Compared with the prior art, the present invention has at least the following beneficial effects:

[0031] In this application, by organizing each limit clause into a limit event, the six types of information, namely "limit object, limit type, value, unit, working condition, and applicable period", are extracted at once to form a fixed structure that can be directly stored in the database and used in calculations, thus solving the problem of scattered information and difficulty in using it for automatic limit verification in existing practices.

[0032] In this application, by designing Chinese prompts and output format constraints, a pre-trained large language model is invoked to extract the above six types of elements from the original clauses and generate standardized expressions. At the same time, the prompts explicitly prohibit modification of the quota object, value, unit and working condition, and only allow uniform wording. Combined with simple rule verification, the automatic conversion from "original clauses" to "standard clauses + structured records" is achieved without any fine-tuning.

[0033] In this application, objects such as "this section," "this line," and "this system" extracted from LLM are aligned with entities in the power grid knowledge graph through string similarity, type constraints, and topological relationships. A clause-level semantic subgraph is then constructed. This structured information, along with quota events, is provided to the large language model for retrieval and generation. This ensures that the generated standard clauses and rules accurately fall on the correct equipment / section, and the entire process is verifiable, auditable, and easy to maintain. Attached Figure Description

[0034] The accompanying drawings, which are included to provide a further understanding of embodiments of the invention and form part of this application, do not constitute a limitation thereof. In the drawings:

[0035] Figure 1 This is a flowchart of the conversion method in an embodiment of the present invention;

[0036] Figure 2 This is a schematic diagram of the connection of the conversion system in an embodiment of the present invention;

[0037] Figure 3 This is a schematic diagram of the connection of an electronic device in an embodiment of the present invention. Detailed Implementation

[0038] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0039] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0040] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0041] In the description of the embodiments of the present invention, "multiple" means at least two.

[0042] Example 1: This example provides a method for standardizing and transforming quota terms based on semantic linking and enhanced retrieval, such as... Figure 1 As shown, the specific steps include the following:

[0043] S1, based on multi-source data from power grid quota business, determines the structured fields of quota terms.

[0044] The first step involves preparing multi-source data and knowledge for power grid quota business, providing a unified and complete semantic foundation for subsequent semantic linking and enhanced retrieval. The determination of structured fields can be achieved through the following steps:

[0045] 1) Collect textual materials from dispatching procedures, operation procedures, safety constraint calculation instructions, power grid operation mode details, over-limit handling procedures, dispatching meeting minutes and dispatching instruction records, etc. Screen and clean the clauses involving cross-section limits, line transmission capacity and system operation mode constraints, and store them in a unified structured or semi-structured text format (such as JSON). Retain metadata such as clause number, version information and scope of application to form a limit clause corpus.

[0046] 2) Based on the existing power grid asset ledger and dispatch master station database, core business objects such as substations, generator units, lines, tie lines, sections, busbars, regional systems, and operating modes are extracted to construct a domain entity dictionary and terminology system, clarifying the standard names, aliases, and hierarchical relationships of various entities (e.g., "line-section-regional system"). On this basis, combined with deep learning entity recognition and relationship extraction methods, power grid objects and their relationships in texts such as dispatch procedures, fault reports, and operating mode change orders are extracted into "entity-relationship-attribute" triples to construct a power grid knowledge graph, realizing a graph representation of knowledge such as equipment topology relationships, membership relationships, and operating constraints. This knowledge graph serves as a unified entity space for subsequent semantic linking, supporting the disambiguation and alignment of "equipment-line-system" references in clauses. Building upon the completed power grid entity knowledge modeling, typical quota clauses and rules for handling exceeding limits are further compiled from normative documents and manually reviewed historical cases to establish a quota business knowledge base and standardized template library.

[0047] 3) Structure and store the elements such as “object identifier, limit type, limit value and unit, applicable working conditions, applicable time period, and exception conditions” from the historical terms into the database to form a searchable set of limit rules.

[0048] 4) Summarize the typical expression styles of the limit clauses under different voltage levels, different cross-sectional types, and different operating modes, and form several standardized text templates to provide structural and sentence constraints for the subsequent retrieval enhancement generation module.

[0049] S2 uses a trained large language model to perform semantic understanding of the quota clause text to be processed, maps the corresponding fields in the quota clause text to structured fields and outputs them.

[0050] Semantic parsing is based on a pre-trained Large Language Model (LLM), meaning it no longer relies on large-scale domain-annotated corpora and additional supervised training. Instead, it directly calls the LLM to understand and structurally extract quota clauses through designed prompt templates and output constraints. Specifically, the input is constructed by combining candidate quota clauses with uniformly designed task instructions, including: a description of the task objective (e.g., "Determine whether the following sentences are grid quota clauses; if so, extract the quota object, quota type, quota value and unit, applicable operating conditions and applicable time period, and output them in the specified JSON format"), a brief definition of the meaning of each field (e.g., "The object is the standard name or text description of a section, line or system, and the quota type is active power, current or transmission capacity, etc."), and several typical example clauses and their expected output forms, used as small-sample demonstrations.

[0051] Given the above input, the pre-trained large language model performs overall semantic understanding of the clause text, identifies whether the clause belongs to the quota-related content, and maps the "object name or designation (e.g., ××—×× section, this system, etc.)", "quota type (active power, current, etc.)", "value and unit (e.g., 800MW, 2000A)", "operating mode or condition (e.g., summer peak mode, N-1 maintenance mode)" and "time period conditions (e.g., June–September each year, long-term validity)" into pre-agreed structured fields, and returns them in the agreed JSON or key-value pair format.

[0052] S3 performs grammatical and business validation on the structured fields output by the large language model, and obtains candidate fields for objects from the validated structured fields.

[0053] When performing syntax and business validation on the structured results: on the one hand, the JSON structure is checked for completeness, missing fields or incorrect types. If problems are found, the model is re-requested for correction with additional prompts. On the other hand, the validity of the extracted results is checked in conjunction with business rules such as the consistency of the physical quantity and unit, and the consistency of the voltage level and object attributes. Unreasonable combinations are marked as requiring manual review or triggering secondary parsing.

[0054] Based on this, the "quota event" structure, parsed and verified by LLM, is used as input for subsequent semantic linking and retrieval enhancement generation modules. This achieves integrated semantic parsing and structured expression of "object—quota type—value—unit—working condition—time period" in quota clauses without requiring additional supervised training for this invention.

[0055] ;in Candidates for quota-limited objects (sections / lines / systems), For quota types (active power, current, etc.), and These are the limit amount and the unit, respectively. For operating conditions, This refers to the applicable time period.

[0056] S4 aligns standard entities in a pre-built power grid entity dictionary and knowledge graph based on object candidate fields through the semantic linking module.

[0057] Among them, for each limit event, the candidate objects The semantic linking module retrieves a set of candidate entities with the same name, whose names contain similarities, or whose aliases match from a pre-built power grid entity dictionary and knowledge graph. and other sets of entities already aligned in the same clause. Define the entity link scoring function under the context constraints: ; in, The string similarity score is normalized based on edit distance or n-gram similarity. For entity type consistency indicator functions, when The value is 1 if the text contains type hints such as "section", "line", "substation", or "system" and matches the candidate entity type; otherwise, it is 0. This represents a compatibility score calculated based on the knowledge graph's topological relationships. It reflects whether a candidate entity maintains topological and hierarchical consistency with entities such as substations and regional systems already aligned within the same clause. (Coefficient) The weights are pre-defined. Based on the scoring function described above, the system selects the optimal entity from the candidate set. ; when ( When the threshold is set to 0, the object is marked as "awaiting manual review" to prevent erroneous linking results from propagating backward. By explicitly introducing entity type constraints and power grid topology constraints on top of string similarity, this invention effectively reduces the risk of confusion between highly similar entities such as lines with the same name and adjacent sections without the need to train a dedicated entity linking model.

[0058] Secondly, to address the numerous references to "this section," "this line," "this system," "other related sections," and "the aforementioned limit" in the quota clauses, the semantic linking module employs a rule-based reference resolution mechanism; within the same chapter, section, or clause group, the system resolves each reference term... Construct a candidate set of preceding entities And define a function to represent the resolution scoring function: ; in, This reflects the relative position and distance between the pronoun and the candidate antecedent entity in the clause structure; the closer the distance, the higher the score. To ensure consistency between the reference type and the entity type, for example, "this section" can only refer back to the section entity, and "this system" can only refer back to the regional system or power grid system entity; Based on the context entities in the knowledge graph related to the current limit event, such as the regional system and voltage level, candidates with consistent topological relationships and compatible operating conditions are assigned higher scores; calculation: ; when ( When the threshold is set, the referential relationship is also marked as awaiting manual confirmation. For set references such as "other related sections of this system", this invention queries all section entities that have an "include / belong to" relationship with the entity "this system" in the knowledge graph, and represents the set as a group of section entities or set identifiers, thereby fully preserving the set semantics. Through the combination of the above-mentioned anti-interference constraints, type constraints and topological constraints, this invention achieves the systematic recovery of the quota object reference chain in the specification text that is "defined once in the preceding text and referenced in multiple places in the following text" without the need to train a general coreference model.

[0059] After completing entity linking and dereference resolution, the semantic linking module denotes the set of standard entities involved in the clause as follows: Let the set of relationships between them be denoted as And construct a clause-level semantic subgraph: ; Among them, the node set It should include at least entities such as cross-sections, lines, substations, regional systems, operating modes, and quota rules, and edge sets. It includes at least relationships such as "belongs to", "contains", "operates under a certain operating mode", and "has a certain limit". This semantic subgraph preserves the topological structure of power grid equipment and regions, and explicitly describes the business constraint relationships of "object-limit-operating condition-time period", serving as the graph structure input and constraint basis for the subsequent retrieval enhancement generation module. Through the above semantic link design, this invention achieves a precise mapping from the limit event structure output by LLM to the standard entities and their referential chains of the power grid knowledge graph, while being based as unsupervised as possible or without additional training. This significantly improves the accuracy, stability, and interpretability of the standardization conversion of limit content.

[0060] S5 inputs the structured fields, after standard entity alignment, into the retrieval enhancement generation module used to achieve the joint generation of structured quota records.

[0061] In this invention, after the semantic parsing and semantic linking module completes the construction of the quota event structure and the entity alignment with the power grid knowledge graph, the retrieval enhancement generation and quota standardization conversion module uses the aligned quota event set... and its clause-level semantic subgraph Using historical normative clauses and quota rule bases as input, this module enables the retrieval of similar clauses, the selection of similar clauses, and the joint generation of standardized clause texts and structured quota records. Without requiring additional supervised training of the language and retrieval models, this module achieves a unified expression and computable transformation of quota content through a pre-built index structure, template library, and the invocation of a pre-trained large language model.

[0062] First, this invention will limit the event. The system inputs the standard entity information linked to it as the "query intent" into the large language model. Specifically, the system constructs prompt information containing the following elements: the standard name of the quota object and its attributes such as voltage level, cross-section type, and regional system; the quota type, value, and unit; the current operating conditions and applicable time period; and explicitly instructs the large language model to "search for several historical clauses most similar to this object, this quota type, and this operating condition from existing regulations and quota rule bases, and generate standardized quota expressions and structured rule records based on the retrieved content." In implementation, the system uses the retrieval interface integrated into the large language model or embeds retrieval components to perform the matching and selection of similar clauses and rules internally, forming a "knowledge context" highly relevant to the current scenario, without requiring the caller to separately train or implement a traditional retrieval model.

[0063] Subsequently, in the same or a subsequent call, the system will display the current limit event. Clause-level semantic subgraph Together with the relevant historical clause summaries returned by the large language model, we issue explicit generation instructions to the large language model, requiring it to: firstly, output a standardized quota clause text in natural language form, expressing "under what working conditions, for what object, what physical quantity is subject to the quota, and the applicable period" in a rigorous, complete, and stylistically consistent manner; secondly, simultaneously output the corresponding structured quota rule record, for example, in JSON or key-value pair format. ; The system explicitly restricts the large language model from arbitrarily modifying key elements such as quota objects, values, units, and working conditions. It can only unify wording and sentence structure without changing the business meaning, thereby ensuring the semantic seriousness and traceability of the generated results.

[0064] Finally, this invention performs formal and business logic validation on the standardized clause text and structured rule records returned by the large language model: on the one hand, it checks whether the output format meets the preset field requirements and whether there are any missing fields or type errors; on the other hand, it combines the object attributes, voltage levels, and regional information recorded in the power grid knowledge graph to perform rule-based validation on the pairing relationship between physical quantities such as "active power – MW, current – ​​A" and units, and the consistency between the quota objects and operating conditions. Results that are obviously unreasonable or conflict with the knowledge graph are marked as requiring manual review. Through the above design, this invention can realize the automatic understanding of original quota clauses, the retrieval and utilization of historical clauses and rules, and the synchronous generation and storage of standardized clause texts with unified standards and structured quota rules.

[0065] Example 2: This application provides a standardization and transformation system for quota terms based on semantic linking and retrieval enhancement, applied to the standardization and transformation method for quota terms based on semantic linking and retrieval enhancement in Example 1, such as... Figure 2 As shown, it includes:

[0066] The field determination module is used to determine the structured fields of the quota terms based on multi-source data of the power grid quota business. The structured fields include object candidate fields, quota type fields, quota value fields, quota unit fields, operating condition fields, and applicable time periods.

[0067] The field mapping module is used to perform semantic understanding on the quota clause text to be processed using a trained large language model, and to map the corresponding fields in the quota clause text to structured fields and output them.

[0068] The field acquisition module is used to perform grammatical and business validation on the structured fields output by the large language model, and to obtain candidate fields for objects from the validated structured fields.

[0069] The entity alignment module is used to align standard entities in a pre-built power grid entity dictionary and knowledge graph based on the object candidate fields through the semantic linking module. The alignment of standard entities is specifically as follows: the entity linking scoring function is used to calculate the comprehensive score of each entity in the power grid entity dictionary and knowledge graph with the entity in the object candidate fields, and the entity with the best comprehensive score is determined as the target entity.

[0070] The retrieval enhancement module is used to input the structured fields, which have been aligned with standard entities, into the retrieval enhancement generation module, which is used to jointly generate structured quota records.

[0071] Example 3: This application provides an electronic device, such as... Figure 3 As shown, it includes: at least one processor, at least one memory, and a data bus;

[0072] In this system, the processor and memory communicate with each other via a data bus; the memory stores program instructions that can be executed by the processor, and the processor calls the program instructions to execute the quota terms content standardization conversion method based on semantic linking and retrieval enhancement as described in Example 1.

[0073] Example 4: This application provides a non-transitory computer-readable storage medium that stores computer instructions, which cause a computer to execute the quota terms content standardization conversion method based on semantic linking and retrieval enhancement described in Example 1.

[0074] 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.

[0075] 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.

[0076] 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.

[0077] 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.

[0078] Those skilled in the art will understand that all or part of the steps in the above facts and methods can be implemented by a program instructing related hardware. The program or the program described therein can be stored in a computer-readable storage medium. When the program is executed, it includes the following steps: at this time, the corresponding method steps are introduced. The storage medium can be ROM / RAM, magnetic disk, optical disk, etc.

[0079] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for standardizing and transforming quota terms based on semantic linking and enhanced retrieval, characterized in that: The specific steps include the following: Based on multi-source data from power grid quota business, determine the structured fields of quota terms; The trained large language model performs semantic understanding on the quota clause text to be processed, maps the corresponding fields in the quota clause text to structured fields and outputs them; Perform grammatical and business validation on the structured fields output by the large language model, and obtain candidate fields for objects from the validated structured fields; The semantic linking module aligns standard entities in a pre-built power grid entity dictionary and knowledge graph based on the candidate fields of the objects. The structured fields, aligned with standard entities, are input into the retrieval enhancement generation module used to jointly generate structured quota records.

2. The method for standardizing and transforming quota terms based on semantic linking and enhanced retrieval according to claim 1, characterized in that, The structured fields include object candidate fields, limit type fields, limit value fields, limit unit fields, operating condition fields, and applicable time periods.

3. The method for standardizing and transforming quota terms based on semantic linking and enhanced retrieval according to claim 1, characterized in that, The alignment of the standard entities is specifically as follows: The entity linking scoring function is used to calculate the comprehensive score of the entity in each candidate field of the power grid entity dictionary and knowledge graph, and the entity with the best comprehensive score is determined as the target entity.

4. The method for standardizing and transforming quota terms based on semantic linking and retrieval enhancement according to claim 3, characterized in that, The entity link scoring function is as follows: ； In the formula, For comprehensive scoring, This is a string similarity score normalized based on edit distance or n-gram similarity. This is an entity type consistency indicator function, which takes the value 1 when the entity type hint matches the candidate entity type, and 0 otherwise. The compatibility score is calculated based on the topological relationships of the knowledge graph. These are pre-set weights.

5. The method for standardizing and transforming quota terms based on semantic linking and enhanced retrieval according to claim 3, characterized in that, The entity link scoring function is as follows: ； In the formula, For comprehensive scoring, To reflect the relative position and distance between the pronoun and the candidate antecedent entity in the clause structure, This indicates a function that indicates consistency between the reference type and the entity type. To match scores, , , These represent the pre-set weights.

6. A standardized conversion system for quota terms based on semantic linking and enhanced retrieval, characterized in that: include: The field determination module is used to determine the structured fields of the quota terms based on multi-source data from power grid quota business. The field mapping module is used to perform semantic understanding on the quota clause text to be processed using a trained large language model, and to map the corresponding fields in the quota clause text to structured fields and output them. The field acquisition module is used to perform grammatical and business validation on the structured fields output by the large language model, and to obtain candidate fields for objects from the validated structured fields. The entity alignment module is used to align standard entities in a pre-built power grid entity dictionary and knowledge graph based on the object candidate fields through the semantic linking module. The retrieval enhancement module is used to input the structured fields, which have been aligned with standard entities, into the retrieval enhancement generation module, which is used to jointly generate structured quota records.

7. The quota terms content standardization and conversion system based on semantic linking and retrieval enhancement according to claim 6, characterized in that, The structured fields include object candidate fields, limit type fields, limit value fields, limit unit fields, operating condition fields, and applicable time periods.

8. The quota terms content standardization and conversion system based on semantic linking and retrieval enhancement according to claim 6, characterized in that, The alignment of the standard entities is specifically as follows: The entity linking scoring function is used to calculate the comprehensive score of the entity in each candidate field of the power grid entity dictionary and knowledge graph, and the entity with the best comprehensive score is determined as the target entity.

9. An electronic device, characterized in that, include: At least one processor, at least one memory, and a data bus; The processor and the memory communicate with each other via the data bus. The memory stores program instructions that can be executed by the processor, which invokes the program instructions to execute the quota terms content standardization conversion method based on semantic linking and retrieval enhancement as described in any one of claims 1-5.

10. A non-transitory computer-readable storage medium, characterized in that, The non-transitory computer-readable storage medium stores computer instructions that cause the computer to execute the quota terms content standardization conversion method based on semantic linking and retrieval enhancement as described in any one of claims 1-5.

Images