Structured scientific paper auxiliary creation method and system based on artificial intelligence

Abstract

The invention discloses a structured scientific paper auxiliary creation method and system based on artificial intelligence, and belongs to the field of artificial intelligence, and the method comprises the steps: receiving paper core elements inputted by a user; decomposing the to-be-written paper into a plurality of logic parts according to a preset rule or user appointment, and defining writing targets of each part; aiming at the at least one logic part, based on the core element and the writing target and according to a preset specific prompting strategy template aiming at different logic parts, generating a structured guiding prompt containing a specific instruction and a constraint; inputting the guiding prompt into an artificial intelligence language model to obtain a preliminary draft; providing an interaction mechanism, receiving structured feedback of the user on the preliminary draft, and generating a revision prompt based on the feedback and the guidance prompt; inputting the revision prompt into the artificial intelligence language model to obtain a revised draft; executing cross-chapter content association and consistency check; and assembling the draft of each logic part to form a complete paper.

Description

Technical Field

[0001] The present invention belongs to the field of artificial intelligence, and in particular relates to an artificial intelligence-based structured scientific paper auxiliary creation method and system. Background Art

[0002] Scientific papers, especially those that need to be published in top journals, have extremely high requirements for the scientificity, innovation, logic, structural standardization and language expression of their content. The traditional paper writing process usually consumes a lot of time and energy of scientific researchers, requiring authors to have not only deep professional knowledge, but also rigorous logical thinking and excellent academic writing skills. In recent years, with the rapid development of artificial intelligence technologies such as large language models (LLMs), many AI-assisted writing tools have emerged. These tools have shown potential in certain general writing tasks. For example, large language models such as OpenAI's ChatGPT, Google's Bard, and Anthropic's Claude have been widely used in tasks such as content creation and text summarization. However, when existing general AI writing tools are directly applied to the creation of rigorous and high-standard scientific papers, there are still many significant technical problems:

[0003] Existing AI-assisted writing tools have numerous shortcomings in the creation of scientific papers. Regarding content quality, the generated text often lacks depth and precision, is prone to factual errors and logical fallacies, and fails to meet the rigorous requirements of scientific research. Regarding structuring capabilities, they struggle to generate full texts that adhere to the standard structure of scientific papers, lacking natural connections and rigorous logical progression between sections, making it difficult to form complete and coherent manuscripts. They also lack domain specificity, with limited understanding of the professional terminology, research paradigms, writing conventions, and contextual differences within specific disciplines, resulting in unprofessional and non-conforming generated content. Regarding innovative expression, the generated text is prone to "template-like" behavior, making it difficult to accurately and powerfully express the author's unique academic insights and research contributions. Human-computer interaction is inefficient, requiring users to repeatedly debug input prompts, resulting in only fragmented, barely usable text fragments. Overall creative efficiency improvements are limited, and significant effort is required for screening, modification, and integration. Regarding meeting the requirements of top journals, existing general-purpose AI tools lack the ability to provide depth of argumentation and innovative expression, making them inadequate for assisting in the creation of top-tier journal papers. Consequently, the acceptance rate of AI-assisted submissions in top journals is significantly lower than that of fully manually written papers. Summary of the Invention

[0004] To solve the above technical problems, the present invention provides an artificial intelligence-based structured scientific paper auxiliary creation method, comprising:

[0005] Receive user input on the core elements of the paper;

[0006] Based on preset rules or user-specified requirements, the paper to be written is broken down into multiple logical parts, and writing objectives and core content points are defined for the multiple logical parts;

[0007] Generate structured guiding prompts containing specific instructions and constraints based on the core elements of the paper, the writing objectives, the core content points and a preset specific prompt strategy template in the logical part;

[0008] constructing an artificial intelligence language model, inputting the structured guiding prompts into the artificial intelligence language model, and obtaining a preliminary draft of the logic portion;

[0009] Processing the preliminary draft of the logic section based on an interactive mechanism to obtain structured feedback;

[0010] generating revision prompts based on the structured feedback and the structured guiding prompts;

[0011] Inputting the revision prompt into the artificial intelligence language model to obtain a revised draft;

[0012] Producing a complete first draft of the paper based on the preliminary draft and the revised draft.

[0013] Preferably, the core elements of the paper include but are not limited to: research topic, research objectives, key assumptions or questions, overview of main research methods, core data and summary of results.

[0014] Preferably, the preset specific prompt strategy template includes specific writing strategies for different logical parts of the paper, and the template for each logical part contains different instruction combinations and priorities.

[0015] Preferably, after obtaining the preliminary draft of the logical part, it also includes: after generating drafts of multiple logical parts, performing the steps of cross-chapter content association and consistency check, the check includes at least one of the following: consistency in terminology use, consistency in introductory questions and conclusion responses, support for methods and results, and consistency between results and discussion explanations.

[0016] Preferably, the consistency check is achieved by extracting key information of each logical part, constructing a logical relationship diagram, and analyzing based on preset rules or by inputting specific check prompts using an artificial intelligence language model. If the check finds inconsistencies, revision prompts are regenerated based on the inconsistent content, and the revision prompts are input into the artificial intelligence language model to obtain a revised draft.

[0017] Preferably, the process of regenerating the revision prompt includes: adjusting specific instructions and constraints in the structured guiding prompt according to the inconsistent content, and regenerating the revision prompt.

[0018] Preferably, the step of adjusting the specific instructions and constraints in the structured guiding prompts includes:

[0019] Modify one or more of the role settings, contextual information, content requirements, style and tone instructions, structural instructions, key information injection, and negative constraints in the structured guiding prompts to resolve the inconsistent content.

[0020] On the other hand, the present invention also provides an artificial intelligence-based structured scientific paper auxiliary writing system, comprising:

[0021] A receiving unit, used to receive the core elements of the paper input by the user;

[0022] A decomposition planning unit is used to decompose the paper to be written into multiple logical parts based on preset rules or user-specified requirements, and define writing goals and core content points for the multiple logical parts;

[0023] A prompt generation engine is configured with a library of specific prompt strategy templates for different logical parts of the paper, and is used to generate structured guiding prompts containing specific instructions and constraints based on the core elements of the paper, the writing goals, the core content points and preset specific prompt strategy templates of the logical parts, and to generate structured guiding prompts containing at least three specified instructions;

[0024] An AI model interface unit, configured to construct an artificial intelligence language model, input the structured guiding prompts into the artificial intelligence language model to obtain a preliminary draft of the logic portion, receive revision prompts and input them into the artificial intelligence language model to obtain a revised draft;

[0025] an interactive processing unit, configured to process the preliminary draft of the logic portion based on an interactive mechanism, obtain structured feedback, and generate revision prompts based on the structured feedback and the structured guiding prompts;

[0026] The assembly unit generates a complete first draft of the paper based on the preliminary draft and the revised draft.

[0027] On the other hand, the present invention further provides an electronic device, comprising a memory, a processor, and a computing program stored in the memory and executable on the processor, wherein the method is implemented when the processor executes the computing program.

[0028] On the other hand, the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and the computer program implements the method when executed by a processor.

[0029] Compared with the prior art, the present invention has the following advantages and technical effects:

[0030] The present invention, through the use of a structured authoring process, guiding prompts generated based on specific strategy templates, interactive iterative refinement, and cross-chapter consistency checking, brings at least the following beneficial effects:

[0031] Significantly improves creative efficiency: Structured decomposition and targeted guidance significantly reduce the time and effort required to conceive and write a first draft from scratch, allowing researchers to focus more on refining core ideas, interpreting experimental data, and reviewing content. Experiments have shown that using this method for scientific paper creation can save approximately 60-70% of writing time compared to traditional methods, while maintaining or improving content quality.

[0032] Effectively improve content quality and standardization: Prompt strategy templates designed based on expert experience and writing patterns guide AI to generate texts that are more in line with academic standards, more rigorous in logic, more in-depth in content, and more professional in expression. The interactive refinement process allows users to precisely control and improve content quality, bringing it closer to the requirements of high-level journals. Comparative experiments show that the paper drafts generated by the system of this invention are approximately 40% more professional, structurally rigorous, and logically consistent than content directly generated by general AI tools (based on objective scores from expert reviewers).

[0033] Ensure structural integrity and logical coherence: Strict structured decomposition and planning ensure a clear and complete framework for the paper. Segment generation and cross-section consistency checking ensure smooth transitions and logical coherence between sections. Using the consistency checking feature of this invention can effectively reduce structural and logical errors by approximately 85%, significantly improving the overall quality of the paper.

[0034] Enhanced user control and personalized expression: By inputting key elements and providing structured feedback, users can effectively guide the AI's output direction and content focus. This avoids the "black box" feel and template-based nature of general AI writing, helping to better incorporate the author's unique insights and research contributions. User surveys show that researchers using this system believe that the final manuscript more accurately reflects their research intentions and academic perspectives, with satisfaction increasing by approximately 50%.

[0035] Lowering the writing barrier: This system provides powerful tools for researchers who struggle with academic writing or are non-native English speakers, helping them overcome language and writing barriers and more effectively organize and publish their valuable research findings. Data shows that after using this system, the language quality scores of non-native English-speaking researchers' papers increased by approximately 65%, approaching the level of native English-speaking researchers.

[0036] Solidify and apply advanced writing methodologies: Proven, high-level essay writing strategies and techniques, such as "specific prompt strategy templates," are solidified into methods and systems through technical means such as prompt strategy templates, making them executable, replicable, and scalable, with strong practical value. This system integrates writing patterns and strategies from hundreds of top journal articles, making this expertise more widely applicable. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The accompanying drawings, which constitute part of this application, are intended to provide a further understanding of this application. The exemplary embodiments and descriptions of this application are intended to explain this application and do not constitute an improper limitation on this application. In the accompanying drawings:

[0038] Figure 1 Flowchart of a structured scientific paper assisted creation method according to an embodiment of the present invention;

[0039] Figure 2 Schematic diagram of a process for generating a guiding prompt according to an embodiment of the present invention;

[0040] Figure 3 Schematic diagram of an interactive iterative refining process according to an embodiment of the present invention;

[0041] Figure 4 This is a functional module block diagram of a structured scientific paper auxiliary creation system according to an embodiment of the present invention;

[0042] Figure 5 This is a diagram illustrating the working principle of cross-chapter consistency checking according to an embodiment of the present invention. DETAILED DESCRIPTION

[0043] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in this application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0044] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0045] Example 1

[0046] like Figure 1-2 As shown, this embodiment provides a structured scientific paper assisted creation method based on artificial intelligence, including:

[0047] Reference Figure 1The flowchart of the structured scientific paper auxiliary creation method of the present invention includes the steps of task definition and initialization (S1), structured decomposition and planning (S2), segmented guided AI content generation (S3), preliminary draft acquisition (S4), interactive iteration and refinement (S5-S7), cross-chapter consistency check (S8) and overall assembly and polishing (S9).

[0048] a) Task Definition and Initialization (S1): Receive the core elements of the paper from the user via an input device. These core elements serve as the foundation for subsequent generation and include at least: research topic, research objectives, key hypotheses or research questions, an overview of the main research methods, and a summary of core experimental data or key results. Optionally, target journal information, pre-defined arguments / conclusions, and a user-provided reference list or summary of key literature may also be received to provide richer context.

[0049] b) Structured decomposition and planning (S2): Based on pre-set, standard scientific paper structure rules (e.g., the IMRaD structure: Introduction, Methods, Results, Discussion) or the user-specified format requirements of the target journal, the paper to be written is logically divided into multiple, orderly logical sections (e.g., Abstract, Introduction, Literature Review, Methods, Results, Discussion, Conclusion, etc.). For each logical section, a clear writing goal is defined (e.g., the introduction should include sub-goals such as background, problem, motivation, and contributions), as well as the core content points to be covered, and possible internal substructures.

[0050] c) Segmented, guided AI content generation (S3): For at least one (usually each) logical part decomposed in step b), based on the core elements of the paper received in step a), the writing objectives and core content points of the logical part, and crucially, according to the preset specific prompt strategy template for the specific logical part (such as introduction, method, discussion, etc.), the system automatically generates structured guided prompts (Guided Prompts). These prompt strategy templates reflect the understanding of the rules of high-level paper writing (for example, integrating the specific writing requirements for different parts in the methodology such as "specific prompt strategy template"). The generated structured guided prompts not only contain the core content information that needs to be expressed in the part, but also contain a series of specific instructions and constraints for accurately guiding the output of the artificial intelligence language model (LLM). These instructions and constraints may include one or more of the following:

[0051] Role setting: Instructs LLM to write in the role of a senior researcher or expert in a specific field (such as materials science, biomedicine).

[0052] Contextual information: Provide necessary contextual information, such as research background, key assumptions, and the target journal's style preferences.

[0053] Content requirements: LLMs are explicitly asked to produce specific types of content (e.g., background introduction, problem statement, methodological details, data interpretation, argument development, comparative analysis, etc.), and may specify specific points to be included or emphasized.

[0054] Style and tone instructions: specify the required academic context (e.g., objectivity, rigor, criticality, persuasiveness) and control the professionalism and formality of the output.

[0055] Structural instructions: LLMs are required to organize the generated content according to a specific internal substructure (such as a four-paragraph structure for the introduction) or a logical sequence (such as presenting the argument first and then providing evidence).

[0056] Key information injection: Accurately embed the core data (e.g., efficiency increased by 30%), key terms (e.g., XX nanomaterials), expected conclusions, etc. provided by the user in step a) into the prompt to ensure that the generated content contains these key information.

[0057] Negative constraints: (Optional) Explicitly instruct the LLM to avoid certain common writing errors, redundant expressions, or inappropriate inferences.

[0058] d) Preliminary Draft Acquisition (S4): The structured guiding prompts generated in step c) are input into a predetermined artificial intelligence language model (e.g., GPT-4, Claude, or other large-scale language models with excellent performance in the field) via an application programming interface (API) or other communication methods. The model is instructed to generate a preliminary text draft corresponding to the logical part according to the prompt requirements.

[0059] e) Interactive Iteration and Refinement: Feedback Receiving (S5): The system presents the preliminary draft obtained in step d) to the user. The user reviews the draft and submits structured feedback on the draft. The structured feedback allows the user to provide precise modification instructions, such as:

[0060] Editing instructions for specific sentences or paragraphs, such as "rewrite," "expand," "simplify," "be more specific," "change tone," "provide reference suggestions," etc.

[0061] Supplementary information, new data, or specific revision suggestions.

[0062] Adjustment instructions regarding logical relationships or strength of argument.

[0063] Notes on factual errors, areas where supporting evidence is needed, or areas that do not conform to core elements.

[0064] f) Interactive Iteration and Refinement: Revision Prompt Generation (S6): The system receives and processes the structured feedback submitted by the user in step e). Based on the feedback and the original structured guidance prompts used to generate the preliminary draft, the system automatically generates a revision prompt. This revision prompt integrates the original requirements and the user's revision intentions.

[0065] g) Interactive iteration and refinement: Draft revision (S7): The revision prompt generated in step f) is input into the artificial intelligence language model again, and it is instructed to generate a revised draft according to the requirements of the revision prompt.

[0066] h) (Optional) Repeat steps e) to g) for multiple rounds of iterative refinement until the user is satisfied with the draft of the logic portion or it meets the preset quality assessment standards.

[0067] i) (Optional) Cross-section association and consistency check (S8): After the drafts of multiple logical sections (e.g., introduction, methods, results, discussion) have been generated to a certain extent or all have been preliminarily completed, a cross-section content association and consistency check is performed. This step aims to ensure the logical rigor and expression consistency of the paper as a whole. The check can be assisted by an artificial intelligence language model (by inputting specific prompts designed specifically for consistency checking) or a rule-based algorithm. Specific implementation methods include:

[0068] Extract key terms, research questions, hypotheses, results, and conclusions from each logical section;

[0069] Construct a logical relationship diagram between the key terms, research questions, hypotheses, results, and conclusions;

[0070] Analyzing potential inconsistencies in the logical relationship diagram based on preset consistency rules;

[0071] Generates a consistency check report that identifies potential inconsistencies and their specific locations.

[0072] Inspection contents may include but are not limited to:

[0073] Whether key terms are used consistently throughout the text.

[0074] Whether the research questions or objectives raised in the introduction are adequately addressed and resolved in the discussion or conclusion sections.

[0075] Is the description in the Methods section sufficient to support the data presented in the Results section?

[0076] Whether the statements in the Results section are consistent with the explanations and arguments in the Discussion section.

[0077] Whether the core ideas and key messages in the abstract, introduction, and conclusion are consistent.

[0078] j) (Optional) If the check in step i) reveals any inconsistencies or logical gaps in the content, the system may prompt the user and, based on the check results, guide the user back to step e) or automatically trigger step g) to adjust the draft of the relevant logical portion or generate necessary transitional statements.

[0079] k) Overall assembly and (optional) polishing (S9): After all logical parts have been refined (possibly multiple times), the final draft is assembled into a complete scientific paper draft according to the paper structure and sequence planned in step b). Optionally, after this step, artificial intelligence language models or integrated third-party tools (such as grammar checking software and reference management software interfaces) can be used to perform final language polishing (e.g., improving fluency and expression), grammar proofreading, spell checking, or automated or semi-automated typesetting adjustments and reference formatting according to the format required by the target journal.

[0080] Example 2

[0081] This embodiment provides an artificial intelligence-based structured scientific paper assisted creation method, including:

[0082] This embodiment describes in detail how to use the method and system of the present invention (such as Figure 4 The system architecture shown in Figure 2 is used to help generate the introduction section of a scientific paper. Suppose a user wants to write a research paper on "XX novel nanocatalyst for efficient ORR (oxygen reduction reaction)" and the target journal is Advanced Energy Materials.

[0083] Step S1: Task definition and initialization;

[0084] The user enters the following core elements of the paper through the system's receiving unit:

[0085] Research topic: Preparation of XX novel nanocatalyst and its application in efficient ORR;

[0086] Research objective: To verify that XX catalyst has higher ORR activity and stability than existing Pt / C catalysts;

[0087] Key hypothesis: The unique Y structure of the XX nanocatalyst can optimize the adsorption energy of reaction intermediates, thereby improving ORR performance;

[0088] Overview of the main methods: XX nanocatalyst was synthesized by solvothermal method and the ORR performance was tested by rotating disk electrode (RDE);

[0089] Key Data / Results Summary: The kinetic current density of the XX catalyst at 0.9 V vs RHE is 1.5 times higher than that of Pt / C, and the activity retention after 5000 cycles is 90%, while that of Pt / C is only 60%.

[0090] Target journal: Advanced Energy Materials;

[0091] (Optional) References: User uploaded 3 key literature summaries on ORR mechanism and advanced catalysts;

[0092] Step S2: structured decomposition and planning;

[0093] Reference Figure 2 In the schematic process, the system's decomposition planning unit decomposes the "Introduction" logical part into the following four internal substructures (or paragraph goals) based on the standard IMRaD structure and the common requirements for the introduction of top journals in the field of energy materials (these rules can be preset in the system):

[0094] P1: Background introduction: Explain the importance of energy conversion and storage, and introduce the urgent needs and challenges of efficient ORR catalysts.

[0095] P2: Research status and deficiencies: This paper reviews the current research progress of ORR catalysts (especially precious metal Pt / C and non-precious metal catalysts), and clearly points out their limitations (such as high cost, insufficient activity / stability, unclear mechanism, etc.), and highlights the knowledge gap of existing technologies.

[0096] P3: Research motivation and introduction of this work: Based on the shortcomings of P2, introduce the motivation of this research and briefly introduce the solution proposed in this paper (i.e., XX new nanocatalyst and its key feature Y structure), as well as the problems expected to be solved by this solution.

[0097] P4: Overview of main contributions and paper structure: Summarize the main research content, key findings (e.g., performance improvements and mechanistic insights) and main contributions of this paper, and briefly explain the arrangement of subsequent chapters of the paper.

[0098] Step S3: Segmented, guided AI content generation;

[0099] Reference Figure 2 Based on the details in S1, the system's prompt generation engine generates structured guiding prompts for each substructure of P1-P4 above, according to the preset specific prompt strategy templates applicable to each paragraph of "Introduction" (these templates embed the strategies for introduction writing in the "Specific Prompt Strategy Template"), and combines the core elements of S1 input.

[0100] Example of guiding prompts for P1:

[0101] json{

[0102] "role":"Senior researcher in the field of energy chemistry",

[0103] "context":"I am writing the first paragraph of the introduction for a paper on novel ORR catalysts to be submitted to Advanced Energy Materials.",

[0104] "content_requirements":[

[0105] "Overview of the challenges facing global energy sustainability and the importance of developing efficient energy conversion technologies such as fuel cells."

[0106] "It is pointed out that the oxygen reduction reaction (ORR) is the key bottleneck reaction in these technologies."

[0107] "This paper discusses the main challenges currently faced by ORR catalysts (especially the widely used Pt / C), such as high cost, resource scarcity, and insufficient stability."

[0108] "This leads to an urgent need for the development of low-cost, highly active, and highly stable non-precious metal ORR catalysts."],

[0109] "style_tone":"Objective, professional, highly summarized, problem-oriented",

[0110] "keywords_inject":["energy conversion","oxygen reduction reaction (ORR)","catalyst","Pt / C","high activity","high stability"],

[0111] "negative_constraints":["Avoid too much technical detail","Don't mention the specific work of this paper in this paragraph"]}

[0112] The guiding prompts for P2 will include instructions such as review requirements, comparison requirements, and clear identification of knowledge gaps. The guiding prompts for P3 will emphasize how to take on the challenges and knowledge gaps raised in the previous paragraph, and introduce the motivation for this study. The guiding prompts for P4 will include instructions such as summarizing contributions and outlining chapters.

[0113] Step S4: Obtaining a preliminary draft

[0114] The system sends the generated P1-P4 guiding prompts one by one or in combination to the predetermined large language model (such as GPT-4) through the AI model interface unit, and obtains the following preliminary draft:

[0115] The global energy crisis and increasingly severe environmental pollution necessitate the development of clean and efficient energy conversion technologies. Fuel cells, devices that directly convert chemical energy into electrical energy, are considered a crucial component of future energy systems due to their high efficiency and zero emissions. However, the sluggish kinetics of the oxygen reduction reaction (ORR) at the cathode severely restrict the overall performance and commercial application of fuel cells. Currently, platinum-based materials (Pt / C) are the most widely used ORR catalysts, but their high cost, resource scarcity, and insufficient long-term stability have made the development of low-cost, highly active, and stable non-precious metal ORR catalysts a key research focus.

[0116] [Preliminary draft contents of P2-P4 omitted below];

[0117] Steps S5, S6, and S7: interactive iteration and refinement;

[0118] The system displays the generated draft of the introduction to the user through the interactive processing unit. After reviewing it, the user finds that:

[0119] The review of non-precious metal catalysts in paragraph P2 is not comprehensive. User feedback: "Please add recent research progress in transition metal nitrides and carbon-based catalysts, and strengthen the comparison with the work in this paper."

[0120] The description of structure Y in paragraph P3 is unclear. The user selected the relevant sentence, provided the "Rewrite" instruction, and added the following explanation: "Emphasize how structure Y theoretically optimizes intermediate adsorption."

[0121] The system receives this structured feedback, generates revision prompts (for example, incorporating the user's supplementary information and "rewrite" instructions into the original prompts of P2 and P3), and calls the AI model again to generate a revised draft:

[0122] [Example of revised P2 paragraph];

[0123] In recent years, researchers have made significant progress in the field of non-precious metal ORR catalysts. Transition metal nitrides (such as Fe-NC and Co-NC) exhibit excellent ORR activity due to their abundant active sites and platinum-like electronic structure. The Fe-NC catalyst developed by Liu et al. (2021) achieved activity close to that of Pt / C in alkaline media, but its performance and long-term stability in acidic media still need to be improved. At the same time, various carbon-based materials (such as doped graphene and carbon nanotubes) have also become research hotspots due to their excellent conductivity and large specific surface area. However, these materials often suffer from low active site utilization and insufficient optimization of intermediate adsorption energy, resulting in their ORR performance still failing to fully meet commercial requirements. In particular, when it comes to optimizing the intermediate adsorption energy of the "O2→OOH→O / OH→OH-" reaction pathway, existing catalyst structure designs have yet to find the optimal balance, which is also a key factor limiting their performance.

[0124] [Example of revised P3 paragraph];

[0125] To address these challenges, this study proposed a novel nanocatalyst with a unique Y-structure, XX. By precisely controlling the electronic structure and coordination environment of the catalyst surface, the Y-structure can theoretically optimize the adsorption energies of O2 and its reaction intermediates (such as OOH*, O*, and OH*) to closer to ideal values, thereby lowering the reaction energy barrier and accelerating the electron transfer rate. Density functional theory (DFT) calculations indicate that the specific atomic coordination in the Y-structure creates an optimal d-band center position, favoring the breaking of O-O bonds and promoting proton-electron transfer, properties essential for efficient ORR catalysts.

[0126] This process can be repeated until the user is satisfied.

[0127] Step S8: (optional) cross-chapter association and consistency check;

[0128] Reference Figure 5 The consistency check mechanism shown here starts after the Introduction, Methods, Results, and Discussion sections are generated. For example, specific prompts can be used to instruct the AI to check: Is the goal of "optimizing ORR performance through Y-structure" proposed in Introduction P3 supported and justified by the results in the Discussion section? Are the test conditions described in the Methods section consistent with the data reported in the Results section? If any inconsistencies are found, the system prompts the user or guides them to make modifications.

[0129] Step S9: overall assembly and (optional) finishing;

[0130] The final drafts of all logical parts are assembled in sequence to form a complete introduction, and even the first draft of the entire paper. Users can choose to use the system's integrated polishing function for language checking and style optimization.

[0131] Example 3

[0132] This embodiment provides an artificial intelligence-based structured scientific paper assisted creation method, including:

[0133] Similarly, for the "Discussion" section, the system's decomposition planning unit will break it down into key sub-goals (such as: summarize the main findings, echo the introduction goal, explain the results (integrate the mechanism), compare with existing research, discuss the limitations of the research, explain the research significance, and look forward to future work). The prompt generation engine will generate guiding prompts based on these sub-goals and the discussion strategy in the "Specific Prompt Strategy Template". For example, for the sub-goal of "Compare with existing research", the prompt will clearly require the AI to:

[0134] Cite key references provided by users or retrieved by AI in S1.

[0135] Compare the key results of this study (core data in S1) with the results reported in these literatures quantitatively or qualitatively.

[0136] Analyze why the results of this study are superior to or different from existing studies, and emphasize the innovation and contribution of this study (combined with the key hypotheses in S1 and the mechanistic analysis generated in S3 / S4).

[0137] Instruct the AI to use a critical and objective tone for comparison.

[0138] Example 4

[0139] This embodiment provides an artificial intelligence-based structured scientific paper assisted creation method, including:

[0140] This example demonstrates how the method and system of the present invention can adapt to the needs of writing scientific papers in different disciplines. Suppose a user is writing a theoretical physics paper on the topological properties of quantum materials, and the target journal is Physical Review Letters.

[0141] Step S1: The user inputs core elements, including the research topic (anomalous quantum Hall effect in topological insulators), research objectives (predicting new topological materials and their properties through first-principles calculations), key assumptions (specific lattice structures can achieve quantum Hall effect at room temperature), etc.

[0142] Step S2: Based on the characteristics of top physics journals, the system breaks down the paper into logical sections appropriate to the field, such as abstract, introduction, theoretical model, computational methods, results analysis, discussion and outlook, and conclusion. For physics papers, the system places particular emphasis on the accuracy of mathematical formulas, the rigor of theoretical models, and the reproducibility of computational methods.

[0143] Step S3: The prompt generation engine generates guiding prompts based on the characteristics of the physics field using a specific prompt strategy template. For example, for the "theoretical model" section, the prompts may include:

[0144] Role setting: Instruct the AI to play the role of a theoretical physicist;

[0145] Content requirements: Detailed description of the Hamiltonian, boundary conditions, scope of application, etc.

[0146] Style directives: Emphasis on mathematical rigor and completeness of logical deduction;

[0147] Structural instructions: first give the general form of the equation, then give the simplified form for the specific case;

[0148] Key information injection: ensure inclusion of key theoretical predictions and calculation parameters provided by users;

[0149] Steps S4-S7: The system generates a preliminary draft, and the user provides feedback, especially correcting the accuracy of complex mathematical formulas and theoretical derivations. The system generates revision prompts and obtains a revised draft.

[0150] Step S8: For theoretical physics papers, the consistency check pays special attention to the consistency between the theoretical model and the calculation results, the consistency of the use of different mathematical symbols, the consistency of physical units, etc.

[0151] Step S9: Assemble the final paper, ensuring that the mathematical formulas are formatted correctly and meet the typesetting requirements of physics journals.

[0152] Example 5

[0153] This embodiment provides an artificial intelligence-based structured scientific paper assisted creation method, including:

[0154] To verify the effectiveness of this invention, the research team conducted a comparative experiment. Fifty researchers from diverse disciplines were divided into an experimental group and a control group, each consisting of 25 people. Both groups were required to write a first draft of a scientific paper based on given research data and materials. The experimental group used the system of this invention, while the control group used traditional methods or a general-purpose AI writing tool.

[0155] The results show:

[0156] Time efficiency: The average completion time for the experimental group was 5.2 hours, while that for the control group was 16.8 hours, an increase of approximately 69%.

[0157] Content quality: Senior scholars reviewed the papers for quality (out of 100 points). The experimental group scored an average of 82 points, while the control group scored 58 points, representing a 41% improvement in quality.

[0158] Structural integrity: The experimental group's paper structure score (out of 100 points) was 90 points, while the control group's was 65 points, an increase of approximately 38%.

[0159] Logical consistency: The experimental group had an average of 2.1 logical consistency defects, while the control group had 14.3 defects, a reduction of about 85%.

[0160] User satisfaction: The experimental group's satisfaction with the final draft (out of 10 points) was 8.5 points, while the control group's was 5.6 points, an increase of approximately 52%.

[0161] These data confirm the significant advantages of the present invention in improving the efficiency and quality of scientific paper creation.

[0162] Example 6

[0163] This embodiment provides an artificial intelligence-based structured scientific paper assisted creation method, including:

[0164] Receiving unit: receives user input (S1 core elements, S5 feedback).

[0165] Decomposition Planning Unit: This module implements S2 and includes built-in IMRaD and other structural rules and journal templates. This module includes adapters for different subject areas and journal types, automatically adjusting the paper structure plan to the specific requirements of the target journal.

[0166] Prompt Generation Engine: The core module, which executes S3, contains a prompt strategy template library for different parts (such as "specific prompt strategy templates"), and generates structured prompts based on input and templates. Based on the analysis of thousands of high-impact factor papers, the engine extracts writing patterns and strategies from different disciplines, different journals, and different paper parts, forming a comprehensive template library, such as Figure 3 shown.

[0167] AI Model Interface Unit: Manages communication with one or more external LLMs (such as the GPT-4 API), sends prompts, and receives results. This module implements load balancing and failover mechanisms to ensure system stability and responsiveness.

[0168] Interaction Processing Unit: Processes user feedback (S5), generates revision prompts (S6), and controls the iteration process (S7). This module supports multiple feedback methods, including text selection, annotation, commenting, and rating, and can intelligently analyze user intent.

[0169] Consistency check module: (Optional) Execute S8, such as Figure 5As shown, it contains consistency checking rules or logic for invoking AI for checking. This module constructs a logical relationship diagram for the entire text, analyzes the coherence between key terms, hypotheses, methods, results, and conclusions, and identifies potential logical breakpoints and inconsistencies.

[0170] Assembly: This module executes S9 to assemble the various sections into a complete document, and provides export and (optional) editing capabilities. This module supports multiple output formats (such as Word, LaTeX, and PDF) and automatically adjusts the layout and citation formatting to the target journal's requirements.

[0171] These modules can be implemented through software programming, run on computer hardware with sufficient computing resources (processor, memory), and interact through internal data buses or APIs.

[0172] The above are merely preferred embodiments of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. A structured scientific paper assisted creation method based on artificial intelligence, characterized in that: include: Receive user input on the core elements of the paper; Based on preset rules or user-specified requirements, the paper to be written is broken down into multiple logical parts, and writing objectives and core content points are defined for the multiple logical parts; Generate structured guiding prompts containing specific instructions and constraints based on the core elements of the paper, the writing objectives, the core content points and a preset specific prompt strategy template in the logical part; constructing an artificial intelligence language model, inputting the structured guiding prompts into the artificial intelligence language model, and obtaining a preliminary draft of the logic portion; Processing the preliminary draft of the logic section based on an interactive mechanism to obtain structured feedback; generating revision prompts based on the structured feedback and the structured guiding prompts; Inputting the revision prompt into the artificial intelligence language model to obtain a revised draft; Producing a complete first draft of the paper based on the preliminary draft and the revised draft.

2. The method according to claim 1, characterized in that The core elements of the paper include but are not limited to: research topic, research objectives, key assumptions or questions, overview of main research methods, core data and summary of results.

3. The method according to claim 1, characterized in that The preset specific prompt strategy template includes specific writing strategies for different logical parts of the paper, and the template for each logical part contains different instruction combinations and priorities.

4. The method according to claim 1, wherein After obtaining the preliminary draft of the logical part, it also includes: after generating drafts of multiple logical parts, performing the steps of cross-chapter content association and consistency check, the check includes at least one of the following: consistency in terminology use, consistency in introductory questions and conclusion responses, support for methods and results, and consistency between results and discussion explanations.

5. The method according to claim 4, characterized in that The consistency check is achieved by extracting key information of each logical part, constructing a logical relationship diagram, and analyzing it based on preset rules or by inputting specific check prompts using an artificial intelligence language model. If the check finds inconsistencies, revision prompts are regenerated based on the inconsistent content, and the revision prompts are input into the artificial intelligence language model to obtain a revised draft.

6. The method according to claim 5, characterized in that The process of regenerating the revision prompt includes: adjusting specific instructions and constraints in the structured guiding prompt according to the inconsistent content, and regenerating the revision prompt.

7. The method according to claim 6, characterized in that The step of adjusting the specific instructions and constraints in the structured guiding prompts includes: Modify one or more of the role settings, contextual information, content requirements, style and tone instructions, structural instructions, key information injection, and negative constraints in the structured guiding prompts to resolve the inconsistent content.

8. An artificial intelligence-based structured scientific paper auxiliary creation system, characterized by: include: A receiving unit, used to receive the core elements of the paper input by the user; A decomposition planning unit is used to decompose the paper to be written into multiple logical parts based on preset rules or user-specified requirements, and define writing goals and core content points for the multiple logical parts; A prompt generation engine is configured with a library of specific prompt strategy templates for different logical parts of the paper, and is used to generate structured guiding prompts containing specific instructions and constraints based on the core elements of the paper, the writing goals, the core content points and preset specific prompt strategy templates of the logical parts, and to generate structured guiding prompts containing at least three specified instructions; An AI model interface unit, configured to construct an artificial intelligence language model, input the structured guiding prompts into the artificial intelligence language model to obtain a preliminary draft of the logic portion, receive revision prompts and input them into the artificial intelligence language model to obtain a revised draft; an interactive processing unit, configured to process the preliminary draft of the logic portion based on an interactive mechanism, obtain structured feedback, and generate revision prompts based on the structured feedback and the structured guiding prompts; The assembly unit generates a complete first draft of the paper based on the preliminary draft and the revised draft.

9. An electronic device comprising a memory, a processor, and a computing program stored in the memory and executable on the processor, wherein: When the processor executes the computing program, the method according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images