Education theory based multi-granularity assessment method, device, medium and product

Abstract

This application provides a method, device, medium, and product for multi-granularity evaluation of academic papers based on educational theory, relating to the field of artificial intelligence technology. The method includes: acquiring paper data to be evaluated; performing text extraction processing on the paper data to obtain structured features; performing multi-granularity evaluation processing on the structured features based on a preset multi-dimensional hierarchical prompting and guiding large language model to obtain a multi-granularity evaluation result for the paper data; the multi-dimensional hierarchical prompting is constructed based on a multi-granularity evaluation dimension system based on educational theory, and the multi-granularity evaluation result includes multi-granularity dimension scores. The technical solution of this application can achieve interpretable paper evaluation.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to a method, device, medium, and product for multi-granularity evaluation of papers based on educational theory. Background Technology

[0002] In recent years, agent technology driven by Large Language Models (LLMs) has rapidly developed and achieved significant application results in the field of educational intelligence. After pre-training on large-scale and diverse corpora, LLMs possess powerful natural language understanding and logical reasoning capabilities, rich cross-domain professional knowledge, and can adapt to diverse educational scenarios and tasks based on contextual learning characteristics. They have gradually become a core technological support for intelligent educational applications such as intelligent thesis evaluation, academic tutoring, and academic diagnosis. Among these, intelligent evaluation of graduation theses is an important research direction for academic quality management and intelligent teaching evaluation in universities.

[0003] When applying Large Language Modeling (LLM) to thesis evaluation, related technologies generally employ a singular and generalized overall scoring approach. For example, they might directly input a simple prompt like "Please evaluate the thesis on a scale of 1-10," obtaining only the single overall score output by the LLM model as the evaluation result. However, academic evaluation of theses inherently requires complex, multi-dimensional reasoning. A general or singular prompting approach cannot fully leverage the LLM model's professional evaluation capabilities in academic dimensions. Furthermore, since the LLM model only outputs an overall score, it fails to characterize the differences in performance across different academic ability dimensions. This makes it impossible to provide students with dimensional, targeted feedback for improvement, nor can it support the logical traceability of the evaluation process.

[0004] In summary, the graduation thesis evaluation technology based on Large Language Model (LLM) has obvious defects. The single overall score prompt cannot adapt to the needs of multi-dimensional complex reasoning, and the evaluation results are only presented in the form of total score, which cannot reflect the subtle differences of each academic dimension, ultimately resulting in poor interpretability of the evaluation results. Summary of the Invention

[0005] The main objective of this application is to propose a multi-granularity evaluation method, device, medium, and product for papers based on educational theory, aiming to achieve interpretable paper evaluation.

[0006] To achieve the above objectives, a first aspect of this application proposes a multi-granularity evaluation method for papers based on educational theory, the method comprising: Obtain the data for the papers to be evaluated; The paper data is subjected to text extraction processing to obtain the structured features of the paper data; Based on the preset multidimensional hierarchical prompts, the large language model performs multi-granularity evaluation processing on the structured features to obtain the multi-granularity evaluation results of the paper data; the multidimensional hierarchical prompts are constructed based on the multi-granularity evaluation dimension system of educational theory, and the multi-granularity evaluation results include multi-granularity dimension scores.

[0007] In some embodiments, the method further includes: Based on Vygotsky's sociocultural theory and Bloom's taxonomy of educational objectives, a multi-granularity assessment dimension system is constructed; the multi-granularity assessment dimension system includes at least two fine-grained assessment dimensions from structure, logic, originality, writing, proficiency, and rigor. Based on the multi-granularity evaluation dimension system, a hierarchical two-stage prompt is constructed; the hierarchical two-stage prompt includes a first-stage prompt and a second-stage prompt. The first-stage prompt is used to guide the large language model to independently evaluate the fine-grained evaluation dimensions in the multi-granularity evaluation dimension system, and the second-stage prompt is used to guide the large language model to integrate the fine-grained evaluation dimensions into an overall evaluation; the multi-dimensional hierarchical prompt includes the first-stage prompt and the second-stage prompt.

[0008] In some embodiments, the multi-granularity evaluation processing of the structured features based on the preset multi-dimensional hierarchical prompting guidance large language model to obtain the multi-granularity evaluation results of the paper data includes: Based on the prompts and guidance of the first stage, the large language model independently evaluates the structured features from the fine-grained evaluation dimensions in the multi-granularity evaluation dimension system to obtain multiple dimension scores for the paper data; the multi-granularity dimension scores include the multiple dimension scores.

[0009] In some embodiments, the step of using a preset multi-dimensional hierarchical prompting large language model to perform multi-granularity evaluation processing on the structured features to obtain the multi-granularity evaluation results of the paper data further includes: Based on the second-stage prompts, the large language model integrates the various fine-grained evaluation dimensions into an overall evaluation; integrating the various fine-grained evaluation dimensions into an overall evaluation includes weighted summation of the scores of the multiple dimensions to obtain the overall score of the paper data, and generating feedback and improvement suggestions; the multi-granularity evaluation results also include the overall score, the feedback, and the improvement suggestions.

[0010] In some embodiments, the method further includes: The large language model is enhanced by context learning based on preset few-shot prompts; the few-shot prompts include paper sample data, and the model evaluation results corresponding to the paper sample data include at least multi-granularity dimension scores and overall scores.

[0011] In some embodiments, the method further includes: The large language model is enhanced by context learning based on preset role-playing prompts; the role-playing prompts are used to guide the large language model to act as a paper reviewer and to reason and evaluate the paper data.

[0012] To achieve the above objectives, a second aspect of this application proposes a multi-granularity evaluation device for papers based on educational theory, the device comprising: The acquisition module is used to acquire the data of the papers to be evaluated; The preprocessing module is used to perform text extraction processing on the paper data to obtain the structured features of the paper data; The multi-granularity evaluation module is used to perform multi-granularity evaluation processing on the structured features based on a preset multi-dimensional hierarchical prompting guidance large language model to obtain the multi-granularity evaluation results of the paper data; the multi-dimensional hierarchical prompting is constructed based on the multi-granularity evaluation dimension system of educational theory, and the multi-granularity evaluation results include multi-granularity dimension scores.

[0013] To achieve the above objectives, a third aspect of this application provides a computer device, the computer device including a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the paper multi-granularity evaluation method based on educational theory described in the first aspect.

[0014] To achieve the above objectives, a fourth aspect of the present application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the paper multi-granularity evaluation method based on educational theory described in the first aspect.

[0015] To achieve the above objectives, a fifth aspect of this application provides a computer program product storing a computer program that, when executed by a processor, implements the multi-granularity evaluation method for papers based on educational theory described in the first aspect.

[0016] The multi-granularity evaluation method, apparatus, computer device, computer-readable storage medium, and computer program product based on educational theory proposed in this application embodiment acquire the paper data to be evaluated; perform text extraction processing on the paper data to obtain the structured features of the paper data; perform multi-granularity evaluation processing on the structured features based on a preset multi-dimensional hierarchical prompting guidance large language model to obtain the multi-granularity evaluation result of the paper data; the multi-dimensional hierarchical prompting is constructed based on the multi-granularity evaluation dimension system of educational theory, and the multi-granularity evaluation result includes multi-granularity dimension scores.

[0017] Compared to traditional methods that rely on a single overall score to guide Large Language Models (LLMs) in paper evaluation, this application's embodiment acquires the paper data to be evaluated, first extracts and processes the text to obtain its structured features, and then uses a multi-dimensional hierarchical prompting system pre-constructed based on educational theory to guide the LLM in performing multi-granularity evaluation of these structured features, resulting in a multi-granularity evaluation result for the paper data. This multi-granularity evaluation result includes multi-granularity dimension scores. Thus, by constructing a multi-dimensional evaluation framework and hierarchical prompting strategy that integrates educational theory, this application's embodiment can provide fine-grained, interpretable evaluations that align with expert assessments, thereby addressing the technical problems of insufficient granularity, lack of educational theory support, and single evaluation dimensions inherent in traditional evaluation methods. Attached Figure Description

[0018] Figure 1 A flowchart illustrating the steps of the multi-granularity evaluation method for papers based on educational theory provided in some embodiments of this application; Figure 2 A flowchart illustrating the steps of the multi-granularity evaluation method for papers based on educational theory provided in this application in other embodiments; Figure 3 for Figure 1 A detailed flowchart of step S103; Figure 4 for Figure 1 A schematic diagram of another detailed step in step S103; Figure 5 A flowchart illustrating the steps of the multi-granularity evaluation method for papers based on educational theory provided in some embodiments of this application; Figure 6 A flowchart illustrating the steps of the multi-granularity evaluation method for papers based on educational theory provided in some embodiments of this application; Figure 7 A schematic diagram of the overall system framework of the paper multi-granularity evaluation method based on educational theory provided in an embodiment of this application in a complete embodiment; Figure 8 A schematic diagram of the structure of the multi-granularity evaluation device for papers based on educational theory provided in the embodiments of this application; Figure 9 This is a schematic diagram of the hardware structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0020] It should be noted that although functional modules are divided in the device schematic diagram and a logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the device or the order in the flowchart. The terms "first," "second," etc., in the specification, claims, and the aforementioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence.

[0021] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0022] First, the overall concept of the embodiments of this application will be explained.

[0023] Undergraduate theses are a crucial indicator of a student's overall academic ability during their undergraduate studies, impacting graduation eligibility and degree classification. Theses are typically lengthy and require evaluation across multiple dimensions, including structural completeness, academic rigor, and methodological soundness. Traditional manual evaluation methods are time-consuming and labor-intensive, and struggle to guarantee timely feedback and constructive personalized improvement suggestions. To address this issue, researchers have attempted to automate intelligent evaluation systems using artificial intelligence. Early natural language processing techniques based on manual rules and more recent neural network-based methods have made some progress in automated essay evaluation; however, their limited ability to extract high-level semantics and structured features makes them unsuitable for undergraduate thesis evaluation.

[0024] In recent years, large language models have made significant progress in the field of educational intelligence. These models, pre-trained on large and diverse corpora, possess powerful language understanding, reasoning abilities, and rich domain knowledge, as well as the capacity for context learning. Researchers have attempted to apply large language models to paper evaluation tasks, typically using simple overall scoring prompts (such as "Please rate the paper on a scale of 1-10") and returning a total score.

[0025] However, the academic evaluation of graduation theses is a task that requires complex reasoning across multiple dimensions. A general or single model is insufficient to fully stimulate the professional evaluation capabilities of the LLM model in the academic dimension. At the same time, the LLM model only outputs an overall score, which cannot characterize the performance differences of the thesis in different academic ability dimensions. It cannot provide students with dimension-based targeted improvement feedback, nor can it support the logical traceability of the evaluation process.

[0026] In summary, the graduation thesis evaluation technology based on Large Language Model (LLM) has obvious defects. The single overall score prompt cannot adapt to the needs of multi-dimensional complex reasoning, and the evaluation results are only presented in the form of total score, which cannot reflect the subtle differences of each academic dimension, ultimately resulting in poor interpretability of the evaluation results.

[0027] Based on this, this application proposes a method, apparatus, computer device, computer-readable storage medium, and computer program product for multi-granularity evaluation of papers based on educational theory. It acquires the paper data to be evaluated; performs text extraction processing on the paper data to obtain its structured features; and performs multi-granularity evaluation processing on the structured features based on a preset multi-dimensional hierarchical prompting and guidance large language model to obtain the multi-granularity evaluation result of the paper data. The multi-dimensional hierarchical prompting is constructed based on a multi-granularity evaluation dimension system of educational theory, and the multi-granularity evaluation result includes multi-granularity dimension scores.

[0028] Compared to traditional methods that rely on a single overall score to guide Large Language Models (LLMs) in paper evaluation, this application's embodiment acquires the paper data to be evaluated, first extracts and processes the text to obtain its structured features, and then uses a multi-dimensional hierarchical prompting system pre-constructed based on educational theory to guide the LLM in performing multi-granularity evaluation of these structured features, resulting in a multi-granularity evaluation result for the paper data. This multi-granularity evaluation result includes multi-granularity dimension scores. Thus, by constructing a multi-dimensional evaluation framework and hierarchical prompting strategy that integrates educational theory, this application's embodiment can provide fine-grained, interpretable evaluations that align with expert assessments, thereby addressing the technical problems of insufficient granularity, lack of educational theory support, and single evaluation dimensions inherent in traditional evaluation methods.

[0029] Based on the overall concept of the embodiments of this application described above, specific embodiments of the multi-granularity evaluation method, apparatus, computer device, computer-readable storage medium, and computer program product based on educational theory provided in the embodiments of this application are proposed. First, the various specific embodiments of the multi-granularity evaluation method for papers based on educational theory in the embodiments of this application are described in detail.

[0030] It should be noted that the embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence (AI) refers to the theories, methods, technologies, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0031] Foundational technologies for artificial intelligence generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interactive systems, and mechatronics. AI software technologies mainly encompass computer vision, robotics, biometrics, speech processing, natural language processing, and machine learning / deep learning.

[0032] Furthermore, in various specific embodiments of this application, when processing data related to user identity or characteristics, such as user information, user behavior data, user historical data, and user location information, user permission or consent is obtained first. Moreover, the collection, use, and processing of this data comply with relevant laws, regulations, and standards. Additionally, when embodiments of this application require access to sensitive personal information of users, separate permission or consent from the user is obtained through pop-ups or redirects to confirmation pages. Only after explicitly obtaining the user's separate permission or consent is the necessary user-related data for the proper functioning of these embodiments acquired.

[0033] Furthermore, the multi-granularity evaluation method for papers based on educational theory provided in this application relates to the field of artificial intelligence technology. This method can be applied to a terminal, a server, or software running on either a terminal or a server. In some embodiments, the terminal can be a base station (BS), base station control and management equipment, a smartphone, tablet, laptop, desktop computer, etc.; the server can be configured as an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms; the software can be an application implementing the multi-granularity evaluation method for papers based on educational theory, but is not limited to the above forms.

[0034] Alternatively, the multi-granularity evaluation method for papers based on educational theory provided in this application can also be used in numerous general-purpose or special-purpose computer system environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer computer devices, network PCs, minicomputers, mainframe computers, and distributed computing environments including any of the above systems or devices. This application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform specific tasks or implement specific abstract data types. This application can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In a distributed computing environment, program modules can reside in local and remote computer storage media, including storage devices.

[0035] For ease of understanding and explanation, the following text will use the application of the educational theory-based multi-granularity evaluation method for papers provided in the embodiments of this application on a terminal device as an example for detailed explanation. The implementation of the educational theory-based multi-granularity evaluation method for papers provided in the embodiments of this application on any of the above-mentioned subject matters can refer to the process of applying the educational theory-based multi-granularity evaluation method for papers on a terminal device as described below.

[0036] Please refer to Figure 1 , Figure 1 The flowchart illustrates the steps of the multi-granularity evaluation method for papers based on educational theory provided in some embodiments of this application. It should be understood that, although... Figure 1 The figure shows the execution order of some method steps, but based on different design needs of practical applications, the multi-granularity evaluation method for papers based on educational theory provided in this application embodiment can of course adopt a different execution order of method steps than that shown in the figure. That is, Figure 1 The order of the steps shown does not constitute a limitation on the execution logic order of the multi-granularity paper evaluation method based on educational theory provided in the embodiments of this application. Any other method based on... Figure 1 Reasonable variations in the order of the steps shown should be included within the protection scope of the multi-granularity evaluation method for papers based on educational theory provided in the embodiments of this application.

[0037] like Figure 1 As shown, in some embodiments, the paper multi-granularity evaluation method based on educational theory provided in this application may include, but is not limited to, steps S101 to S103.

[0038] Step S101: Obtain the data of the paper to be evaluated.

[0039] When performing multi-force evaluation of papers, the terminal device acquires the data of the paper to be evaluated.

[0040] In some embodiments, the thesis data may be the undergraduate thesis mentioned above.

[0041] For example, the terminal device can receive undergraduate graduation theses in PDF format uploaded by users through a pre-set human-computer interaction interface, and thus use the undergraduate graduation thesis as the thesis to be evaluated.

[0042] Step S102: Perform text extraction processing on the paper data to obtain the structured features of the paper data.

[0043] After the terminal device obtains the paper data to be evaluated, it further performs text extraction processing on the paper data to obtain the structured features of the paper data.

[0044] In some embodiments, the structured features can be a structured plain text format that can be processed by a large language model (LLM).

[0045] For example, a terminal device can perform text extraction processing on a PDF-formatted undergraduate thesis through the following three processing stages to obtain the structured features of the undergraduate thesis.

[0046] Phase 1: Use tools such as pdfplumber to extract text elements and layout metadata; Phase Two: Identify chapter boundaries, merge fragmented text, and insert chart placeholders; Phase 3: Generate a structured representation (i.e., structured features) containing metadata such as paper title, chapter tags, figure and table references.

[0047] Step S103: Based on the preset multi-dimensional hierarchical prompting guidance, the large language model performs multi-granularity evaluation processing on the structured features to obtain the multi-granularity evaluation results of the paper data; the multi-dimensional hierarchical prompting is constructed based on the multi-granularity evaluation dimension system of educational theory, and the multi-granularity evaluation results include multi-granularity dimension scores.

[0048] After the terminal device extracts and processes the text of the paper data to obtain its structured features, it inputs these structured features and multi-dimensional hierarchical prompts, which are pre-constructed based on a multi-granularity assessment dimension system of educational theory, into a large language model. Based on these multi-dimensional hierarchical prompts, the large language model is guided to perform multi-granularity assessment processing on the structured features, resulting in a multi-granularity assessment result of the paper data output by the large language model. This multi-granularity assessment result includes at least the multi-granularity dimension score generated by the large language model according to the multi-dimensional hierarchical prompts.

[0049] In this embodiment, when a terminal device performs a multi-dimensional evaluation of a paper, it acquires the paper data to be evaluated. Then, it further performs text extraction processing on the paper data to obtain the structured features of the paper data. Finally, it uses a multi-dimensional hierarchical prompt, which is pre-constructed based on a multi-granularity evaluation dimension system of educational theory, to guide a large language model to perform multi-granularity evaluation processing on the structured features, thereby obtaining the multi-granularity evaluation result of the paper data output by the large language model. The multi-granularity evaluation result includes at least the multi-granularity dimension score generated by the large language model according to the multi-dimensional hierarchical prompt processing of the structured features.

[0050] Compared to traditional methods that rely on a single overall score to guide Large Language Models (LLMs) in paper evaluation, this application's embodiment acquires the paper data to be evaluated, first extracts and processes the text to obtain its structured features, and then uses a multi-dimensional hierarchical prompting system pre-constructed based on educational theory to guide the LLM in performing multi-granularity evaluation of these structured features, resulting in a multi-granularity evaluation result for the paper data. This multi-granularity evaluation result includes multi-granularity dimension scores. Thus, by constructing a multi-dimensional evaluation framework and hierarchical prompting strategy that integrates educational theory, this application's embodiment can provide fine-grained, interpretable evaluations that align with expert assessments, thereby addressing the technical problems of insufficient granularity, lack of educational theory support, and single evaluation dimensions inherent in traditional evaluation methods.

[0051] Please refer to Figure 2 , Figure 2 The flowcharts of the multi-granularity evaluation method for papers based on educational theory provided in this application are illustrated in other embodiments.

[0052] like Figure 2 As shown, in some embodiments, the paper multi-granularity evaluation method based on educational theory provided in this application may further include steps S201 and S202 as shown below.

[0053] Step S201: Based on Vygotsky's sociocultural theory and Bloom's taxonomy of educational objectives, construct a multi-granularity assessment dimension system; the multi-granularity assessment dimension system includes at least two fine-grained assessment dimensions from structure, logic, originality, writing, proficiency and rigor.

[0054] It should be noted that Vygotsky's sociocultural theory emphasizes scaffolded learning (the gradual transition of students from external support to independent academic ability), focusing on learning potential and the developmental process, rather than just current performance, and emphasizing the influence of sociocultural factors on learning. Bloom's Taxonomy of Educational Objectives provides a structured hierarchy of cognitive abilities: memorization → comprehension → application → analysis → evaluation → creation, emphasizing the hierarchical nature of cognitive processes and is widely used in instructional design and assessment.

[0055] Furthermore, among the aforementioned fine-grained assessment dimensions, structure can be S-Structure: used to assess the overall organization and coherence of the paper's chapter arrangement, reflecting the student's ability to plan and manage academic content, and embodying the acquisition process of systematic academic norms; logic can be L-Logic: used to examine the internal consistency and reasoning flow between research objectives, methods, evidence, and conclusions, focusing on the student's ability to construct coherent arguments, corresponding to higher-level cognitive abilities such as "analysis and evaluation"; originality can be O-Originality: used to assess the intellectual independence and degree of innovation demonstrated in the paper, such as innovative perspectives, Problem definition or methodological creativity; Writing can be W-Writing: focusing on the clarity, precision, and standardization of academic expression, including sentence phrasing, terminology use, and adherence to subject-specific writing conventions, reflecting the process of academic socialization; Proficiency can be P-Proficiency: reflecting the student's mastery of subject knowledge and methodological abilities, including the accurate application of technical concepts, analytical skills, and domain-specific tools; Rigor can be R-Rigor: assessing the degree to which the paper adheres to academic standards and norms, including source reliability, citation accuracy, and academic integrity requirements.

[0056] Before performing multi-granularity evaluation of papers, the terminal device can first construct a multi-granularity evaluation dimension system based on Vygotsky's socio-cultural theory and Bloom's taxonomy of educational objectives, including at least two fine-grained evaluation dimensions among structure, logic, originality, writing, proficiency, and rigor.

[0057] In some embodiments, the terminal device may also, when its computing power allows, construct a multi-granularity evaluation dimension system based on Vygotsky's socio-cultural theory and Bloom's taxonomy of educational objectives, including at least two fine-grained evaluation dimensions among structure, logic, originality, writing, proficiency, and rigor, while performing the task of multi-granularity evaluation of papers.

[0058] For example, the terminal device constructs a SLOWPR framework with six fine-grained assessment dimensions based on Vygotsky's socio-cultural theory and Bloom's Taxonomy of Educational Objectives: structure, logic, originality, writing, proficiency, and rigor. This SLOWPR framework is a multi-granularity assessment dimension system.

[0059] Step S202: Construct a hierarchical two-stage prompt based on the multi-granularity evaluation dimension system; the hierarchical two-stage prompt includes a first-stage prompt and a second-stage prompt, the first-stage prompt is used to guide the large language model to independently evaluate the fine-grained evaluation dimensions in the multi-granularity evaluation dimension system, and the second-stage prompt is used to guide the large language model to integrate the fine-grained evaluation dimensions into an overall evaluation; the multi-dimensional hierarchical prompt includes the first-stage prompt and the second-stage prompt.

[0060] After constructing the multi-granularity evaluation dimension system, the terminal device further implements a two-stage hierarchical prompting system, comprising a first-stage prompt and a second-stage prompt. The first-stage prompt guides the large language model to independently evaluate each fine-grained evaluation dimension within the multi-granularity evaluation dimension system. The second-stage prompt guides the large language model to integrate these fine-grained evaluation dimensions into a holistic evaluation. In this way, the terminal device can use both the first-stage and second-stage prompts as multi-dimensional hierarchical prompts to guide the large language model.

[0061] For example, the first-stage suggestion can be the following high-precision independent isolation assessment suggestion built based on the SLOWPR framework described above: Please strictly follow the following rules to complete the assessment task, and do not deviate from the requirements: 1. Evaluation Object: The full text of the formal text to be evaluated (including academic papers, industry reports, official documents, formal plans, etc.) as entered below; 2. Core rule: The six specified fine-grained evaluation dimensions must be evaluated in a completely independent, single-dimensional isolation manner. The evaluation of each dimension is based solely on the definition of that dimension itself and must not be cross-referenced with other dimensions or output any overall judgment in advance. Each dimension should output a complete evaluation result separately. 3. Fixed Evaluation Dimensions and Standards (10 points total, scores must be accurate to 0.5 points, no fractional scores are allowed): ① Structure: Evaluate the completeness of the text framework, the clarity of paragraph hierarchy, the coherence between the beginning and end, and the logical priority of the core content arrangement; ② Logic: Evaluate the matching degree between the text's arguments and evidence, the completeness of the causal chain in the context, the self-consistency of the reasoning process, and the logical progression of the argumentation / narration; ③ Originality: Evaluate the originality of the text's viewpoints, the degree of content non-homogenization, and the proper citation of content. ④ Writing: Evaluate the text's fluency, word accuracy, sentence coherence, grammatical errors and spelling mistakes, and the appropriateness of expression to the context; ⑤ Proficiency: Evaluate the text's use of professional terminology in the relevant field, the depth of industry knowledge, the professional suitability of the content, and the fit with common expression rules in the field; ⑥ Rigor: Evaluate the traceability of the text's data sources, the authenticity and validity of the arguments, the objectivity of the expression, the prudence of the conclusions, and the compliance of avoiding exaggerated / absolute statements; 4. Mandatory output format: Each dimension should be a separate paragraph, strictly following the fixed structure of "Dimension Name + Final Score + Core Evaluation Basis + Existing Core Issues". Dimensions must not be merged, any dimension must not be omitted, and the output order must not be changed.

[0062] In addition, the first-stage hints can also provide the following lightweight independent evaluation hints based on the SLOWPR framework described above: Please complete the evaluation of the texts to be evaluated below (including social media copy, work summaries, speeches, and general daily documents) strictly according to the following requirements: 1. Core requirements: Each of the six fixed dimensions—structure, logic, originality, writing, fluency, and rigor—will be evaluated independently. Results for each dimension will be given separately, without any impact or overlap between them. It is forbidden to output the overall conclusion in advance. 2. Evaluation Rules: The evaluation of each dimension must include 3 fixed items: ① Dimension name ② Evaluation level (only four levels can be selected: Excellent / Good / Satisfactory / Needs Improvement) ③ One sentence as the core reason for the judgment; 3. Mandatory requirements: All 6 dimensions must be fully covered, with no omissions or merging of dimensions. Each dimension must be output on a separate line.

[0063] Furthermore, the second-stage suggestions can be the following deep comprehensive evaluation suggestions built upon the SLOWPR framework described above: Based entirely on the independent evaluation results of the six fine-grained dimensions already presented, please complete the overall comprehensive evaluation of the text to be evaluated, strictly adhering to the following rules: 1. Core principle: All comprehensive evaluation conclusions must be 100% anchored to the independent evaluation results of the six dimensions mentioned above. No new content not mentioned in the single-dimensional evaluation is allowed, and no judgments should be made out of thin air without considering the results of the single dimensions. 2. Fixed weighting rules: The default weighting is 15% for structure, 20% for logic, 20% for originality, 15% for writing, 15% for proficiency, and 15% for rigor. If the weights need to be adjusted according to the text type, the reasons for the adjustment must be clearly stated. 3. Mandatory output content and order must fully cover the following four items: ① Overall comprehensive rating: Calculate the weighted average score based on the individual scores of the six dimensions, and give the overall rating (out of 10: 9-10: S-level Excellent; 7.5-8.9: A-level Good; 6-7.4: B-level Pass; below 6: C-level Needs Improvement); ② Comprehensive strengths and weaknesses judgment: Based on the evaluation results of the six dimensions, systematically summarize the core competitive advantages and core weaknesses of the text. Weaknesses must correspond to the evaluation conclusions of specific dimensions; ③ Overall scenario adaptability conclusion: Clearly determine whether the text is suitable for its target use scenario, give a clear conclusion of suitable / partially suitable / unsuitable, and explain the basis for the judgment based on the dimension evaluation results; ④ Priority improvement plan: Based on the severity of the problems in each dimension, give overall improvement suggestions with high / medium / low priority. High priority improvement items must correspond to the core weakness dimension.

[0064] Furthermore, the second-stage suggestions can also be used for the following lightweight comprehensive evaluation suggestions built on the SLOWPR framework described above: Please strictly base your evaluation on the independent ratings of the six dimensions (structure, logic, originality, writing, fluency, and rigor) already presented above, and complete the overall comprehensive evaluation of the text to be evaluated, adhering to the following requirements: 1. Core rule: All comprehensive conclusions must be based entirely on the independent evaluation results of the six dimensions, and no irrelevant content may be added or judgments may be made arbitrarily without considering the results of a single dimension; 2. Mandatory output content must fully cover the following three items: ① Overall comprehensive level: Combining the independent levels of the six dimensions, give the overall final level of the text (only Excellent / Good / Satisfactory / Needs improvement) and explain the core basis for the level determination; ② Overall highlights summary: Extract the core highlights of the text that stand out in each dimension, no more than 3; ③ Core improvement suggestions: Combining the areas for improvement in each dimension, give the 2-3 most core directions for overall optimization; 3. The language is concise and straightforward, focusing on overall comprehensive judgment without redundant descriptions.

[0065] Please refer to Figure 3 , Figure 3 for Figure 1 A detailed flowchart of step S103.

[0066] like Figure 3 As shown, in some embodiments, the step S103 above, "based on a preset multi-dimensional hierarchical prompting guidance large language model, performs multi-granularity evaluation processing on the structured features to obtain the multi-granularity evaluation result of the paper data", may include step S301 as shown below.

[0067] Step S301: Based on the prompts and guidance of the first stage, the structured features are independently evaluated from the fine-grained evaluation dimensions in the multi-granularity evaluation dimension system to obtain multiple dimension scores of the paper data; the multi-granularity dimension scores include the multiple dimension scores.

[0068] After receiving the first-stage prompt, the terminal device can input this prompt along with the structured features of the paper data into the large language model. Based on this first-stage prompt, the large language model can then independently evaluate the structured features from at least two of the fine-grained evaluation dimensions (structure, logic, originality, writing, fluency, and rigor) within the aforementioned multi-granularity evaluation dimension system, resulting in multi-dimensional scores for the paper data generated by the large language model. In this way, the terminal device can use these multi-dimensional scores as part of the multi-granularity evaluation results obtained from multi-granularity evaluation of the paper data based on educational theory.

[0069] For example, the large language model can independently evaluate the six SLOWPR dimensions according to the algorithm formula shown below, generating a score for each dimension and a corresponding explanation (i.e., multiple dimension scores and their explanations), reducing cross-dimensional interference and achieving more accurate domain knowledge activation.

[0070] .

[0071] in: The input is the graduation thesis text. For the evaluation of the d-th dimension, To generate a large language model for evaluation and justification, For the rating result of the d-th dimension, The reason for the rating of the d-th dimension.

[0072] Please refer to Figure 4 , Figure 4 for Figure 1 A schematic diagram of another detailed step in step S103.

[0073] like Figure 4As shown, in some embodiments, the step S103 above, "based on a preset multi-dimensional hierarchical prompting guidance large language model, performs multi-granularity evaluation processing on the structured features to obtain the multi-granularity evaluation result of the paper data", may also include the following step S401.

[0074] Step S401: Based on the prompts and guidance of the second stage, the large language model integrates the various fine-grained evaluation dimensions into an overall evaluation; the integration of the various fine-grained evaluation dimensions into an overall evaluation includes weighted summation of the scores of the multiple dimensions to obtain the overall score of the paper data, and generating feedback and improvement suggestions; the multi-granularity evaluation results also include the overall score, the feedback and the improvement suggestions.

[0075] After the terminal device independently evaluates the structured features based on the first-stage prompting and guidance of the large language model, obtaining multi-dimensional scores for the paper data, it further inputs the second-stage prompting and guidance into the large language model. Based on this second-stage prompting and guidance, the large language model integrates the various fine-grained evaluation dimensions into a holistic evaluation; that is, it performs a weighted summation of the scores across multiple dimensions to obtain the overall score for the paper data, and generates feedback and improvement suggestions. In this way, the terminal device can further incorporate this overall score, feedback, and improvement suggestions as part of the multi-granularity evaluation results obtained from the paper data's multi-granularity evaluation based on educational theory.

[0076] For example, the large language model can integrate six fine-grained assessments into an overall assessment according to the algorithm formula shown below, wherein the overall score is calculated by weighted summation of the scores of each dimension, the calculation weights can be configured to adapt to different disciplines or assessment standards, and feedback and improvement suggestions are generated at the same time.

[0077] ; .

[0078] in: To evaluate the set of dimensions, The score for the d-th dimension (0-10 points). The final score is 0-10 points. Let d be the weight of the d-th dimension, satisfying , To generate a large language model for feedback. These are the final improvement suggestions.

[0079] Please refer to Figure 5 , Figure 5 The following is a flowchart illustrating the steps of the multi-granularity evaluation method for papers based on educational theory provided in some embodiments of this application.

[0080] like Figure 5 As shown, in some embodiments, the multi-granularity evaluation method for papers based on educational theory provided in this application may further include the following step S501.

[0081] Step S501: Perform context learning enhancement processing on the large language model based on preset few-shot prompts; the few-shot prompts include paper sample data, and the model evaluation results corresponding to the paper sample data include at least multi-granularity dimension scores and overall scores.

[0082] The terminal device can also pre-build few-shot prompts, which include sample data from academic papers. The model evaluation results corresponding to these sample data include at least multi-granularity dimension scores and an overall score. Then, the terminal device performs context learning enhancement processing on the large language model based on these few-shot prompts.

[0083] In some embodiments, to further improve the alignment between the large language model's multi-granularity evaluation of paper data based on educational theory and expert evaluation, the terminal device can integrate contextual learning technology based on few-shot prompts into the prompts for the large language model. That is, by providing example papers (i.e., paper sample data) that conform to the multi-granularity output structure through few-shot prompts, including six SLOWPR dimension scores and an overall score, the model can be helped to internalize the evaluation format through patterned guidance.

[0084] Please refer to Figure 6 , Figure 6 The following is a flowchart illustrating the steps of the multi-granularity evaluation method for papers based on educational theory provided in some embodiments of this application.

[0085] like Figure 6 As shown, in some embodiments, the multi-granularity evaluation method for papers based on educational theory provided in this application may further include the following step S601.

[0086] Step S601: Perform context learning enhancement processing on the large language model based on preset role-playing prompts; the role-playing prompts are used to guide the large language model to act as a paper reviewer and to reason and evaluate the paper data.

[0087] The terminal device can also pre-build role-playing prompts to guide the large language model in acting as a paper reviewer and evaluating the paper data. Then, the terminal device can perform contextual learning enhancement processing on the large language model based on these role-playing prompts.

[0088] In some embodiments, in order to improve the alignment between the large language model's multi-granularity evaluation of thesis data based on educational theory and expert evaluation, the terminal device can also integrate contextual learning technology based on role-playing prompts into the prompts for the large language model. That is, role-playing prompts instruct the large language model to act as an experienced member of the thesis review committee (i.e., thesis review expert), using formal academic tone, subject-appropriate vocabulary, and expert evaluation reasoning methods that meet institutional expectations.

[0089] In some embodiments, to improve the alignment between the large language model's multi-granularity evaluation of thesis data based on educational theory and expert evaluation, the terminal device can simultaneously integrate the above two contextual learning techniques in the prompts: providing example papers that conform to the multi-granularity output structure through few-shot prompting, including six SLOWPR dimension scores and an overall score, thereby helping the model internalize the evaluation format through patterned guidance; and instructing the large language model to act as an experienced member of the thesis review committee (i.e., thesis review expert) through role-playing prompting, using formal academic tone, subject-appropriate vocabulary, and expert evaluation reasoning methods that meet institutional expectations.

[0090] Next, a complete embodiment of the multi-granularity evaluation method for papers based on educational theory provided in this application is presented.

[0091] Please refer to Figure 7 , Figure 7 This is a schematic diagram of the overall system framework of the multi-granularity evaluation method for papers based on educational theory provided in an embodiment of this application.

[0092] like Figure 7 As shown, the terminal device can perform multi-granularity evaluation of undergraduate graduation theses based on educational theory using an undergraduate thesis multi-granularity evaluation system. The specific process of the undergraduate thesis multi-granularity evaluation system can include steps 1 to 7 as shown below: Step 1. Data Preprocessing: Convert the PDF graduation thesis into a structured plain text format that can be processed by LLM. This mainly includes three stages: using tools such as pdfplumber to extract text elements and layout metadata; identifying chapter boundaries, merging fragmented text, and inserting figure and table placeholders; and generating a structured representation that includes metadata such as thesis title, chapter labels, and figure and table references.

[0093] Step 2. Construct a multi-granularity assessment dimension system based on educational theories: Vygotsky's sociocultural theory emphasizes scaffolded learning (students gradually transition from external support to independent academic ability), focusing on learning potential and developmental processes, rather than just current performance, and emphasizing the influence of sociocultural factors on learning. Bloom's Taxonomy of Educational Objectives provides a structured hierarchy of cognitive abilities: memory → comprehension → application → analysis → evaluation → creation, emphasizing the hierarchical nature of cognitive processes, and is widely used in instructional design and assessment.

[0094] Based on Vygotsky's sociocultural theory and Bloom's taxonomy of educational objectives, a SLOWPR framework comprising six fine-grained assessment dimensions is constructed: S-Structure: This assesses the overall organization and coherence of the paper's chapters, reflecting the student's ability to plan and manage academic content, and demonstrating the acquisition of systematic academic norms. L-Logic: Examines the internal consistency and reasoning process between research objectives, methods, evidence, and conclusions, focusing on students' ability to construct coherent arguments, corresponding to higher-level cognitive abilities such as "analysis and evaluation." O-Originality: Assessing the intellectual independence and degree of innovation demonstrated in the paper, such as innovative perspective, problem definition, or methodological creativity. W-Writing focuses on the clarity, precision, and standardization of academic expression, including sentence phrasing, terminology usage, and adherence to disciplinary writing conventions, reflecting the process of academic socialization. P-Proficiency: Reflects a student's mastery of subject knowledge and methodological skills, including the accurate application of technical concepts, analytical techniques, and domain-specific tools. R-Rigor (rigor): assesses the degree to which a paper adheres to academic standards and norms, including source reliability, citation accuracy, and academic integrity requirements.

[0095] Step 3. Design a hierarchical two-stage prompting strategy: Since multi-granularity evaluation needs to integrate multiple evaluation criteria and task-specific instructions into a single input, a hierarchical two-stage prompting mechanism is proposed: Phase 1: The model independently evaluates each of the six SLOWPR dimensions, generating a score for each dimension and a corresponding explanation, reducing cross-dimensional interference and achieving more accurate domain knowledge activation.

[0096] .

[0097] in: The input is the graduation thesis text. For the evaluation of the d-th dimension, To generate a large language model for evaluation and justification, For the rating result of the d-th dimension, The reason for the rating of the d-th dimension.

[0098] The second stage: The model integrates the six fine-grained assessments into an overall assessment. The overall score is calculated by weighted summation of the scores from each dimension. The weights can be configured to adapt to different disciplines or assessment standards, while generating feedback and improvement suggestions.

[0099] ; .

[0100] in: To evaluate the set of dimensions, The score for the d-th dimension (0-10 points). The final score is 0-10 points. Let d be the weight of the d-th dimension, satisfying , To generate a large language model for feedback. These are the final improvement suggestions.

[0101] Step 4. Incorporating Contextual Learning Enhancement Techniques: To further improve alignment with expert evaluation, two contextual learning techniques are integrated into the prompts: Few-shot Prompting: Provides sample papers that conform to a multi-granularity output structure, including six SLOWPR dimension scores and an overall score, helping the model internalize the evaluation format through patterned guidance.

[0102] Role-playing prompts: Instruct the model to act as an experienced member of the thesis review committee, using formal academic terminology, subject-appropriate vocabulary, and expert evaluation reasoning methods that meet institutional expectations.

[0103] Step 5. Generate evaluation report: Combine the preprocessed paper text, SLOWPR-based hierarchical prompts, few-sample examples, and role-playing instructions, and submit them to the large language model for evaluation. Output a standardized evaluation report, including: scores and supporting reasons for each fine-grained dimension, overall score, and improvement suggestions.

[0104] In this embodiment, compared to the traditional assessment scheme that uses simple overall scoring prompts for undergraduate thesis evaluation, this embodiment offers a more granular multi-dimensional assessment based on educational theory. Traditional assessment schemes only return a total score, failing to differentiate the thesis's performance across different competency dimensions. This embodiment, through the SLOWPR six-dimensional framework, enables independent assessment of six dimensions: structure, logic, originality, writing, proficiency, and rigor. It not only provides an overall score but also individual scores and explanations for each dimension, significantly improving the interpretability of the assessment results. Furthermore, this embodiment offers higher assessment quality. The generic prompts of traditional assessment schemes are insufficient to effectively activate the domain knowledge within the Large Language Model (LLM). This embodiment aligns the assessment with established educational theories by integrating Vygotsky's socio-cultural theory and Bloom's Taxonomy of Educational Objectives. The hierarchical two-stage prompting strategy reduces attentional distractions within the large model, while the use of few-sample prompts and role-playing prompts enhances alignment with expert assessment practices. Experiments show that this embodiment significantly outperforms traditional assessment schemes in evaluation metrics such as mean absolute error (MAE), mean squared error (MSE), and Pearson correlation coefficient (PCC). Finally, the feedback output by this embodiment through the large language model is more valuable for instruction. Traditional assessment schemes only provide scores and cannot offer constructive suggestions for improvement. This embodiment not only provides fine-grained scoring but also generates targeted feedback, helping students identify specific shortcomings in areas such as argument organization, methodological clarity, and academic writing, thus providing a basis for teaching improvement.

[0105] Based on the same technical concept as the above-mentioned multi-granularity evaluation method for papers based on educational theory, this application also provides a multi-granularity evaluation device for papers based on educational theory that can implement the above-mentioned multi-granularity evaluation method for papers based on educational theory.

[0106] Please see Figure 8 The multi-granularity evaluation device for papers based on educational theory provided in this application includes: The acquisition module is used to acquire the data of the papers to be evaluated; The preprocessing module is used to perform text extraction processing on the paper data to obtain the structured features of the paper data; The multi-granularity evaluation module is used to perform multi-granularity evaluation processing on the structured features based on a preset multi-dimensional hierarchical prompting guidance large language model to obtain the multi-granularity evaluation results of the paper data; the multi-dimensional hierarchical prompting is constructed based on the multi-granularity evaluation dimension system of educational theory, and the multi-granularity evaluation results include multi-granularity dimension scores.

[0107] In some embodiments, the paper multi-granularity evaluation device based on educational theory provided in this application further includes: The prompting module is used to construct a multi-granularity assessment dimension system based on Vygotsky's socio-cultural theory and Bloom's Taxonomy of Educational Objectives. This system includes at least two fine-grained assessment dimensions from structure, logic, originality, writing, proficiency, and rigor. Furthermore, it constructs a hierarchical two-stage prompting system based on this system. The hierarchical two-stage prompting includes a first-stage prompt and a second-stage prompt. The first-stage prompt guides the large language model to independently assess each fine-grained assessment dimension within the multi-granularity assessment dimension system, while the second-stage prompt guides the large language model to integrate all the fine-grained assessment dimensions into a holistic assessment. The multi-dimensional hierarchical prompting system includes both the first-stage and second-stage prompts.

[0108] In some embodiments, the multi-granularity evaluation module is further configured to independently evaluate the structured features from the fine-grained evaluation dimensions in the multi-granularity evaluation dimension system based on the first-stage prompting guidance of the large language model, to obtain multiple dimension scores for the paper data; the multi-granularity dimension scores include the multiple dimension scores; and, based on the second-stage prompting guidance of the large language model, integrate each of the fine-grained evaluation dimensions into an overall evaluation; the integration of each of the fine-grained evaluation dimensions into an overall evaluation includes weighted summation of the multiple dimension scores to obtain an overall score for the paper data, and generating feedback and improvement suggestions; the multi-granularity evaluation result also includes the overall score, the feedback, and the improvement suggestions.

[0109] In some embodiments, the paper multi-granularity evaluation device based on educational theory provided in this application further includes: The context learning enhancement module is used to perform context learning enhancement processing on the large language model based on preset few-shot prompts; the few-shot prompts include paper sample data, and the model evaluation results corresponding to the paper sample data include at least multi-granularity dimension scores and overall scores.

[0110] In some embodiments, the context learning enhancement module is further configured to perform context learning enhancement processing on the large language model based on preset role-playing prompts; the role-playing prompts are used to guide the large language model to act as a paper reviewer to reason and evaluate the paper data.

[0111] It should be noted that the specific implementation of the paper multi-granularity evaluation device based on educational theory provided in this application is basically the same as the specific implementation of the paper multi-granularity evaluation method based on educational theory described above, and will not be repeated here.

[0112] This application also provides a computer device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the aforementioned multi-granularity evaluation method for papers based on educational theory. This computer device can be any smart terminal, including tablet computers, in-vehicle computers, etc.

[0113] Please see Figure 9 , Figure 9 This illustration shows the hardware structure of a computer device in one embodiment, the computer device including: The processor 901 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this application. The memory 902 can be implemented as a read-only memory (ROM), static storage device, dynamic storage device, or random access memory (RAM). The memory 902 can store the operating system and other application programs. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 902 and is called and executed by the processor 901 to implement the multi-granularity paper evaluation method based on educational theory of the embodiments of this application. The input / output interface 903 is used to implement information input and output; The communication interface 904 is used to enable communication and interaction between this device and other devices. Communication can be achieved through wired means (such as USB, Ethernet cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.). Bus 905 transmits information between various components of the device (e.g., processor 901, memory 902, input / output interface 903, and communication interface 904); The processor 901, memory 902, input / output interface 903, and communication interface 904 are connected to each other within the device via bus 905.

[0114] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described multi-granularity evaluation method for papers based on educational theory.

[0115] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. Furthermore, memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory may optionally include memory remotely located relative to the processor, and these remote memories can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0116] This application also provides a computer program product that stores a computer program that, when executed by a processor, implements the above-described multi-granularity evaluation method for papers based on educational theory.

[0117] The paper multi-granularity evaluation method, device, computer equipment, computer-readable storage medium, and computer program product based on educational theory provided in this application embodiment utilize the dual beam splitting phenomenon in the spatial and frequency domains based on sparse arrays to achieve super-resolution beam alignment with limited spectrum resources, greatly improving the accuracy of beam training. Furthermore, by utilizing central subarray activation, this application embodiment decouples angle and distance estimation, significantly reducing the pilot overhead required for beam training, thereby achieving a super-resolution, low-overhead paper multi-granularity evaluation method based on educational theory.

[0118] The embodiments described in this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided by the embodiments of this application. As those skilled in the art will know, with the evolution of technology and the emergence of new application scenarios, the technical solutions provided by the embodiments of this application are also applicable to similar technical problems.

[0119] Those skilled in the art will understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of this application, and may include more or fewer steps than shown, or combine certain steps, or different steps.

[0120] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0121] Those skilled in the art will understand that all or some of the steps in the methods disclosed above, as well as the functional modules / units in the systems and devices, can be implemented as software, firmware, hardware, or suitable combinations thereof.

[0122] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0123] It should be understood that in this application, "at least one (item)" means one or more, and "more than" means two or more. "And / or" is used to describe the relationship between related objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one (item) of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0124] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of the units described above is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0125] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0126] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0127] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes multiple instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing programs, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0128] The preferred embodiments of the present application have been described above with reference to the accompanying drawings, but this does not limit the scope of the claims of the present application. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and substance of the embodiments of the present application shall be within the scope of the claims of the present application.

Claims

1. A multi-granularity evaluation method for papers based on educational theory, characterized in that, The method includes: Obtain the data for the papers to be evaluated; The paper data is subjected to text extraction processing to obtain the structured features of the paper data; Based on the preset multidimensional hierarchical prompts, the large language model performs multi-granularity evaluation processing on the structured features to obtain the multi-granularity evaluation results of the paper data; the multidimensional hierarchical prompts are constructed based on the multi-granularity evaluation dimension system of educational theory, and the multi-granularity evaluation results include multi-granularity dimension scores.

2. The method according to claim 1, characterized in that, The method further includes: Based on Vygotsky's sociocultural theory and Bloom's taxonomy of educational objectives, a multi-granularity assessment dimension system is constructed; the multi-granularity assessment dimension system includes at least two fine-grained assessment dimensions from structure, logic, originality, writing, proficiency, and rigor. Based on the multi-granularity evaluation dimension system, a hierarchical two-stage prompt is constructed; the hierarchical two-stage prompt includes a first-stage prompt and a second-stage prompt. The first-stage prompt is used to guide the large language model to independently evaluate the fine-grained evaluation dimensions in the multi-granularity evaluation dimension system, and the second-stage prompt is used to guide the large language model to integrate the fine-grained evaluation dimensions into an overall evaluation; the multi-dimensional hierarchical prompt includes the first-stage prompt and the second-stage prompt.

3. The method according to claim 2, characterized in that, The pre-defined multi-dimensional hierarchical prompting guides the large language model to perform multi-granularity evaluation processing on the structured features, obtaining multi-granularity evaluation results for the paper data, including: Based on the prompts and guidance of the first stage, the large language model independently evaluates the structured features from the fine-grained evaluation dimensions in the multi-granularity evaluation dimension system to obtain multiple dimension scores for the paper data; the multi-granularity dimension scores include the multiple dimension scores.

4. The method according to claim 3, characterized in that, The method of using a pre-defined multi-dimensional hierarchical prompting guidance large language model to perform multi-granularity evaluation processing on the structured features to obtain the multi-granularity evaluation results of the paper data also includes: Based on the second-stage prompts, the large language model integrates the various fine-grained evaluation dimensions into an overall evaluation; integrating the various fine-grained evaluation dimensions into an overall evaluation includes weighted summation of the scores of the multiple dimensions to obtain the overall score of the paper data, and generating feedback and improvement suggestions; the multi-granularity evaluation results also include the overall score, the feedback, and the improvement suggestions.

5. The method according to any one of claims 1 to 4, characterized in that, The method further includes: The large language model is enhanced by context learning based on preset few-shot prompts; the few-shot prompts include paper sample data, and the model evaluation results corresponding to the paper sample data include at least multi-granularity dimension scores and overall scores.

6. The method according to claim 5, characterized in that, The method further includes: The large language model is enhanced by context learning based on preset role-playing prompts; the role-playing prompts are used to guide the large language model to act as a paper reviewer and to reason and evaluate the paper data.

7. A multi-granularity evaluation device for papers based on educational theory, characterized in that, The device includes: The acquisition module is used to acquire the data of the papers to be evaluated; The preprocessing module is used to perform text extraction processing on the paper data to obtain the structured features of the paper data; The multi-granularity evaluation module is used to perform multi-granularity evaluation processing on the structured features based on a preset multi-dimensional hierarchical prompting guidance large language model to obtain the multi-granularity evaluation results of the paper data; the multi-dimensional hierarchical prompting is constructed based on the multi-granularity evaluation dimension system of educational theory, and the multi-granularity evaluation results include multi-granularity dimension scores.

8. A computer device, characterized in that, The computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the multi-granularity evaluation method for papers based on educational theory as described in any one of claims 1 to 6.

9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the multi-granularity evaluation method for papers based on educational theory as described in any one of claims 1 to 6.

10. A computer program product, said computer program product storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the multi-granularity evaluation method for papers based on educational theory as described in any one of claims 1 to 6.

Images