Prompt message construction method and related equipment

By standardizing and mining the user's input query content, and combining the adaptation rules of the content processing model to construct prompt information, the problem of inaccurate capture of user intent in existing technologies is solved, and high-quality content generation is achieved.

CN123115390APending Publication Date: 2026-11-17SHENZHEN IWIN VISUAL TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611424760.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-09-14
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing content processing models fail to accurately capture user intent when faced with colloquial, fragmented, or ambiguous user needs, resulting in generated content that deviates from expectations and requires users to make repeated adjustments.

Method used

By acquiring the query content input by the user, we perform standardized processing, intent mining, and model adaptation rule construction to provide prompt information, including semantic correction, logical restructuring, and format standardization. We then combine user intent and model rules to generate professional prompt words.

Benefits of technology

It achieves fully automated construction of the entire process from the user's original query to professional prompts, reducing the user's need for professional knowledge and improving the accuracy, relevance and usability of the generated results.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123115390A_ABST
    Figure CN123115390A_ABST
Patent Text Reader

Abstract

This application relates to the field of computer technology and discloses a method and related equipment for constructing prompt information. The method can acquire query content input by a user for a content processing model; standardize the query content to obtain standardized target query content; perform intent mining processing on the target query content to obtain user intent; and construct prompt information based on the model adaptation rules of the content processing model, the user intent, and the target query content. The prompt information is used to input into the content processing model for analysis and reasoning to obtain query results corresponding to the query content. This application can achieve the automatic construction of high-quality prompt information, thereby effectively improving the accuracy, relevance, and usability of the results generated by the content processing model.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This application relates to the field of computer technology, specifically to a method for constructing prompt messages and related equipment. Background Technology

[0002] With the rapid development of artificial intelligence technology, various content processing models have been widely applied in many fields such as text generation, code writing, data analysis, and creative creation. However, in practical applications, users often express their needs in a conversational, fragmented, or even vague way, such as "Help me write a more formal proposal" or "Make a diagram, it needs to look good." Existing systems often directly input these raw inputs as prompts into the model. There is a significant gap between user input and model understanding, causing the model to fail to accurately capture the user's true intent. The generated content deviates from expectations, and users need to make repeated manual adjustments. Summary of the Invention

[0003] This application provides a method for constructing prompt information and related equipment; the related equipment may include a prompt information construction device, an electronic device, and a computer-readable storage medium. This application can realize the automatic construction of high-quality prompt information, thereby effectively improving the accuracy, relevance, and usability of the results generated by the content processing model.

[0004] This application provides a method for constructing prompt information, including: Obtain the query content input by the user for the content processing model; The query content is standardized to obtain the standardized target query content; The target query content is subjected to intent mining processing to obtain the user intent; Based on the model adaptation rules of the content processing model, the user intent, and the target query content, a prompt message is constructed. The prompt message is used to input the content processing model for analysis and reasoning to obtain the query results corresponding to the query content.

[0005] Accordingly, embodiments of this application provide a prompt information construction device, including: The acquisition unit is used to acquire the query content input by the user for the content processing model; A standardization unit is used to standardize the query content to obtain the standardized target query content; The intent mining unit is used to perform intent mining processing on the target query content to obtain the user intent; The prompt construction unit is used to construct prompt information based on the model adaptation rules of the content processing model, the user intent, and the target query content. The prompt information is used to input the content processing model for analysis and reasoning to obtain the query results corresponding to the query content.

[0006] Optionally, in some embodiments of this application, the standardization unit may include a semantic correction subunit, a logical reconstruction subunit, and a format standardization subunit, as follows: The semantic correction subunit is used to perform semantic correction processing on the query content to obtain the corrected query content; The logical reconstruction subunit is used to perform logical reconstruction processing on the corrected query content to obtain the reconstructed query content. The format standardization subunit is used to perform format standardization processing on the reconstructed query content to obtain the standardized target query content.

[0007] Optionally, in some embodiments of this application, the prompt construction unit may include a scene label determination subunit and a construction subunit, as follows: The scene tag determination subunit is used to extract keywords from the target query content and determine the usage scene tag based on the keywords. A sub-unit is constructed to build prompt information based on the usage scenario tags, the model adaptation rules of the content processing model, the user intent, and the target query content.

[0008] Optionally, in some embodiments of this application, the scene tag determination subunit may be specifically used to perform matching calculations between the keyword and candidate scene tags in the scene tag lexicon to obtain the matching degree between the keyword and the candidate scene tags; detect whether there are candidate scene tags in the scene tag lexicon with a matching degree higher than a preset threshold; if there are, determine the usage scene tag based on the candidate scene tags with a matching degree higher than the preset threshold; if there are no, output the candidate scene tags for the user to select and confirm, and use the scene tag confirmed by the user as the usage scene tag.

[0009] Optionally, in some embodiments of this application, the prompt construction unit may include a mapping set acquisition subunit, a query subunit, and a prompt construction subunit, as follows: The mapping set acquisition sub-unit is used to acquire the model rule mapping relationship set. The model rule mapping relationship set includes a preset model and a preset model adaptation rule. The preset model adaptation rule indicates the usage rules, output characteristics, and prompt construction standards of the corresponding model in a preset scenario. The query subunit is used to query the model adaptation rules of the content processing model in the model rule mapping relationship set; The prompt construction subunit is used to construct prompt information based on the model adaptation rules of the content processing model, the user intent, and the target query content.

[0010] Optionally, in some embodiments of this application, the standardization unit can be specifically used to standardize the query content by constructing a model through prompts, so as to obtain the standardized target query content; The prompt information construction device further includes a quality analysis unit, a feedback evaluation unit, and an update unit, as follows: The quality analysis unit is used to perform quality analysis on the query results output by the content processing model to obtain quality analysis results; The feedback evaluation unit is used to determine the feedback evaluation results based on the user's operation records and usage feedback during the query process; The update unit is used to update the prompt construction model based on the quality analysis results and the feedback evaluation results to obtain the updated prompt construction model.

[0011] Optionally, in some embodiments of this application, the updating unit may be specifically used to determine a comprehensive evaluation index for the prompt information based on the quality analysis results and the feedback evaluation results; construct a prompt sample based on the comprehensive evaluation index; and update the prompt construction model according to the prompt sample to obtain an updated prompt construction model.

[0012] Optionally, in some embodiments of this application, the prompt information construction device may further include a requirement parameter determination unit, a missing condition identification unit, and a completion unit, as follows: The requirement parameter determination unit is used to determine the structured requirement parameters of the target query content based on the user intent. The missing condition identification unit is used to identify implicit missing conditions in the structured requirement parameters based on the domain knowledge graph. The completion unit is used to perform the content completion operation on the target query content for the implicit missing conditions, so as to obtain the completed target query content.

[0013] Optionally, in some embodiments of this application, the completion unit may specifically be used to perform confidence assessment processing on the implicit missing conditions to obtain a confidence score; when the confidence score does not exceed a preset threshold, an interactive follow-up question is generated based on the implicit missing conditions and output, so that the user can supplement the content based on the interactive follow-up question, and perform a content completion operation on the target query content according to the supplemented content to obtain the completed target query content; when the confidence score exceeds a preset threshold, the content completion operation of the implicit missing conditions is performed on the target query content to obtain the completed target query content.

[0014] In the embodiments of this application, the terms "module" or "unit" refer to a computer program or part of a computer program that has a predetermined function and works with other related parts to achieve a predetermined goal, and can be implemented wholly or partially using software, hardware (such as processing circuitry or memory), or a combination thereof. Similarly, a processor (or multiple processors or memory) can be used to implement one or more modules or units. Furthermore, each module or unit can be part of an overall module or unit that includes the functionality of that module or unit.

[0015] An electronic device provided in this application includes a processor and a memory. The memory stores multiple instructions, and the processor loads the instructions to execute the steps in the prompt message construction method provided in this application.

[0016] This application also provides a computer-readable storage medium storing a computer program thereon, wherein the computer program, when executed by a processor, implements the steps in the prompt message construction method provided in this application.

[0017] Furthermore, embodiments of this application also provide a computer program product, including a computer program or instructions, which, when executed by a processor, implement the steps in the prompt information construction method provided in embodiments of this application.

[0018] This application provides a method and related equipment for constructing prompt information. The method can acquire query content input by a user for a content processing model; standardize the query content to obtain standardized target query content; perform intent mining processing on the target query content to obtain user intent; and construct prompt information based on the model adaptation rules of the content processing model, the user intent, and the target query content. The prompt information is used to input into the content processing model for analysis and reasoning to obtain query results corresponding to the query content.

[0019] This application standardizes the user's original query content, effectively eliminating issues such as colloquial expressions and logical inconsistencies, resulting in standardized target query content and narrowing the semantic gap between user input and model understanding. Building upon this, intent mining processing of the target query content accurately captures the user's core needs, avoiding intent deviations or comprehension biases caused by directly constructing prompts based on the original input. Furthermore, by introducing model adaptation rules from the content processing model, and combining the identified user intent with the standardized target query content to construct prompts, the generated prompts can automatically adapt to the specific model used. Overall, this method achieves fully automated construction of professional prompts from the user's original query, requiring no professional prompting engineering knowledge from the user, significantly lowering the barrier to entry, while effectively improving the accuracy, relevance, and usability of the results generated by the content processing model. Attached Figure Description

[0020] To more clearly illustrate the technical features of the embodiments of this application, the drawings used in the embodiments of this application will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1a This is a flowchart of the prompt information construction method provided in the embodiments of this application; Figure 1b This is an explanatory diagram of the prompt information construction method provided in the embodiments of this application; Figure 2 This is another flowchart of the prompt information construction method provided in the embodiments of this application; Figure 3 This is a schematic diagram of the structure of the prompt information construction device provided in the embodiments of this application; Figure 4 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0022] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0023] This application provides a method for constructing prompt messages and related equipment. The related equipment may include a prompt message construction device, an electronic device, and a computer-readable storage medium. Specifically, the prompt message construction device may be integrated into an electronic device, such as a terminal.

[0024] It should be noted that the order of description of the following embodiments is not intended to limit the preferred order of the embodiments.

[0025] This embodiment will be described from the perspective of a prompt information construction device, which can be integrated into an electronic device, such as a terminal or server.

[0026] Terminals can include mobile phones, vehicle terminals, aircraft, tablets, laptops, or personal computers (PCs). Clients can also be installed on the terminal; these clients can be application clients or browser clients, etc.

[0027] A server can be a standalone physical server, a server cluster or distributed system consisting of multiple physical servers, or a cloud server that provides cloud computing services.

[0028] In the description of this application, it should be understood that the numbering itself, such as "first", "second", etc., is only used to distinguish the described objects and has no sequential or technical meaning, and should not be construed as specifying or implying the importance of the described objects.

[0029] In the description of this application, the term "multiple" refers to two or more. Furthermore, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this document generally indicates that the preceding and following related objects have an "or" relationship.

[0030] In the description of this application, it should be noted that, unless otherwise defined, all technical and scientific terms used in this application have the same meaning as commonly understood by one of ordinary skill in the art. The terminology used in this application is for the purpose of describing specific embodiments only and is not intended to limit the application. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.

[0031] It is understood that in the specific embodiments of this application, data such as user information are involved. When the above embodiments of this application are applied to specific products or technologies, user permission or consent is required, and the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0032] like Figure 1a As shown, the specific process for constructing this prompt message can be as follows: 101. Obtain the query content input by the user for the content processing model.

[0033] The content modality of the query content can be voice, text, image, or document, etc., and this embodiment does not impose any restrictions on it.

[0034] Specifically, a content processing model refers to an artificial intelligence model used to understand, analyze, and reason about input content and generate corresponding output results, including but not limited to large language models (LLM), multimodal large models, code generation models, and text-to-graph models. In this embodiment, the content processing model is the final recipient and executor of the prompt information.

[0035] In specific scenarios, the system can receive raw query content through the user interface. The query content can be text manually entered by the user or text converted from voice input through speech recognition. For example, the user's query content could be: "Help me write something about something the company needs next week, for that new product launch, something formal, for the boss to see."

[0036] 102. Standardize the query content to obtain the standardized target query content.

[0037] Since user input often contains colloquial expressions, redundant information, and logical inconsistencies, this embodiment can first standardize the query content, converting it into target query content that is clearly structured, semantically clear, and formatted.

[0038] Optionally, in this embodiment, the step "standardizing the query content to obtain standardized target query content" may include: The query content is semantically corrected to obtain the corrected query content; The corrected query content is logically reconstructed to obtain the reconstructed query content; The reconstructed query content is then formatted to obtain the standardized target query content.

[0039] In this embodiment, the standardization process adopts a three-level progressive process, including three sub-steps: semantic correction, logical reconstruction, and format standardization.

[0040] Semantic correction processing specifically involves semantic error correction and clarification of the original query content. This includes: identifying and correcting colloquial expressions (e.g., mapping "write something" to "write a document"), eliminating semantic ambiguity (e.g., determining from the context that "new product launch" is the document topic rather than the event schedule), and filtering invalid and redundant information (e.g., removing filler words such as "that one" and "help me" that do not affect the core semantics).

[0041] For example, after semantically correcting a user's input "Help me write something about the company's new product launch next week, something formal for the boss to see," the result is: "Write a formal document about the company's new product launch for the boss to review and use next week."

[0042] Based on semantic correction, the system reorganizes the logical structure of the query content, i.e., logical reconstruction, to make it conform to the standard logical framework of "action-object-constraint-purpose". The system can identify the core action (writing), the target object (new product launch document), the constraints (formal style, to be used next week), and the target audience (leaders) in the query, and reorganize the statements according to the preset logical template.

[0043] For example, after logical restructuring, we can obtain: "[Action] Write documents related to the new product launch conference; [Topic] Company new product launch conference; [Style Requirements] Formal and rigorous; [Audience] Company leaders; [Time Constraint] To be used next week." Finally, a format standardization process is performed, which converts the logically restructured query content into a unified standard format within the system, such as structured text or JSON, so that subsequent modules can parse it by machine. The standardized format includes predefined fields such as query_type, subject, style, audience, and constraints.

[0044] In a specific embodiment, this standardization process can be automatically completed by a prompting model, which can be a finely tuned lightweight language model (such as one based on BERT). Its training data includes a large number of colloquial queries and their corresponding standardized query content pairs. After receiving the original query, the model directly outputs the standardized target query content.

[0045] 103. Perform intent mining processing on the target query content to obtain the user intent.

[0046] Specifically, this embodiment can use a neural network model to perform deep semantic analysis on the standardized target query content to identify the user's core needs. This neural network model can be the aforementioned prompt construction model. For example, from the above example, the user intent is mined as: "Generate a formal document related to a new product launch conference for company leaders."

[0047] 104. Based on the model adaptation rules of the content processing model, the user intent, and the target query content, construct prompt information. The prompt information is used to input the content processing model for analysis and reasoning to obtain the query results corresponding to the query content.

[0048] In this embodiment, the system retrieves the corresponding model adaptation rules based on the content processing model selected by the user or the system's default model. Combining the identified user intent with the standardized target query content, the system constructs professional prompt information that adapts to the characteristics of the model. This prompt information can be directly input into the content processing model for analysis and reasoning to obtain high-quality query results.

[0049] Model adaptation rules can specifically be a set of standardized requirements pre-configured for a specific content processing model to guide the construction of prompt words. Specifically, model adaptation rules can include: usage rules for the corresponding model (such as optimal input length, sensitive word restrictions, and special marker usage guidelines), output characteristics (such as preferred answer styles, default output formats, and common illusion types), and prompt construction standards in preset scenarios (such as the role setting template, thought chain construction method, and output format constraints recommended by the model in the "official document writing" scenario). Model adaptation rules can be stored in a set of model rule mapping relationships in the form of structured data.

[0050] Optionally, in this embodiment, the step "constructing prompt information based on the model adaptation rules of the content processing model, the user intent, and the target query content" may include: Extract keywords from the target query content, and determine usage scenario tags based on the keywords; Based on the usage scenario tags, the model adaptation rules of the content processing model, the user intent, and the target query content, a prompt message is constructed.

[0051] Specifically, usage scenario tags can be category markers used to identify the application domain to which the user's query content belongs, such as "official document writing," "code generation," "marketing copywriting," "academic papers," and "data analysis." Usage scenario tags are used to trigger corresponding scenario-specific optimization strategies.

[0052] In specific implementations, core keywords can be extracted from the target query content using TF-IDF, TextRank, or deep learning-based keyword extraction models.

[0053] Optionally, in this embodiment, the step "determine usage scenario tags based on the keywords" may include: The matching degree between the keyword and the candidate scene tag in the scene tag thesaurus is calculated by matching the keyword with the candidate scene tag. Detect whether there are candidate scene tags in the scene tag lexicon with a matching degree higher than a preset threshold; If they exist, determine which scene tag to use based on candidate scene tags with a matching degree higher than a preset threshold; If no candidate scene tag exists, the candidate scene tag is output for the user to select and confirm, and the scene tag confirmed by the user is used as the usage scene tag.

[0054] The preset threshold can be set according to the actual situation, and this embodiment does not impose any restrictions on it.

[0055] Specifically, the scene tag lexicon pre-stores multi-level candidate scene tags, for example: Primary tags: Official document writing, code development, marketing creativity, academic research; Secondary tags (under official document writing): Speeches by leaders, work reports, notices and announcements, meeting minutes.

[0056] The system can perform semantic matching between extracted keywords and candidate scene tags. This matching calculation can be based on cosine similarity calculation based on word vectors, semantic similarity matching based on semantic models, or rule-based keyword matching.

[0057] For example, the extracted keywords include: "leader's speech", "new product launch", "formal", "market effect", and "company leader". Calculations show a match of 0.92 with the candidate scenario tag "official document writing - leader's speech"; 0.65 with the candidate scenario tag "marketing creativity - product promotion"; and 0.12 with the candidate scenario tag "academic research - paper abstract".

[0058] The system sets a preset threshold (e.g., 0.80), and the detection found that the matching degree of "official document writing - leader's speech draft" was 0.92, which is higher than the threshold.

[0059] The system identifies "Official Document Writing - Leadership Speech" as a usage scenario tag. This tag will trigger the following scenario-specific optimization strategies: automatically applying the standard structure of a leadership speech (opening remarks → product introduction → market outlook → thank-you speech), enabling formal and solemn language style constraints, and adding rules to filter colloquial expressions, etc.

[0060] In another embodiment, if the user input is extremely vague (e.g., "write some copy"), and the matching degree between the keyword and multiple candidate scenario tags is lower than a preset threshold (e.g., "marketing copy" 0.45, "social media copy" 0.42, "advertising slogan" 0.38), then the system will sort the candidate scenario tags and their corresponding matching degrees and output them for the user to select and confirm. User interface display: The system has detected a scenario that you might need (please select the one that best matches): Marketing copywriting (45% match) Social media copywriting (42% match) Ad slogan (38% match) If none of the above apply, I will manually enter " After the user confirms their selection, the system will use the user-confirmed scenario tag as the usage scenario tag.

[0061] In some embodiments, the usage scenario tags can be manually selected by the user. Specifically, after entering their query, the user can easily select the corresponding usage scenario and AI model. In other words, the usage scenario tags are selected by the user through clicking or dropping down on the front-end interface, and the system is only responsible for invoking the corresponding optimization strategy based on the user's selected scenario.

[0062] Optionally, in this embodiment, the step "constructing prompt information based on the model adaptation rules of the content processing model, the user intent, and the target query content" may include: Obtain a set of model rule mapping relationships, which includes a preset model and preset model adaptation rules. The preset model adaptation rules indicate the usage rules, output characteristics, and prompt construction standards of the corresponding model in a preset scenario. In the set of model rule mapping relationships, query the model adaptation rules of the content processing model; Based on the model adaptation rules of the content processing model, the user intent, and the target query content, a prompt message is constructed.

[0063] Specifically, the system maintains a set of model rule mapping relationships in the database, which is stored in the form of a structured data table or a knowledge base. Each record contains: a preset model identifier and its corresponding preset model adaptation rule.

[0064] For example, the model adaptation rules for the GPT model can be as follows: Usage rules: Supports long contexts (up to 128K tokens), excels at handling complex reasoning tasks, and responds well to role prompts; Output characteristics: The default output is quite detailed and may contain too much explanatory content; it is sensitive to the "step-by-step thinking" command; Scenario Construction Standards (Leadership Speech Scenario): It is recommended to adopt a four-part prompt structure of "role setting + background information + structural constraints + output format"; it is suggested to explicitly require "solemn language and avoid colloquialisms"; Markdown format can be used to constrain the output.

[0065] Specifically, the step "constructing prompt information based on the model adaptation rules of the content processing model, the user intent, and the target query content" may include: assembling the model adaptation rules, the user intent, and the target query content to obtain prompt information.

[0066] Optionally, in this embodiment, the step "standardizing the query content to obtain standardized target query content" may include: By constructing a model based on prompts, the query content is standardized to obtain the standardized target query content. The method for constructing this prompt message may also include: The quality analysis results are obtained by performing quality analysis on the query results output by the content processing model. The feedback evaluation results are determined based on the user's operation records and usage feedback during the query process; Based on the quality analysis results and the feedback evaluation results, the prompt construction model is updated to obtain the updated prompt construction model.

[0067] Specifically, the suggestion building model can be a lightweight artificial intelligence model or rule engine used to perform query content standardization, intent mining, keyword extraction, and suggestion information construction. This model can be continuously updated based on quality analysis results and feedback evaluation results through supervised learning or reinforcement learning.

[0068] In this embodiment, the operation record specifically represents the user's implicit behavioral feedback to the prompt words, while the usage feedback specifically represents explicit feedback. Implicit behavioral feedback can include prompt word copying, prompt word editing, result regeneration, and result export. For the feedback evaluation result, differentiated weights can be assigned to explicit feedback and various types of implicit behavioral feedback, with explicit feedback having a higher weight than implicit behavioral feedback. Then, based on the scores corresponding to explicit feedback and various types of implicit behavioral feedback, and their respective weights, a weighted sum is performed to obtain the feedback evaluation result.

[0069] Specifically, the semantic interpretation of implicit behavioral feedback can be as follows: interpreting the prompt word editing behavior as a negative signal that the generated result does not fully meet the requirements, and interpreting the generated result export behavior as a positive signal that the generated result meets the requirements.

[0070] In a specific embodiment, the query results can be subjected to multi-dimensional quality analysis, including but not limited to: accuracy, completeness, style fit, and format compliance. Accuracy refers to whether the generated content accurately reflects the product information requested by the user; completeness reflects whether it covers all elements specified in the structured requirement parameters; style fit refers to whether the language style meets the requirement of "formal and dignified"; and format compliance refers to whether the output format meets preset constraints.

[0071] By weighting and fusing the quality scores of each analytical dimension, a comprehensive quality score can be obtained, which serves as the final quality analysis result.

[0072] In specific scenarios, the system can record user actions during the query process, including: whether the generated results were manually edited (and the amount of editing), whether "regenerate" was clicked, whether the results were copied and used, and the duration of user interaction after use. Simultaneously, the system collects subjective user feedback through the interface, such as five-star ratings and textual reviews ("perfectly meets requirements" or "needs to be more concise"). For example, if a user gives a five-star rating and comments "The content is excellent, I adopted it directly," the feedback evaluation result would be: user satisfaction 100%, adoption rate 100%.

[0073] Optionally, in this embodiment, the step "updating the prompt construction model based on the quality analysis results and the feedback evaluation results to obtain the updated prompt construction model" may include: Based on the quality analysis results and the feedback evaluation results, a comprehensive evaluation index for the prompt information is determined; Based on a comprehensive evaluation index, construct a sample of prompts; Based on the given prompt sample, the prompt construction model is updated to obtain the updated prompt construction model.

[0074] The comprehensive evaluation index is a composite indicator used to quantify the quality of prompt information. It can be calculated by weighting the quality score (such as accuracy, completeness, and relevance) output by the content processing model caused by the prompt information with the user's subjective feedback score, and the value range can be from 0 to 100.

[0075] When the comprehensive evaluation index exceeds a preset excellent threshold (e.g., 90 points), the system constructs a high-quality prompt sample from the entire process data. This prompt sample can include: original query content → standardized processing → intent mining results → implicit missing conditions → completed target query content → usage scenario tags → model adaptation rules → final prompt information → quality analysis results → user feedback. This sample is labeled as a "positive example".

[0076] When the comprehensive evaluation index is lower than the preset improvement threshold (e.g., 60 points), the system constructs "negative example" samples and marks the specific defect dimensions (e.g., "inconsistent style" or "incomplete information").

[0077] Specifically, the system can periodically use accumulated prompt samples as training data to fine-tune and update the prompt-building model. Update strategies may include: Reinforcement learning based on positive examples: Improve the model's standardized processing logic and prompt construction strategy in similar scenarios; Negative example-based error correction learning: Adjusting the model's bias in specific intent recognition or condition completion; Adjusting model adaptation rules based on user feedback: If user feedback for a certain model remains low in a specific scenario, then update the model adaptation rules for that scenario (such as adjusting output format requirements or style constraint statements).

[0078] Through the aforementioned iterative mechanism, the system can continuously accumulate high-quality prompt word cases, update the recognition rules and optimization logic in reverse, and achieve a self-evolution effect of "becoming more accurate with use".

[0079] Optionally, in this embodiment, before the step "constructing prompt information based on the model adaptation rules of the content processing model, the user intent, and the target query content", the following may be included: Based on the user intent, determine the structured requirement parameters for the target query content; Based on the domain knowledge graph, identify the implicit missing conditions in the structured requirement parameters; The target query content is then filled with the implicit missing conditions to obtain the filled target query content.

[0080] In specific scenarios, users often have a lot of implicit conditions (such as target audience, format specifications, constraints, etc.) when expressing their needs. This embodiment can perform deep semantic analysis and implicit condition mining on user input to obtain complete prompt word information.

[0081] Specifically, a domain knowledge graph refers to a semantic network structure built for a specific application domain, containing entities within that domain (such as document types, industry terms, and roles), entity attributes (such as formatting requirements and style characteristics), and relationships between entities (such as the relationship between "product launch" and "speech draft," or the constraint relationship between "leadership report" and "formal format"). Domain knowledge graphs are used to identify implicit needs that are not explicitly mentioned in user queries but are strongly related to the scenario.

[0082] Structured requirement parameters can be data structures with clearly defined fields, formed after parsing user intent, such as {Document Type: "Speech", Audience: "Company Leaders", Style: "Formal", Topic: "New Product Launch", Word Count Requirement: null, Deadline: "Next Week"}. Structured requirement parameters are used to systematically express the various dimensions of user needs.

[0083] Implicit missing conditions specifically refer to requirement elements in structured requirement parameters that should exist based on the current use case and domain knowledge but have not been provided or have been vaguely described by the user. For example, a user requests "write a solution" but does not specify the solution type, target audience, format requirements, etc. These unspecified elements are implicit missing conditions.

[0084] In specific scenarios, the system maps user intents into structured requirement parameters based on the intent mining results. Taking the aforementioned example, the structured requirement parameters initially extracted by the system are: Document type: Not specified (user only said "thing") Subject: New Product Launch Style: Formal Target audience: Company leaders Purpose: To be used next week Word count requirement: Not mentioned Formatting guidelines: Not mentioned Key selling points: Not mentioned Then, the system calls the domain knowledge graph corresponding to the "enterprise document writing" domain. The knowledge graph can store entity relationships: "New product launch" → common document types → ["press release", "leader's speech", "project proposal", "invitation letter"]; "leader's review" → document style constraints → ["complete structure", "data support", "decision suggestion"]; "use next week" → time constraints → ["must include time node", "reserve approval period"].

[0085] Through graph reasoning, the system identified the following latent missing conditions: The document type is unclear (is it a speech draft or an event plan?). Specific information about the new product is missing (product name, core functions, target market); The decision-making dimensions that leaders focus on are missing (budget, expected results, risk assessment).

[0086] Finally, the system can assess the confidence level of the identified implicit missing conditions to determine whether there is sufficient evidence to automatically fill in the missing information, or whether it is necessary to initiate interactive follow-up questions to the user.

[0087] Optionally, in this embodiment, the step "performing the implicit missing condition completion operation on the target query content to obtain the completed target query content" may include: The confidence score is obtained by performing a confidence assessment on the latent missing conditions. When the confidence score does not exceed a preset threshold, an interactive follow-up question is generated and output based on the implicit missing condition, so that the user can supplement the content based on the interactive follow-up question, and perform a content completion operation on the target query content according to the supplemented content to obtain the completed target query content. When the confidence score exceeds a preset threshold, the target query content is filled with the implicit missing conditions to obtain the filled target query content.

[0088] In specific scenarios, the system can calculate a confidence score (0-100 points) for each implicit missing condition. The scoring criteria include: the frequency of occurrence of the condition in the corresponding scenario, the default value of the condition in the user's historical behavior data, and the semantic relevance of the current query context.

[0089] Specifically, the system can automatically complete constraints with high confidence levels (confidence score > preset threshold, such as 70 points). For example, for some general constraints, the system can directly complete the constraints based on default values. For instance, in the "leadership review" scenario, the system will automatically complete the format constraint that "formal documents must include five parts: summary, background, solution details, expected results, and execution plan."

[0090] Low-confidence interactive follow-up questions (confidence score ≤ preset threshold): When core information is missing (such as document type, product details), the system generates interactive follow-up questions and outputs them to the user.

[0091] Taking the aforementioned example, the system generates the following follow-up question: To better generate the document for you, please add the following information: What type of document do you need? (e.g., speech drafts for leaders / event execution plans / press releases) What is the name and key selling points of the new product? Are leaders more focused on budget control or market performance forecasting? The user added, "It's a speech draft for a leader. The product is called 'xxx,' and it's market-oriented AI assistant. The leader is concerned about its market performance." Based on this additional information, the system completed the target query, and the updated structured requirements parameters were fully populated with fields such as document type, product information, and key focus.

[0092] The prompts obtained through the above construction can be directly input into the corresponding content processing model. Since the prompts have undergone semantic correction, logical reconstruction, implicit condition completion, scene label adaptation, and model rule adaptation, the model output is significantly better than the effect of directly using the original colloquial input in terms of structural integrity, style fit, and content accuracy.

[0093] This application provides a method for constructing prompt information. By standardizing the user's original query content, it effectively eliminates problems such as colloquial expressions, semantic ambiguity, and logical inconsistencies, transforming fragmented natural language input into structurally standardized and semantically clear target query content, significantly narrowing the semantic gap between user input and model understanding. Furthermore, by performing intent mining on the target query content, it accurately captures the user's core needs, avoiding intent deviations or comprehension biases caused by directly constructing prompt information based on the original input. Further, by introducing model adaptation rules from content processing models, and combining the identified user intent with the standardized target query content to construct prompt information, the generated prompt information can automatically adapt to the usage rules, output characteristics, and prompt construction standards of different models. This achieves a universal conversion of the same original query to multiple content processing models, solving the problem of mismatch between traditional prompt words and specific models. Overall, this method achieves fully automated construction of professional prompt information from the user's original query, requiring no professional prompt engineering knowledge from the user, significantly lowering the usage threshold, and effectively improving the accuracy, relevance, and usability of the results generated by the content processing model.

[0094] Specifically, the prompt information construction method provided in this application adopts a lightweight modular architecture, which includes three core functional modules: intelligent parsing and preprocessing module, cross-model adaptive optimization module, and data iterative operation and maintenance module.

[0095] The intelligent parsing and preprocessing module is mainly responsible for processing the user's raw input. Regardless of whether the user types or speaks, the system first converts it into text, automatically filters out irrelevant and invalid content, understands what the user really wants to do, and fills in any implicit needs or missing conditions that the user did not clearly express. It organizes the messy, colloquial requirements into standardized requirements parameters that the system can recognize, laying the foundation for subsequent optimizations.

[0096] The cross-model adaptive optimization module is responsible for truly optimizing the prompts. The system pre-stores the usage rules and output characteristics of various AI models. Based on the AI ​​model selected by the user and the usage scenario, it automatically performs five tasks: correcting ambiguous semantics, reconstructing the logical structure of the prompts, supplementing scenario-specific constraints, adapting to the corresponding model parameters, and standardizing the output format. Ultimately, it transforms ordinary, casual descriptions into rigorous and user-friendly professional prompts adapted to the current AI model.

[0097] The data iteration and maintenance module is responsible for the system's self-upgrade. The system records the user experience and the quality of AI-generated content, accumulating high-quality prompt word examples. This data, in turn, continuously updates the system's recognition rules and optimization logic, making the system more accurate with use and solving the problem of traditional tools remaining unchanged and unable to improve.

[0098] The system operation process may include: receiving raw user input → performing standardized preprocessing and intent mining → identifying and filling in implicit missing conditions → determining usage scenario tags → querying the adaptation rules of the target model → constructing professional prompt information → input content processing model to obtain query results → collecting feedback and quality data → iteratively optimizing system rules.

[0099] In specific scenarios, refer to Figure 1b The intelligent parsing preprocessing module preprocesses user-input queries, including redundancy removal, requirement completion, and structuring. The cross-model adaptive optimization module adapts to specific model rules, determining the appropriate rules for the current model and performing semantic correction, structural reconstruction, format standardization, and model adaptation on the preprocessed queries. This outputs optimized professional suggestions, which are then input into the connected AI model (the content processing model mentioned above) to generate content. Finally, the data iteration and maintenance module uses the AI ​​model-generated content and user feedback to accumulate samples. These samples are then used to update the rules for building the model and iteratively upgrade it.

[0100] The system supports multiple input methods, including but not limited to: manual keyboard input and voice input (the system automatically converts speech into text using speech recognition technology). Users can access the system via mobile phones, computers, and other terminal devices. The input method is simple and requires no knowledge of professional prompt words.

[0101] In terms of hardware deployment, for everyday lightweight use scenarios, the system can be deployed on a basic server configured with 4 cores and 8GB of RAM; for commercial high-concurrency scenarios, an 8-core, 16GB cloud server cluster can be used, along with conventional load balancing strategies. On the software side, the system is built on a conventional server operating system, backend runtime environment, and relational / non-relational database, requiring neither high-end computing power nor complex development.

[0102] As can be seen from the above, this embodiment can obtain the query content input by the user for the content processing model; perform standardization processing on the query content to obtain the standardized target query content; perform intent mining processing on the target query content to obtain the user intent; and construct prompt information based on the model adaptation rules of the content processing model, the user intent, and the target query content. The prompt information is used to input the content processing model for analysis and reasoning to obtain the query results corresponding to the query content.

[0103] This application standardizes the user's original query content, effectively eliminating issues such as colloquial expressions and logical inconsistencies, resulting in standardized target query content and narrowing the semantic gap between user input and model understanding. Building upon this, intent mining processing of the target query content accurately captures the user's core needs, avoiding intent deviations or comprehension biases caused by directly constructing prompts based on the original input. Furthermore, by introducing model adaptation rules from the content processing model, and combining the identified user intent with the standardized target query content to construct prompts, the generated prompts can automatically adapt to the specific model used. Overall, this method achieves fully automated construction of professional prompts from the user's original query, requiring no professional prompting engineering knowledge from the user, significantly lowering the barrier to entry, while effectively improving the accuracy, relevance, and usability of the results generated by the content processing model.

[0104] refer to Figure 2 This is another flowchart of the method for constructing the prompt message in this application. The specific description is as follows: 201. Obtain the query content input by the user for the content processing model.

[0105] 202. Standardize the query content to obtain the standardized target query content.

[0106] 203. Perform intent mining processing on the target query content to obtain the user intent.

[0107] 204. Obtain a set of model rule mapping relationships, which includes a preset model and preset model adaptation rules. The preset model adaptation rules indicate the usage rules, output characteristics, and prompt construction standards of the corresponding model in a preset scenario.

[0108] 205. In the set of model rule mapping relationships, query the model adaptation rules of the content processing model.

[0109] 206. Extract keywords from the target query content, and determine usage scenario tags based on the keywords.

[0110] 207. Based on the usage scenario tags, the model adaptation rules of the content processing model, the user intent, and the target query content, construct prompt information. The prompt information is used to input the content processing model for analysis and reasoning to obtain the query results corresponding to the query content.

[0111] To better implement the above methods, embodiments of this application also provide a prompt information construction device, such as... Figure 3 As shown, the prompt information construction device may include an acquisition unit 301, a standardization unit 302, an intent mining unit 303, and a prompt construction unit 304, as follows: (1) Obtain unit 301; The retrieval unit is used to retrieve the query content input by the user for the content processing model.

[0112] (2) Standardized unit 302; The standardization unit is used to standardize the query content to obtain the standardized target query content.

[0113] Optionally, in some embodiments of this application, the standardization unit may include a semantic correction subunit, a logical reconstruction subunit, and a format standardization subunit, as follows: The semantic correction subunit is used to perform semantic correction processing on the query content to obtain the corrected query content; The logical reconstruction subunit is used to perform logical reconstruction processing on the corrected query content to obtain the reconstructed query content. The format standardization subunit is used to perform format standardization processing on the reconstructed query content to obtain the standardized target query content.

[0114] (3) Intent mining unit 303; The intent mining unit is used to perform intent mining processing on the target query content to obtain the user intent.

[0115] (4) Prompt to construct unit 304; The prompt construction unit is used to construct prompt information based on the model adaptation rules of the content processing model, the user intent, and the target query content. The prompt information is used to input the content processing model for analysis and reasoning to obtain the query results corresponding to the query content.

[0116] Optionally, in some embodiments of this application, the prompt construction unit may include a scene label determination subunit and a construction subunit, as follows: The scene tag determination subunit is used to extract keywords from the target query content and determine the usage scene tag based on the keywords. A sub-unit is constructed to build prompt information based on the usage scenario tags, the model adaptation rules of the content processing model, the user intent, and the target query content.

[0117] Optionally, in some embodiments of this application, the scene tag determination subunit may be specifically used to perform matching calculations between the keyword and candidate scene tags in the scene tag lexicon to obtain the matching degree between the keyword and the candidate scene tags; detect whether there are candidate scene tags in the scene tag lexicon with a matching degree higher than a preset threshold; if there are, determine the usage scene tag based on the candidate scene tags with a matching degree higher than the preset threshold; if there are no, output the candidate scene tags for the user to select and confirm, and use the scene tag confirmed by the user as the usage scene tag.

[0118] Optionally, in some embodiments of this application, the prompt construction unit may include a mapping set acquisition subunit, a query subunit, and a prompt construction subunit, as follows: The mapping set acquisition sub-unit is used to acquire the model rule mapping relationship set. The model rule mapping relationship set includes a preset model and a preset model adaptation rule. The preset model adaptation rule indicates the usage rules, output characteristics, and prompt construction standards of the corresponding model in a preset scenario. The query subunit is used to query the model adaptation rules of the content processing model in the model rule mapping relationship set; The prompt construction subunit is used to construct prompt information based on the model adaptation rules of the content processing model, the user intent, and the target query content.

[0119] Optionally, in some embodiments of this application, the standardization unit can be specifically used to standardize the query content by constructing a model through prompts, so as to obtain the standardized target query content; The prompt information construction device further includes a quality analysis unit, a feedback evaluation unit, and an update unit, as follows: The quality analysis unit is used to perform quality analysis on the query results output by the content processing model to obtain quality analysis results; The feedback evaluation unit is used to determine the feedback evaluation results based on the user's operation records and usage feedback during the query process; The update unit is used to update the prompt construction model based on the quality analysis results and the feedback evaluation results to obtain the updated prompt construction model.

[0120] Optionally, in some embodiments of this application, the updating unit may be specifically used to determine a comprehensive evaluation index for the prompt information based on the quality analysis results and the feedback evaluation results; construct a prompt sample based on the comprehensive evaluation index; and update the prompt construction model according to the prompt sample to obtain an updated prompt construction model.

[0121] Optionally, in some embodiments of this application, the prompt information construction device may further include a requirement parameter determination unit, a missing condition identification unit, and a completion unit, as follows: The requirement parameter determination unit is used to determine the structured requirement parameters of the target query content based on the user intent. The missing condition identification unit is used to identify implicit missing conditions in the structured requirement parameters based on the domain knowledge graph. The completion unit is used to perform the content completion operation on the target query content for the implicit missing conditions, so as to obtain the completed target query content.

[0122] Optionally, in some embodiments of this application, the completion unit may specifically be used to perform confidence assessment processing on the implicit missing conditions to obtain a confidence score; when the confidence score does not exceed a preset threshold, an interactive follow-up question is generated based on the implicit missing conditions and output, so that the user can supplement the content based on the interactive follow-up question, and perform a content completion operation on the target query content according to the supplemented content to obtain the completed target query content; when the confidence score exceeds a preset threshold, the content completion operation of the implicit missing conditions is performed on the target query content to obtain the completed target query content.

[0123] As can be seen from the above, in this embodiment, the acquisition unit 301 acquires the query content input by the user for the content processing model; the standardization unit 302 performs standardization processing on the query content to obtain the standardized target query content; the intent mining unit 303 performs intent mining processing on the target query content to obtain the user intent; and the prompt construction unit 304 constructs prompt information based on the model adaptation rules of the content processing model, the user intent, and the target query content. The prompt information is used to input into the content processing model for analysis and reasoning to obtain the query results corresponding to the query content.

[0124] This application standardizes the user's original query content, effectively eliminating issues such as colloquial expressions and logical inconsistencies, resulting in standardized target query content and narrowing the semantic gap between user input and model understanding. Building upon this, intent mining processing of the target query content accurately captures the user's core needs, avoiding intent deviations or comprehension biases caused by directly constructing prompts based on the original input. Furthermore, by introducing model adaptation rules from the content processing model, and combining the identified user intent with the standardized target query content to construct prompts, the generated prompts can automatically adapt to the specific model used. Overall, this method achieves fully automated construction of professional prompts from the user's original query, requiring no professional prompting engineering knowledge from the user, significantly lowering the barrier to entry, while effectively improving the accuracy, relevance, and usability of the results generated by the content processing model.

[0125] This application also provides an electronic device, such as... Figure 4 The diagram shows a structural schematic of an electronic device involved in an embodiment of this application. This electronic device can be a terminal or a server, specifically: The electronic device may include components such as a processor 401 with one or more processing cores, a memory 402 with one or more computer-readable storage media, a power supply 403, and an input unit 404. Those skilled in the art will understand that... Figure 4 The electronic device structure shown does not constitute a limitation on the electronic device and may include more or fewer components than shown, or combine certain components, or have different component arrangements. Wherein: The processor 401 is the control center of the electronic device, connecting various parts of the device via various interfaces and lines. It executes software programs and / or modules stored in the memory 402, and calls data stored in the memory 402, to perform various functions and process data. Optionally, the processor 401 may include one or more processing cores; preferably, the processor 401 may integrate an application processor and a modem processor, wherein the application processor mainly handles the operating system, user interface, and applications, and the modem processor mainly handles wireless communication. It is understood that the modem processor may not be integrated into the processor 401.

[0126] The memory 402 can be used to store software programs and modules. The processor 401 executes various functional applications and data processing by running the software programs and modules stored in the memory 402. The memory 402 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, application programs required for at least one function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the electronic device, etc. In addition, the memory 402 may include high-speed random access memory, and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other volatile solid-state storage device. Accordingly, the memory 402 may also include a memory controller to provide the processor 401 with access to the memory 402.

[0127] The electronic device also includes a power supply 403 that supplies power to the various components. Preferably, the power supply 403 can be logically connected to the processor 401 through a power management system, thereby enabling functions such as charging, discharging, and power consumption management through the power management system. The power supply 403 may also include one or more DC or AC power supplies, recharging systems, power fault detection circuits, power converters or inverters, power status indicators, and other arbitrary components.

[0128] The electronic device may also include an input unit 404, which can be used to receive input digital or character information, and generate keyboard, mouse, joystick, optical or trackball signal inputs related to user settings and function control.

[0129] Although not shown, the electronic device may also include a display unit, etc., which will not be described in detail here. Specifically, in this embodiment, the processor 401 in the electronic device loads the executable files corresponding to the processes of one or more applications into the memory 402 according to the following instructions, and the processor 401 runs the applications stored in the memory 402 to execute the steps in any of the prompt information construction methods provided in the embodiments of this application.

[0130] As can be seen from the above, this embodiment can obtain the query content input by the user for the content processing model; perform standardization processing on the query content to obtain the standardized target query content; perform intent mining processing on the target query content to obtain the user intent; and construct prompt information based on the model adaptation rules of the content processing model, the user intent, and the target query content. The prompt information is used to input the content processing model for analysis and reasoning to obtain the query results corresponding to the query content.

[0131] This application standardizes the user's original query content, effectively eliminating issues such as colloquial expressions and logical inconsistencies, resulting in standardized target query content and narrowing the semantic gap between user input and model understanding. Building upon this, intent mining processing of the target query content accurately captures the user's core needs, avoiding intent deviations or comprehension biases caused by directly constructing prompts based on the original input. Furthermore, by introducing model adaptation rules from the content processing model, and combining the identified user intent with the standardized target query content to construct prompts, the generated prompts can automatically adapt to the specific model used. Overall, this method achieves fully automated construction of professional prompts from the user's original query, requiring no professional prompting engineering knowledge from the user, significantly lowering the barrier to entry, while effectively improving the accuracy, relevance, and usability of the results generated by the content processing model.

[0132] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be performed by instructions, or by instructions controlling related hardware. These instructions can be stored in a computer-readable storage medium and loaded and executed by a processor.

[0133] Therefore, embodiments of this application provide a computer-readable storage medium storing a plurality of instructions that can be loaded by a processor to execute the steps in any of the prompt message construction methods provided in embodiments of this application.

[0134] The computer-readable storage medium may include: read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0135] Since the instructions stored in the computer-readable storage medium can execute the steps in any of the prompt message construction methods provided in the embodiments of this application, the beneficial effects that any of the prompt message construction methods provided in the embodiments of this application can achieve can be realized. For details, please refer to the previous embodiments, which will not be repeated here.

[0136] According to one aspect of this application, a computer program product is provided, which includes a computer program that, when executed by a processor, can implement the steps in the above-described prompt information construction method provided in the embodiments of this application.

[0137] The processor of the electronic device reads computer instructions from a computer-readable storage medium, executes the computer instructions, and causes the electronic device to perform the methods provided in the various alternative implementations of the above-mentioned prompt information construction.

[0138] The above description is merely a preferred embodiment of this application, but the scope of protection of this application is not limited thereto. It should be noted that, for those skilled in the art, several equivalent obvious modifications and / or equivalent substitutions can be made without departing from the technical principles of this application, and these obvious modifications and / or equivalent substitutions should also be considered within the scope of protection of this application.

Claims

1. A method for constructing prompt information, characterized in that, include: Obtain the query content input by the user for the content processing model; The query content is standardized to obtain the standardized target query content; The target query content is subjected to intent mining processing to obtain the user intent; Based on the model adaptation rules of the content processing model, the user intent, and the target query content, a prompt message is constructed. The prompt message is used to input the content processing model for analysis and reasoning to obtain the query results corresponding to the query content.

2. The method according to claim 1, characterized in that, The standardization process for the query content to obtain the standardized target query content includes: The query content is semantically corrected to obtain the corrected query content; The corrected query content is logically reconstructed to obtain the reconstructed query content; The reconstructed query content is then formatted to obtain the standardized target query content.

3. The method according to claim 1, characterized in that, The prompt information is constructed based on the model adaptation rules of the content processing model, the user intent, and the target query content, including: Extract keywords from the target query content, and determine usage scenario tags based on the keywords; Based on the usage scenario tags, the model adaptation rules of the content processing model, the user intent, and the target query content, a prompt message is constructed.

4. The method according to claim 3, characterized in that, The step of determining usage scenario tags based on the keywords includes: The matching degree between the keyword and the candidate scene tag in the scene tag thesaurus is calculated by matching the keyword with the candidate scene tag. Detect whether there are candidate scene tags in the scene tag lexicon with a matching degree higher than a preset threshold; If they exist, determine which scene tag to use based on candidate scene tags with a matching degree higher than a preset threshold; If no candidate scene tag exists, the candidate scene tag is output for the user to select and confirm, and the scene tag confirmed by the user is used as the usage scene tag.

5. The method according to claim 1, characterized in that, The prompt information is constructed based on the model adaptation rules of the content processing model, the user intent, and the target query content, including: Obtain a set of model rule mapping relationships, which includes a preset model and preset model adaptation rules. The preset model adaptation rules indicate the usage rules, output characteristics, and prompt construction standards of the corresponding model in a preset scenario. In the set of model rule mapping relationships, query the model adaptation rules of the content processing model; Based on the model adaptation rules of the content processing model, the user intent, and the target query content, a prompt message is constructed.

6. The method according to claim 1, characterized in that, The standardization process for the query content to obtain the standardized target query content includes: By constructing a model based on prompts, the query content is standardized to obtain the standardized target query content. The method further includes: The quality analysis results are obtained by performing quality analysis on the query results output by the content processing model. The feedback evaluation results are determined based on the user's operation records and usage feedback during the query process; Based on the quality analysis results and the feedback evaluation results, the prompt construction model is updated to obtain the updated prompt construction model.

7. The method according to claim 1, characterized in that, Before constructing the prompt information based on the model adaptation rules of the content processing model, the user intent, and the target query content, the method further includes: Based on the user intent, determine the structured requirement parameters for the target query content; Based on the domain knowledge graph, identify the implicit missing conditions in the structured requirement parameters; The target query content is then filled with the implicit missing conditions to obtain the filled target query content.

8. A device for constructing prompt information, characterized in that, include: The acquisition unit is used to acquire the query content input by the user for the content processing model; A standardization unit is used to standardize the query content to obtain the standardized target query content; The intent mining unit is used to perform intent mining processing on the target query content to obtain the user intent; The prompt construction unit is used to construct prompt information based on the model adaptation rules of the content processing model, the user intent, and the target query content. The prompt information is used to input the content processing model for analysis and reasoning to obtain the query results corresponding to the query content.

9. An electronic device, characterized in that, It includes a memory and a processor; the memory stores an application program, and the processor runs the application program within the memory to perform the operations in the prompt message construction method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a plurality of instructions adapted for loading by a processor to perform the steps of the prompt message construction method according to any one of claims 1 to 7.