Chinese composition intelligent reviewing method, system and equipment based on large language model and storage medium

Abstract

The invention belongs to the technical field of artificial intelligence, and particularly discloses a Chinese composition reviewing method, system and device based on a large language model and a storage medium, and the method comprises the steps: carrying out the image preprocessing and OCR recognition of a paper composition, and obtaining a composition material and composition content; concurrently inputting the composition material, the composition content and the prompt into different preset large language models for reviewing, and outputting to obtain different reviewing results; determining a priority sequence of each preset large language model review result, integrating the corresponding specific analysis results according to the sequence to obtain a composition review analysis result, and adding the corresponding scores to obtain a total composition score; and outputting the adjusted composition review analysis result and the composition total score to a request end. According to the method, the reliability of Chinese composition review scores and the composition review efficiency are improved through a unified evaluation standard.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, specifically to technical fields such as computer vision, deep learning, and large models, and in particular to a method, system, device, and storage medium for intelligent review of Chinese compositions based on a large language model. Background Art

[0002] In Chinese language education, essay reviewing has traditionally been a time-consuming and subjective task, primarily performed manually by teachers. While this method provides targeted feedback to students, it is inefficient. Furthermore, as class sizes increase, teachers' essay review workload increases dramatically, leading to a decline in the quality and timeliness of feedback. Furthermore, the subjectivity of manual essay reviewing means that grading criteria can vary from person to person, lacking uniformity and objectivity.

[0003] To improve the efficiency and consistency of essay grading, automatic essay scoring (AES) systems have been gradually developed and applied. These systems initially relied on simple statistical machine learning algorithms that could detect basic grammatical and spelling errors, but they performed poorly in understanding the deeper semantics and logic of essays. This meant that the systems struggled to provide constructive feedback, especially when dealing with complex styles and creative expressions, and their scoring accuracy was limited. Furthermore, the lack of personalized feedback was a significant drawback of early automatic scoring systems.

[0004] With the maturity of deep learning technology, and in particular the emergence of pre-trained large language models, automatic essay scoring systems have seen significant improvements. These deep learning-based systems, trained on large corpora, are able to more accurately understand the semantic structure and logic of essays, providing more nuanced scoring and more targeted feedback. Large language models address the challenge of maintaining consistent scoring for long, complex essays. Summary of the Invention

[0005] The present application provides a method, system, device and storage medium for intelligent review of Chinese compositions based on a large language model.

[0006] According to a first aspect of the present disclosure, a method for intelligent review of Chinese compositions based on a large language model is provided, comprising: performing image preprocessing and OCR recognition on paper compositions to obtain composition materials and composition content; concurrently inputting the composition materials and composition content into different preset large language models for review, and outputting different review results; determining specific analysis results and corresponding scores of the reviews of each preset large language model, integrating the corresponding specific analysis results in sequence to obtain a composition review analysis result, and adding up the corresponding scores to obtain a total composition score; and outputting the adjusted composition review analysis result and the composition total score to a requesting end.

[0007] According to a second aspect of the present disclosure, a Chinese composition intelligent review system based on a large language model is provided, comprising: an input module configured to perform image preprocessing and OCR recognition on paper compositions to obtain composition materials and composition content; a review module configured to concurrently input composition materials and composition content into different preset large language models for review, and output different review results; an output module configured to determine specific analysis results and corresponding scores of the reviews of each preset large language model, integrate the corresponding specific analysis results in sequence to obtain a composition review analysis result, and add up the corresponding scores to obtain a total composition score; and output the adjusted composition review analysis result and composition total score to a requesting end.

[0008] According to a third aspect of the present disclosure, an electronic device is provided, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the method described in any implementation manner in the first aspect.

[0009] According to a fourth aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, where the computer instructions are used to cause a computer to execute the method described in any implementation manner of the first aspect.

[0010] According to a fifth aspect of the present disclosure, a computer program product is provided, comprising a computer program, which implements the method described in any implementation manner of the first aspect when executed by a processor.

[0011] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present disclosure, nor are they intended to limit the scope of the present disclosure. Other features of the present disclosure will become readily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] The accompanying drawings are provided to facilitate a better understanding of the present invention and do not constitute a limitation of the present disclosure.

[0013] Figure 1 is an exemplary system architecture diagram in which the present disclosure may be applied;

[0014] Figure 2 is a flow chart of an embodiment of the method for intelligent review of Chinese composition according to the present disclosure;

[0015] Figure 3 is a flow chart of an embodiment of the method for intelligent review of Chinese composition according to the present disclosure;

[0016] Figure 4 It is a structural block diagram of a basic electronic device according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0017] The following description of exemplary embodiments of the present disclosure is made in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding. These details should be considered as merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0018] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in the present disclosure can be combined with each other. The present disclosure will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0019] Figure 1 An exemplary system architecture 100 is shown to which an embodiment of the essay correction method or essay correction apparatus of the present disclosure can be applied.

[0020] like Figure 1 As shown, system architecture 100 may include terminal devices 101, 102, 103, and 104, a network 105, and a server 106. Network 105 is a medium for providing communication links between terminal devices 101, 102, 103, and 104 and server 106. Network 105 may include various connection types, such as wired or wireless communication links or fiber optic cables.

[0021] Users can use terminal devices 101, 102, 103, 104 to interact with server 106 via network 105 to receive or send information, etc. Various client applications can be installed on terminal devices 101, 102, 103, 104.

[0022] Terminal devices 101, 102, 103, and 104 can be either hardware or software. When terminal devices 101, 102, 103, and 104 are hardware, they can be various electronic devices, including but not limited to smartphones, tablet computers, laptop computers, and desktop computers. When terminal devices 101, 102, 103, and 104 are software, they can be installed in the aforementioned electronic devices. They can be implemented as multiple software programs or software modules, or as a single software program or software module. This is not specifically limited here.

[0023] The server 106 can provide various services. For example, the server 106 can analyze and process the essays to be reviewed obtained from the terminal devices 101, 102, 103, and 104, and generate processing results (for example, output the review results to the requesting end).

[0024] It should be noted that server 106 can be either hardware or software. When server 106 is hardware, it can be implemented as a distributed server cluster consisting of multiple servers, or as a single server. When server 106 is software, it can be implemented as multiple software programs or software modules (for example, to provide distributed services), or as a single software program or software module. No specific limitations are given herein.

[0025] It should be noted that the Chinese composition review method provided in the embodiment of the present disclosure is generally executed by the server 106 , and accordingly, the Chinese composition review device is generally set in the server 106 .

[0026] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is merely illustrative and any number of terminal devices, networks and servers may be provided according to actual needs.

[0027] Continue to refer Figure 2 , which shows an embodiment of the process 200 of the intelligent method for reviewing Chinese composition of the present disclosure. The intelligent method for reviewing Chinese composition includes the following steps:

[0028] Step 201: perform image preprocessing and OCR recognition on the paper composition to obtain composition materials and composition content.

[0029] In this embodiment, the implementation subject of the Chinese composition intelligent review method (for example Figure 1 The server 106 shown performs image preprocessing and OCR on the paper composition to obtain composition materials and composition content. Composition materials refer to the given composition materials and writing requirements, while composition content refers to the composition to be reviewed. The subject information corresponding to the composition materials and composition content is Chinese language, and the specific academic level can be elementary school, junior high school, or high school, which is not specifically limited in this embodiment.

[0030] In step 202, the composition materials, composition content and prompts are input into different preset large language models for review, and different review results are output.

[0031] In this embodiment, the execution subject (for example Figure 1The server 106 shown will concurrently input the composition materials, composition content and prompts into different preset large language models, thereby obtaining different review results. That is, based on the multi-model review method, combined with the characteristics of different review dimensions, different preset large language models are requested. Here, the review dimensions may include: thinking level dimension, logical structure dimension, content element dimension, language feature dimension, and emotional expression dimension. Different review dimensions correspond to different prompt content and different preset large language models. Different open source large language models (such as GPT, BERT) are fine-tuned to become the preset large language model.

[0032] After obtaining the composition materials and composition content, the above-mentioned execution entity will concurrently request different preset large language models, that is, concurrently input the composition materials, composition content and prompts into different preset large language models, so that multiple preset large language models can correct the compositions at the same time, thereby outputting corresponding multiple correction results.

[0033] It should be noted that large language models (LLMs) are AI models designed to understand and generate human language. These models, with their massive number of parameters, often exceeding one billion, and trained on massive text corpora, demonstrate unprecedented language understanding and generation capabilities. They are capable of performing a wide range of complex language tasks, covering a wide range of application scenarios, from text summarization and machine translation to sentiment analysis. The impressive performance of LLMs stems from their deep neural network architecture, particularly the Transformer-based design. This architecture revolutionizes traditional sequence modeling methods by effectively capturing long-range dependencies through the self-attention mechanism, improving model efficiency and effectiveness. In recent years, with the continuous advancement of computing hardware and the continuous improvement of deep learning algorithms, the size and complexity of LLMs have grown exponentially. This not only prolongs training cycles but also improves the models' generalization and understanding of subtle linguistic structures. Common LLMs include, but are not limited to, GPT, BERT, T5, and ERNIE.

[0034] In some optional implementations of this embodiment, the review dimension may include one or more of the following dimensions: a thinking level dimension, a logical structure dimension, a content element dimension, a language feature dimension, and an emotional expression dimension. That is, the above-mentioned execution subject may conduct a thinking level review, a logical structure review, a content element review, a language feature review, and an emotional expression review on the composition content. More specifically, the above-mentioned execution subject may concurrently input the composition materials, composition content, and prompts into the thinking level preset large language model, the logical structure preset large language model, the content element preset large language model, the language feature preset large language model, and the emotional expression preset large language model, thereby outputting the corresponding thinking level review results, logical structure review results, content element review results, language feature review results, and emotional expression review results. The review result of each dimension includes the review analysis result and the specific score of the composition content under that dimension.

[0035] Step 203: determine the priority order of each preset large language model review result, integrate the corresponding specific analysis results in order to obtain the composition review analysis result, and add up the corresponding scores to obtain the total score of the composition.

[0036] In some optional implementations of this embodiment, the review dimensions include: thinking level dimension, logical structure dimension, content element dimension, language feature dimension and emotional expression dimension. In a multi-model request link, multiple types of data are responded to concurrently, but for the display of the output end page, concurrent display is unreasonable. Therefore, based on the user's review habits, a priority order can be set for multiple review dimensions, for example, the priority order is: thinking level dimension> logical structure> content element> language feature> emotional expression. The review results of each dimension include the review analysis results and the score of the composition under this dimension. Then, before processing each review result, it will be judged whether the review result with a higher priority is output. If it does not meet the priority judgment (that is, the review result with a higher priority than the review result has not been output), the review result will be put back into the queue to be processed, that is, the review result with a high priority must be output first, and after its output is completed, the review result will be output.

[0037] Step 204: Output the adjusted composition review analysis results and the composition total score to the requesting end.

[0038] In some optional implementations of this embodiment, the system outputs the review and analysis results and total score of the essay image to the requesting end for reference by students, parents, and teachers. Simultaneously, the system uses the review results to optimize and train each pre-set large language model, further improving the system's processing capabilities.

[0039] The intelligent review method for Chinese compositions provided by the disclosed embodiment first performs image preprocessing and OCR recognition on paper compositions to obtain composition materials and composition content; then, the composition materials, composition content, and prompts are concurrently input into different preset large language models for review, and different review results are output; then, the priority order of the review results of each preset large language model is determined, the corresponding specific analysis results are sequentially integrated to obtain the composition review analysis result, and the corresponding scores are added to obtain the total score of the composition; finally, the adjusted composition review analysis result and the total score of the composition are output to the request end. The intelligent review method for Chinese compositions in this embodiment requests different preset large language models according to different review dimensions and their characteristics. At the same time, in order to ensure the response time, the existing serial streaming link request method is changed to a multi-model concurrent request, thereby improving the efficiency of composition review; in addition, the output rendering is performed in sequence according to the priority order of each review dimension, thereby reducing the user's waiting time and improving the user experience.

[0040] Continue to refer Figure 3 , Figure 3 FIG300 shows another embodiment of the intelligent review method for Chinese composition according to the present disclosure. The composition review method includes the following steps:

[0041] Step 301: The paper composition is subjected to image preprocessing and OCR recognition to obtain composition materials and composition content.

[0042] In this embodiment, the execution subject of the Chinese composition intelligent review method (for example Figure 1 The server 105 shown in the figure performs image preprocessing and OCR recognition on the paper composition to obtain the composition materials and composition content. Step 301 is basically the same as step 201 in the above embodiment. The specific implementation method can refer to the above description of step 201 and will not be repeated here.

[0043] Step 302: Input the composition materials, composition content and thinking level review prompts into a preset large language model, and output the thinking level review results.

[0044] In this embodiment, the execution entity inputs the composition materials, composition content, and thought-level review prompts into a pre-set large language model, which then outputs the thought-level review results. Because the thought-level review of an essay depends on context, using a pre-set large language model is more effective. Furthermore, in terms of the performance of domestic and foreign large language models, domestic large language models are more effective. Therefore, a fine-tuned domestic large language model (e.g., Qwen-72B) is used as the pre-set large language model for thought-level review.

[0045] Step 303: Input the composition content and logical structure review prompts into a preset large language model, and output the logical structure review results.

[0046] In this embodiment, the execution entity inputs the essay content and logical structure review prompts into a preset large language model, which then outputs the logical structure review results. Because the review of an essay's logical structure depends on context, using a preset large language model is more effective. Furthermore, domestic large language models are more effective than foreign large language models, so a fine-tuned domestic large language model (e.g., Qwen-72B) is used as the preset large language model for logical structure review.

[0047] In step 304, the composition content and content element review prompts are input into a preset large language model, and the content element review results are output.

[0048] In this embodiment, the execution entity inputs the essay content and content element review prompts into a preset large language model, which then outputs the content element review results. Because the review of essay content elements depends on context, using a preset large language model is more effective. Furthermore, in terms of the performance of domestic and foreign large language models, domestic large language models are more effective. Therefore, a fine-tuned domestic large language model (e.g., Qwen-72B) is used as the preset large language model for content element review.

[0049] Step 305: Input the composition content and language feature review prompts into a preset large language model, and output the language feature review results.

[0050] In this embodiment, the execution entity inputs the essay content and language feature review prompts into a preset large language model, which then outputs the language feature review results. Because the review of essay language features depends on context, using a preset large language model is more effective. Furthermore, domestic large language models are superior in terms of performance compared to foreign large language models. Therefore, a fine-tuned domestic large language model (e.g., Qwen-72B) is used as the preset large language model for language feature review.

[0051] Step 306: Input the composition content and the emotion expression review prompt into a preset large language model, and output the emotion expression review result.

[0052] In this embodiment, the execution entity inputs the essay content and emotional expression review prompts into a preset large language model, which then outputs the language feature review results. Because the emotional expression review of essays depends on context, using a preset large language model is more effective. Furthermore, in terms of the performance of domestic and foreign large language models, domestic large language models are more effective. Therefore, a fine-tuned domestic large language model (e.g., Qwen-72B) is used as the preset large language model for emotional expression review.

[0053] It should be noted that steps 302 to 306 are executed concurrently. Based on the multi-model review method, the characteristics of different review dimensions are combined to request different model services, thereby improving the efficiency of essay review.

[0054] Step 307 , determine the priority order of each preset large language model review result, integrate the corresponding specific analysis results in order to obtain the composition review analysis result, and add up the corresponding scores to obtain the total score of the composition.

[0055] In this embodiment, the execution entity first determines the priority order corresponding to each review dimension; then, based on this priority order, it determines the output order of the multiple review results. In a multi-model request chain, multiple types of data are responded to concurrently, but for the display of the output page, concurrent display is unreasonable. Therefore, a priority order can be set for multiple review dimensions based on the user's review habits. For example, the priority order is: thinking level dimension > logical structure dimension > content element dimension > language feature dimension > emotional expression dimension.

[0056] Step 308: Output the adjusted composition review analysis results and the composition total score to the requesting end.

[0057] In this embodiment, the execution entity outputs the review and analysis results and the total score of the essay image to the requesting end for reference by students, parents, and teachers. Simultaneously, the system uses the review results to optimize and train each pre-set large language model, further improving the system's processing capabilities.

[0058] from Figure 3 It can be seen that with Figure 2 Compared with the corresponding embodiments, the intelligent review method for Chinese compositions in this embodiment highlights the steps of using different preset large language models to concurrently review the compositions to be reviewed and outputting multiple review results in sequence on the request end, thereby achieving multiple responses for one request based on multi-model concurrent requests, and rendering multiple responses in batches, reducing the user's waiting time and improving the user experience.

[0059] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device.

[0060] Figure 4 A schematic block diagram of an example electronic device 400 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are provided as examples only and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0061] Specifically, the processor 402 may include a central processing unit (CPU), or an application-specific integrated circuit (ASIC), or may be configured to implement one or more integrated circuits of the embodiments of the present application.

[0062] Among them, the memory 403 may include a large-capacity memory for data or instructions. By way of example and not limitation, the memory 403 may include a hard disk drive (HDD), a floppy disk drive, a solid-state drive (SSD), a flash memory, an optical disk, a magneto-optical disk, a magnetic tape, or a universal serial bus (USB) drive, or a combination of two or more of these. Where appropriate, the memory 403 may include a removable or non-removable (or fixed) medium. Where appropriate, the memory 403 may be inside or outside the data processing device. In a specific embodiment, the memory 403 is a non-volatile memory. In a specific embodiment, the memory 403 includes a read-only memory (ROM) and a random access memory (RAM). Where appropriate, the ROM may be a mask-programmed ROM, a programmable ROM (PROM), an erasable PROM (EPROM), an electrically erasable PROM (EEPROM), an electrically alterable ROM (EAROM) or a flash memory (FLASH), or a combination of two or more of these. Under appropriate circumstances, the RAM can be a static random access memory (SRAM) or a dynamic random access memory (DRAM), where the DRAM can be a fast page mode dynamic random access memory (FPMDRAM), an extended data output dynamic random access memory (EDODRAM), a synchronous dynamic random access memory (SDRAM), etc.

[0063] The memory 403 may be used to store or cache various data files that need to be processed and / or used for communication, as well as possible computer program instructions executed by the processor 11 .

[0064] The processor 402 reads and executes the computer program instructions stored in the memory 403 to implement any one of the large language model-based intelligent review methods for Chinese compositions in the above embodiments.

[0065] In one embodiment, the electronic device may further include a communication interface 404 and a bus 401. Figure 4 As shown, the processor 402 , the memory 403 , and the communication interface 404 are connected via a bus 401 and communicate with each other.

[0066] The communication interface 404 is used to enable communication between the various modules, devices, units, and / or equipment in the embodiments of the present application. The communication interface 404 can also enable data communication with other components such as external devices, image / data acquisition equipment, databases, external storage, and image / data processing workstations.

[0067] Bus 401 includes hardware, software, or both, and couples components of the electronic device to each other. Bus 401 includes, but is not limited to, at least one of the following: a data bus, an address bus, a control bus, an expansion bus, and a local bus. By way of example and not limitation, bus 401 may include an Accelerated Graphics Port (AGP) or other graphics bus, an Extended Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), a Hyper Transport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an InfiniBand interconnect, a Low Pin Count (LPC) bus, a memory bus, a Micro Channel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local Bus (VLB) bus, or other suitable buses, or a combination of two or more of these. Where appropriate, bus 401 may include one or more buses. Although embodiments herein describe and illustrate a particular bus, this application contemplates any suitable bus or interconnect.

[0068] The embodiment of the present application provides a computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements Figure 1 and Figure 2 The intelligent review method for Chinese composition based on large language model is provided in.

[0069] Among them, the readable storage medium can more specifically include but is not limited to: a portable disk, a hard disk, a random access memory, a read-only memory, an erasable programmable read-only memory, an optical storage device, a magnetic storage device or any suitable combination of the above.

[0070] In a possible implementation, the present invention can also be implemented in the form of a program product, which includes program code. When the program product is run on a terminal device, the program code is used to enable the terminal device to execute the steps of the intelligent review method for Chinese compositions based on a large language model provided in the first aspect.

[0071] The program code for executing the present invention may be written in any combination of one or more programming languages, and the program code may be executed entirely on the user device, partially on the user device, as an independent software package, partially on the user device and partially on a remote device, or entirely on the remote device.

[0072] The technical features of the above-mentioned embodiments can be combined arbitrarily. In order to make the description concise, not all possible combinations of the technical features in the above-mentioned embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0073] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art could make various modifications and improvements without departing from the spirit of the present application, all of which fall within the scope of protection of the present application. Therefore, the scope of protection of the present patent application shall be determined by the appended claims.

Claims

1. A method for intelligent review of Chinese composition based on a large language model, characterized in that: include: Perform image preprocessing and OCR recognition on paper essays to obtain essay materials and content; The composition materials, composition content and prompts are input into different preset large language models for review, and different review results are output; Determine the priority order of the review results of each preset large language model, integrate the corresponding specific analysis results in order to obtain the composition review analysis results, and add up the corresponding scores to obtain the total score of the composition; The adjusted essay review analysis results and the essay's total score are output to the requesting end.

2. The intelligent review method for Chinese composition according to claim 1, characterized in that: The image preprocessing and OCR recognition of the paper composition includes: Perform tilt correction and splicing on composition images; Perform OCR recognition on the corrected composition image.

3. The intelligent review method for Chinese composition according to claim 1, characterized in that: The review includes: review of thinking level, review of logical structure, review of content elements, review of language features, and review of emotional expression; and The composition materials and content are input into different preset large language models for review, and different review results are output, including: Inputting composition materials, composition content, and thought-level review prompts into a preset large language model, and outputting thought-level review results; and / or Inputting the composition content and logical structure review prompts into a preset large language model, and outputting the logical structure review results; and / or Inputting the composition content and content element review prompts into a preset large language model, and outputting the content element review results; and / or Inputting the composition content and language feature review prompts into a preset large language model, and outputting the language feature review results; and / or The composition content and emotion expression review prompts are input into the preset large language model, and the emotion expression review results are output.

4. The intelligent review method for Chinese composition according to claim 3, characterized in that: The priority order of each preset large language model review result is determined, the corresponding specific analysis results are integrated in order to obtain the composition review analysis result, and the corresponding scores are added to obtain the total composition score, including: Determine the priority order of the review results of each preset large language model; Integrate the corresponding specific analysis results based on the priority order to obtain the composition review analysis results; The total score of the composition is obtained by adding up the review scores of each preset large language model.

5. The intelligent review method for Chinese composition according to claim 3, characterized in that: The preset large language model includes: LLaMA, ChatGLM, and / or Tongyi Qianwen Large Language Model.

6. The intelligent review method for Chinese composition according to claim 4, characterized in that: Outputting the adjusted essay review analysis results and the essay total score to the requesting end includes: Upload the essay review analysis results and the total score to the teacher review terminal; The composition review results and total composition scores are used to optimize and train each preset large language model.

7. A Chinese composition intelligent review system, comprising: The input module is configured to perform image preprocessing and OCR recognition on the paper composition to obtain composition materials and composition content; The review module is configured to input the composition materials and composition content into different preset large language models for review, and output different review results; The output module is configured to determine the specific analysis results and corresponding scores of each preset large language model review, integrate the corresponding specific analysis results in order to obtain the composition review analysis results, and add up the corresponding scores to obtain the total composition score; the adjusted composition review analysis results and total composition score are output to the request end.

8. The system according to claim 7, characterized in that The review includes: review of thinking level, review of logical structure, review of content elements, review of language features, and review of emotional expression; and The review module includes: The thinking level review submodule is configured to input the composition materials, composition content and thinking level review prompts into a preset large language model, and output the thinking level review results; and / or A logical structure review submodule is configured to input the composition content and the logical structure review prompts into a preset large language model and output the logical structure review results; and / or A content element review submodule is configured to input the composition content and content element review prompts into a preset large language model and output a content element review result; and / or A language feature review submodule is configured to input the composition content and language feature review prompts into a preset large language model and output a language feature review result; and / or The emotion expression review submodule is configured to input the composition content and emotion expression review prompts into a preset large language model, and output the emotion expression review results.

9. The system according to claim 8, characterized in that The output module includes: A priority order submodule is configured to determine the priority order of each preset large language model review result; The composition review and analysis submodule is configured to integrate the corresponding specific analysis results based on the priority order to obtain a composition review and analysis result; The composition total score molecular module is configured to add up the review scores of the corresponding preset large language models to obtain the total score of the composition. The output submodule is configured to output the composition review analysis results and the total composition score to the teacher review page.

10. The intelligent review device for Chinese composition based on a large language model is characterized by: include: a memory for storing instructions; A processor is used to read the instructions stored in the memory and execute the Chinese composition intelligent review method based on a large language model according to any one of claims 1 to 6.

11. A computer-readable storage medium, characterized in that The computer-readable storage medium stores instructions, and when the instructions are executed on a computer, the computer is caused to execute the intelligent review method for Chinese compositions based on a large language model as described in any one of claims 1 to 6.

Images