Hyper-relation extraction method based on collaboration of multiple large language models and its application

Abstract

This application relates to the field of data processing technology, and in particular to a hyper-relation extraction method based on the collaboration of multiple large language models and its application. By leveraging the powerful learning capabilities of the large language model itself to optimize the extraction prompt words used as extraction rules and providing it with a correction mechanism to correct the generated hyper-relationships, fast and accurate text hyper-relationship extraction can be achieved without fine-tuning the model parameters through supervised learning. This approach aims to address the problem of how to improve the efficiency of hyper-relationship extraction from text.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a hyper-relation extraction method based on the collaboration of multiple large language models. Background Art

[0002] Traditional knowledge graphs are limited to simple triple structures and are mainly used to represent binary relationships. Hyper-relational knowledge graphs can represent more complex n-ary relationships and additional attributes. By associating each fact with relevant metadata (such as source document information, temporal context, and confidence scores), they can more effectively capture and express rich contextual information, thereby improving the performance of knowledge graphs in entity disambiguation and complex query processing.

[0003] One of the commonly used super-relation extraction methods at present is the fine-tuning method, which is to add a task-specific layer on top of the pre-trained model and fine-tune the model parameters through supervised learning. However, the fine-tuning method requires a lot of time and labor costs to be invested in the early stage for supervision, and has the limitation of low extraction efficiency.

[0004] In view of this, this application proposes a super-relation extraction method that does not rely on supervised learning to improve the efficiency of super-relation extraction of text. Summary of the Invention

[0005] The main purpose of this application is to provide a hyper-relation extraction method based on the collaboration of multiple large language models, aiming to solve the problem of how to improve the efficiency of hyper-relation extraction of text.

[0006] To achieve the above objectives, the present application provides a hyper-relation extraction method based on the collaboration of multiple large language models, the method comprising:

[0007] Pre-set extraction prompt words, optimization prompt words and correction prompt words;

[0008] Acquire a text to be extracted, and input the text to be extracted and the optimized prompt word into the prompt word optimization model, so that the prompt word optimization model analyzes the content of the text to be extracted based on the optimized prompt word to generate prompt word optimized content;

[0009] Obtaining the optimized content of the prompt word, and updating the extracted prompt word according to the optimized content of the prompt word to obtain the optimized extracted prompt word;

[0010] Inputting the optimized extraction prompt words and the text to be extracted into the super-relationship extraction model, so that the super-relationship extraction model extracts the super-relationship in the text to be extracted according to the optimized extraction prompt words;

[0011] An initial super relation is obtained, and the initial super relation and the correction prompt word are input into the correction model, so that the correction model proofreads the initial super relation based on the correction prompt word, obtains the correction super relation, and outputs it.

[0012] Optionally, the extracted prompt word satisfies the following constraints:

[0013] The task type is information extraction task;

[0014] The output format includes a super-relationship quadruple of (head entity, relation, tail entity, auxiliary key-value pair);

[0015] The output includes a hint about the auxiliary model narrowing down the scope of the auxiliary key-value pairs.

[0016] Optionally, the optimization prompt word satisfies the following constraints:

[0017] The task type is optimization task;

[0018] The output content includes potential auxiliary key-value pair information in the text to be extracted;

[0019] Contains guiding instructions to guide the large language model to first extract entities in the sentence, then classify the entities, and finally output information other than people and places as potential auxiliary key-value pairs.

[0020] Optionally, the modified prompt word satisfies the following constraints:

[0021] The task type is a correction task;

[0022] The output format includes a super-relationship quadruple of (head entity, relation, tail entity, auxiliary key-value pair);

[0023] The format correction scheme satisfies the following requirements: 4-tuples with missing information are deleted, and 4-tuples with excessive information are pruned while the main triples (head entity, relationship, tail entity) remain unchanged, thus removing redundant information.

[0024] The semantic correction scheme satisfies: comparing the sentence composed of the super-relationship quadruple itself with the original corpus, and checking whether the super-relationship quadruple correctly translates the meaning of the original corpus.

[0025] In addition, to achieve the above-mentioned purpose, the present application also provides a super-relation extraction device, which includes:

[0026] A prompt word optimization model is pre-set with optimized prompt words, and is used to analyze the content of the text to be extracted input by the user based on the optimized prompt words to generate prompt word optimized content;

[0027] An extraction prompt word optimization module, configured to update the extraction prompt word according to the prompt word optimization content to obtain an optimized extraction prompt word;

[0028] A super-relation extraction model is pre-set with extraction prompt words, and is used to extract the super-relation in the text to be extracted according to the optimized extraction prompt words to obtain an initial super-relation;

[0029] The correction model is pre-set with correction prompt words, and is used to proofread the initial super-relationship based on the correction prompt words to obtain a corrected super-relationship.

[0030] In addition, to achieve the above-mentioned purpose, the present application also provides a computer system, which includes: a memory, a processor, and a computer program stored on the memory and runnable on the processor. When the computer program is executed by the processor, the steps of the hyper-relation extraction method based on the collaboration of multiple large language models as described in any one of the above items are implemented.

[0031] In addition, to achieve the above-mentioned purpose, the present application also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the hyper-relation extraction method based on the collaboration of multiple large language models as described in any of the above items are implemented.

[0032] This application has at least the following beneficial effects:

[0033] By leveraging the powerful learning ability of the large language model to optimize the extraction prompt words used as extraction rules and giving it a correction mechanism to correct the generated hyper-relationships, fast and accurate text hyper-relationship extraction can be achieved without fine-tuning the model parameters through supervised learning. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 This is a flow chart of a first embodiment of a method for extracting super relations based on collaboration of multiple large language models according to an embodiment of the present application;

[0035] Figure 2 Schematic diagram of the architecture of the hyper-relation extraction device involved in the embodiment of the present application;

[0036] Figure 3 This is a schematic diagram of the architecture of the hardware operating environment of the computer system involved in the embodiments of the present application.

[0037] The realization of the objectives, functional features and advantages of this application will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0038] To better understand the above technical solutions, exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments described herein. Rather, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0039] First embodiment

[0040] Reference Figure 1 In this embodiment, a super-relation extraction method based on the collaboration of multiple large language models is provided. The multiple large language models include a prompt word optimization model, a super-relation extraction model and a correction model.

[0041] It is worth noting that the prompt word optimization model, the super-relation extraction model and the correction model are all large language models. In this embodiment, in order to facilitate distinction, they are named according to the functions performed in each step. Therefore, the three models can actually be three different language models or the same language model, which is not limited in this embodiment.

[0042] In this embodiment, the hyper-relation extraction method based on the collaboration of multiple large language models includes the following steps:

[0043] Step S10, presetting the extraction prompt words, optimization prompt words and correction prompt words;

[0044] In this embodiment, the extraction prompt words, optimization prompt words and correction prompt words are all text instructions input into the large language model for the model to generate corresponding super-relation extraction rules, so that the large language model generates corresponding super-relation extraction rules based on these prompt words.

[0045] Step S20: obtaining a text to be extracted, and inputting the text to be extracted and the optimized prompt word into the prompt word optimization model, so that the prompt word optimization model analyzes the content of the text to be extracted based on the optimized prompt word to generate prompt word optimized content;

[0046] Step S30, obtaining the optimized content of the prompt word, and updating the extracted prompt word according to the optimized content of the prompt word to obtain the optimized extracted prompt word;

[0047] In this embodiment, a text input by the user or collected by the system from which a hyperrelation needs to be extracted is obtained as the text to be extracted. The text to be extracted and the optimization prompt words are input into the prompt word optimization model. The prompt word optimization model analyzes the text to be extracted based on the rules defined in the optimization prompt words and generates prompt word optimization content.

[0048] In this embodiment, the prompt word optimization content refers to the data output by the large language model for optimizing and adjusting the user-preset extraction prompt words. For the large language model, the prompt word optimization content updates the extraction prompt words, that is, updates the super-relation extraction rules.

[0049] The prompt word optimization model refers to a large language model used to optimize the extracted prompt words and generate prompt word optimized content.

[0050] It should be noted that the purpose of steps S20 and S30 is to provide a training target (i.e., more accurate extraction of prompt words) to the large language model, so that the large model can adjust its own super-relationship extraction rules according to the training target, and use its own powerful learning and computing capabilities to quickly and accurately identify and extract super-relationships in the text to be extracted.

[0051] Step S40: inputting the optimized extraction prompt words and the text to be extracted into the super-relationship extraction model, so that the super-relationship extraction model extracts the super-relationship in the text to be extracted according to the optimized extraction prompt words;

[0052] In this embodiment, after obtaining the optimized extraction prompt words output by the large language model, the optimized extraction prompt words and the text to be extracted are input into the hyper-relation extraction model.

[0053] The super-relation extraction model refers to a large language model used to extract super-relationships from the text to be extracted based on optimized extraction prompt words.

[0054] Step S50 , obtaining an initial super relation, inputting the initial super relation and the correction prompt word into the correction model, so that the correction model can proofread the initial super relation based on the correction prompt word to obtain a correction super relation.

[0055] In this embodiment, after obtaining the initial super-relation output by the super-relation extraction model, a correction model is introduced to verify the obtained initial super-relation.

[0056] The correction model refers to a large language model used to verify the initial super relation and output the verified corrected super relation.

[0057] In some optional implementations, the revised model can be verified from the following two dimensions:

[0058] 1. Check whether the hyperrelation format is correct based on the correction prompt word. If the format is incorrect, correct the hyperrelation according to the correct format;

[0059] Second, based on the correction prompt words, check whether the hyperrelationship content correctly reflects the information contained in the input text. If the content is incorrect, correct the hyperrelationship according to the text.

[0060] The initial super-relationship that is found to have errors is corrected, and the initial super-relationship that is not found to have errors is retained to obtain the final corrected super-relationship.

[0061] In the technical solution provided in this embodiment, the large language model's own powerful learning ability is used to optimize the extraction prompt words used as extraction rules, and a correction mechanism is given to correct the generated super-relationships. This allows for fast and accurate text super-relationship extraction without the need for fine-tuning the model parameters through supervised learning.

[0062] In the second embodiment

[0063] Based on the first embodiment, this embodiment provides an embodiment for setting prompt words based on the super-relation extraction method based on the collaboration of multiple large language models described in the first embodiment, which is specifically as follows:

[0064] (1) The extracted prompt words meet the following constraints:

[0065] The task type is information extraction task;

[0066] The output format includes a super-relationship quadruple of (head entity, relation, tail entity, auxiliary key-value pair);

[0067] The output includes a hint about the auxiliary model narrowing down the scope of the auxiliary key-value pairs.

[0068] For example:

[0069] "You are a natural language processing expert who first needs to extract entities and relations from the input sentence. Additionally, you must extract auxiliary key-value pairs. Auxiliary key-value pairs are additional information that complements entities and relations. Auxiliary key-value pairs add another relation-entity pair to a triple. Auxiliary key-value pairs are always 'key:value' pairs, where the key is one of the auxiliary key-value pair relations and the value is an entity.

[0070] Here, let me give you a few examples:

[0071] Example 1: Acadia University is an undergraduate university located in Wolfville, Nova Scotia, Canada, with some master's programs and one doctoral program.

[0072] Output: (Acadia University, Wolfville, Nova Scotia, Headquarters, Country: Canada)

[0073] The following is information about candidate secondary key-value pairs for the current input statement. This information can help you determine the secondary key-value pairs more accurately:

[0074] (2) The optimized prompt words meet the following constraints:

[0075] The task type is optimization task;

[0076] The output content includes potential auxiliary key-value pair information in the text to be extracted;

[0077] Contains guiding instructions to guide the large language model to first extract entities in the sentence, then classify the entities, and finally output information other than people and places as potential auxiliary key-value pairs.

[0078] For example:

[0079] “You need to understand the input sentence and identify the categories of information, which may include time, identity, location, quantity, organization, and other specific categories such as finance.

[0080] example:

[0081] She is best known for her two best-selling novels, The Fountainhead (1943) and Atlas Shrugged (1957), and for developing a philosophical system she called Objectivism.

[0082] Output: This sentence contains information about time and work. "

[0083] (3) The modified prompt word satisfies the following constraints:

[0084] The output format includes a super-relationship quadruple of (head entity, relation, tail entity, auxiliary key-value pair);

[0085] The format correction scheme satisfies the following requirements: 4-tuples with missing information are deleted, and 4-tuples with excessive information are pruned while the main triples (head entity, relationship, tail entity) remain unchanged, thus removing redundant information.

[0086] The semantic correction scheme satisfies: comparing the sentence composed of the super-relationship quadruple itself with the original corpus, and checking whether the super-relationship quadruple correctly translates the meaning of the original corpus.

[0087] For example:

[0088] You are an expert in natural language processing. You need to determine whether the hyperrelation correctly reflects the semantic information of the sentence and whether there is a problem with its format. If there is a problem, fix it.

[0089] The correct format is: (Entity1, Entity2, Relation, Qualifier Key: QualifierValue)

[0090] Here is an example:

[0091] Example 1: Input: Sentence: Acadia University is an undergraduate university located in Wolfville, Nova Scotia, Canada, with some master's programs and one doctoral program.

[0092] Hyperlink: (Acadia University, Wolfville, Nova Scotia, Headquarters, Country: Canada)

[0093] Output: (Acadia University, Wolfville, Nova Scotia, Headquarters, Country: Canada)

[0094] Example 2: Input: Sentence: After the Norman Reconquest, Palermo became the capital of a new kingdom (from 1130 to 1816), the Kingdom of Sicily and the Holy Roman Empire, ruled by Pope XXXII of the Holy Roman Empire and Pope XXXVI of Germany.

[0095] (XXXIV of Germany, Palermo, capital, position: Roman XX)

[0096] Output: (Palermo, capital of the Kingdom of Sicily, 1130-1816)

[0097] Example 3: Input example sentence: In 1975, XX and XX founded Microsoft Corporation to develop computer software business.

[0098] Super Relationship: (Microsoft, XX·XX, founding member; Microsoft, XX·XX, founding member, time: 1975)

[0099] Output: (Microsoft, XX·XX, founder, time: 1975)

[0100] (Microsoft, XX·XX, founding member, time: 1975)

[0101] If there are errors, please correct them and only output the corrected hyperrelations. You should adhere strictly to this format. "

[0102] In addition, as an implementation solution, this embodiment also provides a super-relation extraction result obtained by using the super-relation extraction method based on the collaboration of multiple large language models as described in any of the above items.

[0103] In addition, as an implementation solution, refer to Figure 2 This embodiment further provides a super-relation extraction device, which includes:

[0104] The prompt word optimization model 100 is pre-set with optimized prompt words, and is used to analyze the content of the text to be extracted input by the user based on the optimized prompt words to generate prompt word optimized content;

[0105] An extraction prompt word optimization module 200 is used to update the extraction prompt word according to the prompt word optimization content to obtain an optimized extraction prompt word;

[0106] A super-relation extraction model 300 is pre-set with extraction prompt words, and is used to extract the super-relationship in the text to be extracted according to the optimized extraction prompt words to obtain an initial super-relationship;

[0107] The correction model 400 is pre-set with correction prompt words, and is used to proofread the initial super-relationship based on the correction prompt words to obtain a corrected super-relationship.

[0108] In addition, as an implementation solution, Figure 3 This is a schematic diagram of the architecture of the hardware operating environment of the computer system involved in the embodiment of the present application.

[0109] like Figure 3 As shown, the computer system may include: a processor 1001, such as a CPU, a memory 1005, a user interface 1003, a network interface 1004, and a communication bus 1002. The communication bus 1002 is used to implement communication between these components. The user interface 1003 may include a display and an input unit such as a keyboard. Optionally, the user interface 1003 may also include a standard wired interface or a wireless interface. The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface). The memory 1005 may be a high-speed RAM memory or a non-volatile memory, such as a disk storage device. The memory 1005 may also be a storage device independent of the processor 1001.

[0110] Those skilled in the art will understand that Figure 3 The computer system architecture shown in the figure does not constitute a limitation of the computer system, and may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently.

[0111] like Figure 3 As shown, the memory 1005 as a storage medium may include an operating system, a network communication module, a user interface module and a computer program. Among them, the operating system is a program that manages and controls the hardware and software resources of the computer system, the operation of the computer program and other software or programs.

[0112] exist Figure 3In the computer system shown, the user interface 1003 is mainly used to connect to the terminal and communicate data with the terminal; the network interface 1004 is mainly used to communicate data with the background server; the processor 1001 can be used to call the computer program stored in the memory 1005.

[0113] In this embodiment, the computer system includes: a memory 1005, a processor 1001, and a computer program stored in the memory and executable on the processor, wherein:

[0114] When the processor 1001 calls the computer program stored in the memory 1005, it performs the following operations:

[0115] Pre-set extraction prompt words, optimization prompt words and correction prompt words;

[0116] Acquire a text to be extracted, and input the text to be extracted and the optimized prompt word into the prompt word optimization model, so that the prompt word optimization model analyzes the content of the text to be extracted based on the optimized prompt word to generate prompt word optimized content;

[0117] Obtaining the optimized content of the prompt word, and updating the extracted prompt word according to the optimized content of the prompt word to obtain the optimized extracted prompt word;

[0118] Inputting the optimized extraction prompt words and the text to be extracted into the super-relationship extraction model, so that the super-relationship extraction model extracts the super-relationship in the text to be extracted according to the optimized extraction prompt words;

[0119] An initial super relation is obtained, and the initial super relation and the correction prompt word are input into the correction model, so that the correction model proofreads the initial super relation based on the correction prompt word, obtains the correction super relation, and outputs it.

[0120] Furthermore, those skilled in the art will appreciate that all or part of the steps in the method of the above-described embodiment can be implemented by instructing the relevant hardware through a computer program. The computer program includes program instructions, which can be stored in a storage medium that is a computer-readable storage medium. The program instructions are executed by at least one processor in a computer system to implement the steps in the process of the above-described method embodiment.

[0121] Therefore, the present application also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the various steps of the super-relation extraction method based on the collaboration of multiple large language models as described in the above embodiment.

[0122] The computer-readable storage medium may be any computer-readable storage medium that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a magnetic disk, or an optical disk.

[0123] It should be noted that since the storage medium provided in the embodiments of this application is the storage medium used to implement the method of the embodiments of this application, based on the method described in the embodiments of this application, those skilled in the art will be able to understand the specific structure and deformation of the storage medium, and therefore will not be described in detail here. All storage media used in the method of the embodiments of this application fall within the scope of protection to be provided by this application.

[0124] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0125] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems) and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0126] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0127] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 The steps for the function specified in one or more boxes.

[0128] It should be noted that in the claims, any reference signs placed between parentheses shall not be construed as limiting the claims. The word "comprising" does not exclude the presence of components or steps not listed in the claim. The word "a" or "an" preceding a component does not exclude the presence of a plurality of such components. The present application may be implemented by means of hardware comprising several different components and by means of a suitably programmed computer. In a unit claim enumerating several means, several of these means may be embodied by one and the same item of hardware. The use of the words first, second, and third etc. does not indicate any order. These words may be interpreted as names.

[0129] Although the preferred embodiments of the present application have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present application.

[0130] Obviously, those skilled in the art may make various changes and modifications to this application without departing from the spirit and scope of this application. Thus, if these modifications and variations of this application fall within the scope of the claims of this application and their equivalents, this application is intended to include these modifications and variations.

Claims

1. A hyper-relation extraction method based on the collaboration of multiple large language models, characterized in that: The large language model includes a prompt word optimization model, a super-relation extraction model and a correction model. The method includes the following steps: Pre-set extraction prompt words, optimization prompt words and correction prompt words; Acquire a text to be extracted, and input the text to be extracted and the optimized prompt word into the prompt word optimization model, so that the prompt word optimization model analyzes the content of the text to be extracted based on the optimized prompt word to generate prompt word optimized content; Obtaining the optimized content of the prompt word, and updating the extracted prompt word according to the optimized content of the prompt word to obtain the optimized extracted prompt word; Inputting the optimized extraction prompt words and the text to be extracted into the super-relationship extraction model, so that the super-relationship extraction model extracts the super-relationship in the text to be extracted according to the optimized extraction prompt words; A super relation is obtained and used as an initial super relation, and the initial super relation and the correction prompt word are input into the correction model so that the correction model can proofread the initial super relation based on the correction prompt word, obtain the correction super relation and output it.

2. The method according to claim 1, wherein The extracted prompt words meet the following constraints: The task type is information extraction task; The output format includes a super-relationship quadruple of "head entity, relationship, tail entity, and auxiliary key-value pairs"; The output includes a hint about the auxiliary model narrowing down the scope of the auxiliary key-value pairs.

3. The method according to claim 1, wherein The optimization prompt word meets the following constraints: The task type is optimization task; The output content includes potential auxiliary key-value pair information in the text to be extracted; Contains guiding instructions to guide the large language model to first extract entities in the sentence, then classify the entities, and finally output information other than people and places as potential auxiliary key-value pairs.

4. The method according to claim 1, wherein The modified prompt word meets the following constraints: The task type is a correction task; The output format includes a super-relationship quadruple of "head entity, relationship, tail entity, and auxiliary key-value pairs"; The format correction scheme satisfies the following requirements: 4-tuples with missing information are deleted, and 4-tuples with excessive information are trimmed while the main triples (head entity, relationship, tail entity) remain unchanged, removing redundant information from the auxiliary key-value pairs. The semantic correction scheme satisfies: comparing the sentence composed of the super-relationship quadruple itself with the original corpus, and checking whether the super-relationship quadruple correctly translates the meaning of the original corpus.

5. A super-relation extraction device, characterized in that: The super-relation extraction device comprises: A prompt word optimization model is pre-set with optimized prompt words, and is used to analyze the content of the text to be extracted input by the user based on the optimized prompt words to generate prompt word optimized content; An extraction prompt word optimization module, configured to update the extraction prompt word according to the prompt word optimization content to obtain an optimized extraction prompt word; A super-relation extraction model is pre-set with extraction prompt words, and is used to extract the super-relation in the text to be extracted according to the optimized extraction prompt words, obtain the super-relation and use it as the initial super-relation; The correction model is pre-set with correction prompt words, and is used to proofread the initial super-relationship based on the correction prompt words to obtain a corrected super-relationship.

6. A computer system, characterized in that: The computer system includes: a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, the steps of the hyper-relation extraction method based on collaboration of multiple large language models as described in any one of claims 1 to 4 are implemented.

7. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the super-relation extraction method based on collaboration of multiple large language models according to any one of claims 1 to 4.

Images