A large model-based automatic pragmatics model construction method

Abstract

This application discloses an automatic pragmatic model construction method, apparatus, device, and storage medium based on a large model, belonging to the field of model generation. The method includes: acquiring a requirement description text for the target pragmatic model; extracting target knowledge information from a knowledge base based on the target information extraction domain represented by the requirement description text; and generating a target pragmatic model based on the target knowledge information for extracting information fields under the target information extraction domain. By combining knowledge base-driven semantic understanding with a model generation mechanism, this application not only simplifies the pragmatic model construction process but also improves the model's extraction capability and adaptability, reduces the cost of manual intervention, and enhances the system's processing efficiency and intelligence level in diverse semantic scenarios. It solves the problems of complex pragmatic model design processes, weak model extraction capabilities, and the inability to automate the process.

Description

Technical Field

[0001] This application belongs to the field of model generation, specifically relating to a method, apparatus, device, and storage medium for automatically constructing pragmatic models based on large models. Background Technology

[0002] With the rapid development of artificial intelligence technology, pragmatic models are playing an increasingly important role in natural language processing tasks such as information extraction, intelligent question answering, and semantic understanding. Users often want to describe their needs using natural language, and the system can automatically understand and construct corresponding pragmatic models to achieve accurate extraction and semantic mapping of specific information fields.

[0003] In existing technologies, the construction of pragmatic models typically relies on manually designing extraction rules or training on small-scale models. Specifically, developers must first analyze the semantic requirements of the target task, manually define the information field types and their extraction logic, and then train them using traditional machine learning or lightweight deep learning models to generate a pragmatic model with a certain extraction capability. Furthermore, some methods attempt to utilize predefined templates or domain dictionaries to assist model construction, thereby improving semantic coverage and extraction accuracy.

[0004] However, on the one hand, the manual definition of rules and templates is difficult to adapt to diverse and dynamically changing semantic needs, resulting in insufficient generalization ability of the model; on the other hand, training methods based on small-scale models perform poorly when faced with complex semantic structures, making it difficult to accurately identify and extract target information fields; finally, existing methods cannot achieve the automated generation of pragmatic models. Summary of the Invention

[0005] This application aims to provide an automatic pragmatic model construction method, apparatus, device, and storage medium based on large models, which at least solves the problems of complex pragmatic model design process, weak model extraction capability, and inability to automate existing pragmatic models.

[0006] In a first aspect, embodiments of this application disclose an automatic pragmatic model construction method based on a large model, including:

[0007] Obtain the text describing the requirements for the target pragmatic model;

[0008] Based on the target information extraction domain represented by the requirement description text, target knowledge information is extracted from the knowledge base; the target knowledge information is used to represent multiple information field types under the target information extraction domain;

[0009] Based on the target knowledge information, a target pragmatic model is generated for extracting information fields in the target information extraction domain.

[0010] Secondly, embodiments of this application also disclose an automatic pragmatic model construction apparatus based on a large model, characterized in that it includes:

[0011] The requirement elicitation module is used to obtain the requirement description text for the target pragmatic model;

[0012] The knowledge extraction module is used to extract target knowledge information from the knowledge base according to the target information extraction domain represented by the requirement description text; the target knowledge information is used to represent multiple information field types under the target information extraction domain;

[0013] The model generation module is used to generate a target pragmatic model based on the target knowledge information for extracting information fields in the target information extraction domain.

[0014] Thirdly, embodiments of this application also disclose an electronic device, including a processor and a memory, wherein the memory stores a program or instructions that can run on the processor, and the program or instructions, when executed by the processor, implement the steps of the method described in the first aspect.

[0015] Fourthly, embodiments of this application also disclose a readable storage medium storing a program or instructions that, when executed by a processor, implement the steps of the method described in the first aspect.

[0016] In summary, in this embodiment, by acquiring the requirement description text for the target pragmatic model, the system can accurately understand the user's semantic intent, thereby avoiding the manual analysis of semantic requirements in traditional methods, lowering the development threshold and improving the flexibility of model construction. This lays a semantic foundation for subsequent knowledge extraction and model generation, contributing to improved overall automation. Furthermore, by mapping the requirement description text to the target information extraction domain and extracting corresponding target knowledge information from the knowledge base, multiple information field types in that domain can be automatically identified and summarized. This effectively utilizes structured knowledge resources, enhances the model's ability to understand complex semantic structures, and thus improves the accuracy and coverage of information extraction. Finally, based on the extracted target knowledge information, a pragmatic model is generated, enabling the model to accurately extract information fields for specific domains. This improves the model's generalization ability and extraction performance, avoiding the poor extraction results caused by fixed rule templates or limited model size in traditional methods. Therefore, the method based on the embodiments of this application, by combining knowledge base-driven semantic understanding and model generation mechanisms, not only simplifies the pragmatic model construction process, but also improves the model's extraction capability and adaptability, reduces the cost of manual intervention, enhances the system's processing efficiency and intelligence level in diverse semantic scenarios, and solves the problems of complex pragmatic model design process, weak model extraction capability, and inability to automate. Attached Figure Description

[0017] In the attached diagram:

[0018] Figure 1 This is a flowchart illustrating the steps of an automatic pragmatic model construction method based on a large model, as provided in an embodiment of this application.

[0019] Figure 2 This is a flowchart of another method for constructing an automatic pragmatic model based on a large model, provided in an embodiment of this application.

[0020] Figure 3 This is a domain example of a pragmatic model based on embodiments of this application;

[0021] Figure 4 A complete pragmatic model generation process based on the embodiments of this application.

[0022] Figure 5 This is a block diagram of an automatic pragmatic model construction device based on a large model provided in an embodiment of this application;

[0023] Figure 6 This is a block diagram of an electronic device provided in one embodiment of this application. Detailed Implementation

[0024] 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, 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.

[0025] The terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such use of data can be interchanged where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class and the number of objects is not limited; for example, a first object can be one or more. Furthermore, in the specification and claims, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0026] like Figure 1 The image shows an automatic pragmatic model construction method based on a large model provided in an embodiment of this application.

[0027] The method may include the following steps:

[0028] Step 101: Obtain the requirement description text for the target pragmatic model.

[0029] In some embodiments of this application, to enable the system to understand the user's needs for constructing a pragmatic model, thereby providing a semantic foundation for subsequent information extraction and model generation, a demand description text for the target pragmatic model is acquired. This text typically expresses the content or semantic goal that the user wants the system to extract in natural language, such as "Please identify the payment terms in the contract" or "Extract diagnostic information from the medical record." Here, the "demand description text" refers to the natural language content used to express the information extraction target, possessing semantic features such as context, intent, and linguistic behavior. By acquiring this text, the system can establish a preliminary understanding of the user's semantic intent, thereby avoiding the process of manually analyzing semantic needs, reducing the complexity of model construction, and improving automated processing capabilities.

[0030] In a specific example, a user wants to build a pragmatic model to identify diagnostic fields in medical records. The user's input description text is "Please extract the diagnostic results and related symptoms from the medical records." Upon receiving this text, the system parses it as semantic input and identifies "diagnostic results" and "related symptoms" as the semantic references to the target information fields. This allows the system to obtain the semantic foundation for subsequent knowledge extraction and model generation, enabling it to extract diagnosis-related knowledge information in later steps and build a pragmatic model with extraction capabilities.

[0031] Step 102: Extract target knowledge information from the knowledge base based on the target information extraction domain represented by the requirement description text.

[0032] Among them, target knowledge content is used to represent multiple information field types in the target information extraction domain.

[0033] In some embodiments of this application, to enable the system to retrieve structured knowledge related to the target information extraction domain from a knowledge base based on the user's semantic needs, thereby providing semantic support for the generation of pragmatic models, target knowledge information is extracted from the knowledge base based on the target information extraction domain represented by the demand description text. This process relies on the semantic parsing results of the demand text, mapping it to domain tags or concept nodes in the knowledge base to locate knowledge entries related to that domain. Target knowledge information refers to structured content used to define and describe multiple information field types within a specific information extraction domain, such as field names, field attributes, and logical relationships between fields. In this way, the system can establish an information structure framework within the domain, thereby enhancing the subsequent pragmatic model's ability to understand complex semantic structures and improving the accuracy and coverage of information extraction.

[0034] In a specific example, the user inputs the requirement description text "Please extract the payment terms and liability for breach of contract from the contract." The system identifies "contract" as the target information extraction domain and locates the relevant entry for "contract law" in the knowledge base. Subsequently, the system extracts field types related to "payment terms" and "liability for breach of contract" from this entry, such as "payment time," "payment method," and "method of calculating liquidated damages." In this way, the system obtains target knowledge information to represent the information structure of the contract domain, providing a structured semantic foundation for subsequent pragmatic model generation.

[0035] Step 103: Generate a target pragmatic model based on the target knowledge information to extract information fields in the target information extraction domain.

[0036] In some embodiments of this application, in order for the system to construct a pragmatic model with extraction capabilities based on extracted structured knowledge information to complete the task of identifying and extracting target information fields, it is necessary to generate a target pragmatic model for extracting information fields in the target information extraction domain based on the target knowledge information. This process typically involves calling a large model with semantic modeling capabilities, combining the field types, field attributes, and their semantic relationships in the target knowledge information, to generate a domain-adaptive extraction model. A pragmatic model refers to a semantic structure model that can automatically identify and extract specific information fields based on input text, possessing capabilities such as semantic understanding, field location, and extraction rule reasoning. By generating this model, the system can automatically process target domain text in subsequent applications, thereby improving the accuracy and generalization ability of extraction and reducing the cost of manual intervention.

[0037] In a specific example, the system has extracted target knowledge information in the "contract" domain from the knowledge base, including field types such as "payment time," "payment method," and "liability calculation method." The system then calls a Large Language Model (LLM), using these field types as semantic constraints as input, and trains or fine-tunes them in conjunction with contract text corpora to generate a pragmatic model capable of recognizing the aforementioned fields in contract text. Ultimately, the system obtains a pragmatic model that can be used for automatic contract text extraction. This model can identify payment terms and liability fields in contracts in practical applications, thereby automating the information extraction task.

[0038] In summary, in this embodiment, by acquiring the requirement description text for the target pragmatic model, the system can accurately understand the user's semantic intent, thereby avoiding the manual analysis of semantic requirements in traditional methods, lowering the development threshold and improving the flexibility of model construction. This lays a semantic foundation for subsequent knowledge extraction and model generation, contributing to improved overall automation. Furthermore, by mapping the requirement description text to the target information extraction domain and extracting corresponding target knowledge information from the knowledge base, multiple information field types in that domain can be automatically identified and summarized. This effectively utilizes structured knowledge resources, enhances the model's ability to understand complex semantic structures, and thus improves the accuracy and coverage of information extraction. Finally, based on the extracted target knowledge information, a pragmatic model is generated, enabling the model to accurately extract information fields for specific domains. This improves the model's generalization ability and extraction performance, avoiding the poor extraction results caused by fixed rule templates or limited model size in traditional methods. Therefore, the method based on the embodiments of this application, by combining knowledge base-driven semantic understanding and model generation mechanisms, not only simplifies the pragmatic model construction process, but also improves the model's extraction capability and adaptability, reduces the cost of manual intervention, enhances the system's processing efficiency and intelligence level in diverse semantic scenarios, and solves the problems of complex pragmatic model design process, weak model extraction capability, and inability to automate.

[0039] Figure 2 This is another method for constructing an automatic pragmatic model based on a large model, provided in the embodiments of this application.

[0040] The method may include the following steps:

[0041] Step 201: Obtain the requirement description text for the target pragmatic model.

[0042] The method shown in this step has been explained in step 101 and will not be repeated here.

[0043] Step 202: Extract multiple pieces of information from the knowledge base based on the requirement description text.

[0044] In some embodiments of this application, to enable the system to retrieve multiple information resources related to user needs from the knowledge base, providing a material foundation for subsequent semantic filtering and model construction, multiple pieces of information knowledge are extracted from the knowledge base based on the demand description text. This process typically associates the demand text with entries in the knowledge base through semantic matching, keyword expansion, or domain tag mapping, thereby obtaining a broader knowledge set. Information knowledge refers to semantic content stored in the knowledge base in a structured or semi-structured form, which may include domain terms, field definitions, hierarchical relationships, typical expressions, etc. By extracting multiple pieces of information knowledge, the system can establish a preliminary semantic candidate set, providing the necessary corpus foundation for subsequent semantic information-driven target knowledge filtering.

[0045] In a specific example, the user inputs the request text "Please extract the diagnosis results and related symptoms from the medical records." The system identifies "medical records" as the target domain and retrieves entries related to "diagnosis results" and "symptoms" from the medical knowledge base. The extracted information may include multiple entries such as "disease name," "clinical manifestations," "diagnostic criteria," and "symptom description," with each entry containing field definitions, typical terms, and their semantic relationships. In this way, the system obtains a knowledge set covering diagnostic semantics, providing a foundational resource for subsequently filtering out the target knowledge information that best matches the user's intent.

[0046] Optionally, step 202 includes the following sub-steps:

[0047] Sub-step 2021 involves performing similarity matching between the semantic vector of the requirement description text and the semantic vector of each piece of information knowledge in the knowledge base.

[0048] In some embodiments of this application, to enable the system to determine the semantic relevance between the requirement description text and the information knowledge in the knowledge base through semantic-level quantitative representation, thereby providing a matching basis for subsequent information filtering, the semantic vector of the requirement description text is matched with the semantic vector of each piece of information knowledge in the knowledge base based on similarity. This process typically involves converting the text content into semantic vectors using a vectorization model and then using vector similarity calculation methods (such as cosine similarity or Euclidean distance) for matching evaluation. A semantic vector is a numerical representation that maps natural language text to a high-dimensional semantic space, preserving the semantic features and contextual relationships of the text. In this way, the system can establish a ranking of the semantic relevance between the requirement text and knowledge items, providing a foundation for subsequently extracting information knowledge that meets semantic conditions.

[0049] In a specific example, the user inputs the requirement description text "Please extract the payment terms from the contract." The system converts this text into a semantic vector and performs similarity matching on each semantic vector in the knowledge base containing entries such as "payment time," "payment method," and "penalty calculation." During the matching process, the system can use the cosine similarity calculation formula. Then, the system obtains the semantic matching degree between each piece of information and the requirement text, providing a quantitative basis for subsequently filtering information that meets the preset similarity conditions.

[0050] Sub-step 2022 involves extracting the information and knowledge corresponding to semantic vectors in the knowledge base that have a matching similarity with the semantic vectors of the requirement description text that meet the preset similarity conditions.

[0051] In some embodiments of this application, to enable the system to filter out information knowledge that meets semantic relevance requirements from the knowledge set that has undergone semantic matching, thereby constructing semantically targeted knowledge input, the information knowledge corresponding to semantic vectors in the knowledge base whose matching similarity with the semantic vectors of the requirement description text reaches a preset similarity condition is extracted. This process typically filters the semantic matching degrees calculated in previous steps by setting a similarity threshold, retaining those knowledge items that have a high semantic consistency with the requirement description text. The preset similarity condition can be adjusted according to task complexity or domain requirements, for example, setting a cosine similarity of not less than 0.85, and selecting the top 10 items. Information knowledge refers to structured items associated with semantic vectors, including field definitions, semantic tags, hierarchical relationships, etc. In this way, the system can construct a knowledge set that is highly relevant to user needs, providing semantic support for the subsequent extraction of target knowledge information and the generation of pragmatic models.

[0052] In a specific example, the user inputs the requirement description text "Please extract the payment terms from the contract." In previous steps, the system matched the semantic vector of this text with the semantic vectors of multiple entries in the knowledge base and calculated the similarity value. The system sets a preset similarity threshold of 0.85 and filters out entries with a matching degree reaching or exceeding this threshold, including "payment time," "payment method," and "payment cycle." The system then extracts the information knowledge corresponding to these entries, including field names, typical expressions, and their semantic tags. Ultimately, the system obtains a set of information knowledge highly consistent with the semantics of the requirement text, providing structured semantic input for subsequently building a pragmatic model for the payment terms.

[0053] Step 203: Extract target information knowledge from multiple pieces of information knowledge based on the semantic information represented by the requirement description text to form target knowledge information.

[0054] Among them, target knowledge information is used to characterize multiple information field types in the target information extraction domain; semantic information includes at least one of the following: the direct textual intent or predictive textual intent of the demand description text, textual context, and textual language behavior information.

[0055] In some embodiments of this application, in order for the system to filter out the content that best matches the user's semantic needs from multiple pieces of acquired information knowledge, thereby forming target knowledge information with extraction guidance significance, the system extracts target information knowledge from multiple pieces of information knowledge based on the semantic information represented by the demand description text. This process relies on modeling and parsing the semantic information of the demand text. The semantic information includes at least one of the following: direct intent of the text (i.e., the extraction target explicitly expressed by the user), predicted intent of the text (i.e., the potential demand inferred by the system based on the context), text context (i.e., the semantic environment in which the text exists, such as domain background or contextual relationships), and text linguistic behavior information (i.e., the type of language behavior embodied in the text, such as requests, statements, instructions, etc.). By identifying these semantic features, the system can perform semantic matching and filtering on multiple pieces of information knowledge, extracting the knowledge content most relevant to the user's intent. Target knowledge information refers to a structured semantic set used to represent multiple information field types within the target information extraction domain, possessing domain adaptability and extraction guidance. The execution of this step helps improve the semantic accuracy and domain coverage of the pragmatic model, providing semantic support for subsequent model generation.

[0056] In a specific example, the user inputs the request description text "Please extract the diagnosis results and related symptoms from the medical records." In previous steps, the system has extracted multiple pieces of information from the medical knowledge base, including entries such as "disease name," "clinical manifestations," "diagnostic criteria," "treatment plan," and "symptom description." The system further analyzes the semantic information of the text, identifying the direct intent as "diagnosis results" and "related symptoms," the text context as "medical records," and the language behavior type as "request." Based on these semantic features, the system filters entries related to "diagnosis results" and "symptom description" from the multiple pieces of information, forming target knowledge information, including field types such as "primary diagnosis," "secondary diagnosis," "typical symptoms," and "accompanying symptoms." Ultimately, the system obtains semantically targeted target knowledge information, providing a structured semantic foundation for the subsequent construction of a pragmatic model.

[0057] Optionally, step 203 includes the following sub-steps:

[0058] Sub-step 2031: When the semantic information is represented as containing the intention to extract macro information, the data of the first field type in the multiple pieces of information knowledge is identified as the target information knowledge.

[0059] The first field type is used to represent the semantic boundary conditions of the requirement description text.

[0060] In some embodiments of this application, to enable the system to identify and determine the field types used to represent semantic boundary conditions in semantic scenarios where the user expresses a macro-level extraction intent, thereby constructing target knowledge information with abstract semantic coverage, when the semantic information is represented as containing a macro-level information extraction intent, data of the first field type among multiple pieces of information knowledge is identified as target information knowledge. This process relies on the classification and judgment of semantic information. The macro-level information extraction intent is usually reflected in the user's desire to obtain overall, background, or identity information, rather than specific details or behavioral content. The first field type refers to the set of fields used to describe semantic boundary conditions, such as identity attributes, role tags, or basic background information, and its role is to limit the semantic applicability of the pragmatic model. By performing this step, the system can construct target knowledge information covering macro-level semantics, providing boundary constraints and a semantic framework for subsequent model generation.

[0061] In a specific example, such as Figure 3 As shown, the user-input requirement text is "Please extract basic information from the resume." The system parses the semantic information of this text, determining that the intent is macro-level information extraction, with the goal of obtaining the applicant's identity and background information. The system identifies the first field type related to this intent in the knowledge base, including entries such as "Name," "Applied Position," "Work Experience," and "Skills." The system then determines the information knowledge corresponding to these fields as target knowledge information and uses it for the subsequent construction of the pragmatic model. After completing this execution process, the system obtains a knowledge set to represent semantic boundary conditions, enabling the generated pragmatic model to extract basic information from the resume.

[0062] Sub-step 2032: When the semantic information is represented as containing a specific information extraction intent, the data of the second field type in multiple pieces of information knowledge is identified as the target information knowledge.

[0063] The second field type is used to represent the information extraction conditions of the requirement description text.

[0064] In some embodiments of this application, to enable the system to identify and determine the field types used to characterize the conditions for information extraction content in semantic scenarios where the user expresses a specific extraction intent, thereby constructing target knowledge information with fine-grained semantic coverage, data of the second field type among multiple pieces of information knowledge will be identified as target information knowledge when the semantic information is characterized as containing a specific information extraction intent. This process relies on the classification and judgment of semantic information. The specific information extraction intent is usually reflected in the user's desire to obtain a certain type of detailed content, attribute, or behavioral information. The second field type refers to a set of fields used to describe the conditions for information extraction content, such as educational background, transaction details, diagnostic results, etc., and its role is to limit the extraction target and semantic focus of the pragmatic model. In this way, the system can construct target knowledge information that covers specific semantic goals, providing extraction directions and field definitions for the subsequent generation of pragmatic models.

[0065] In a specific example, such as Figure 3 As shown, the user-input request text is "Please extract educational background information from the resume." The system parses the semantic information of this text, determining that the intent is specific information extraction, with the goal of obtaining content related to the applicant's educational experience. The system identifies a second field type related to this intent in the knowledge base, including entries such as "educational background," "degree major," "graduating institution," and "graduation year." The system then identifies the information knowledge corresponding to these fields as target knowledge information and uses it for the subsequent construction of the pragmatic model. After completing this execution process, the system obtains a knowledge set representing the conditions for information extraction, enabling the generated pragmatic model to extract educational background information.

[0066] Step 204: Generate a target pragmatic model based on the target knowledge information to extract information fields in the target information extraction domain.

[0067] The method shown in this step has been explained in step 103 and will not be repeated here.

[0068] Optionally, based on sub-steps 2031 and 2032, step 204 includes the following sub-steps:

[0069] Sub-step 2041: When the semantic information is represented as containing the intention to extract macro information, the output data of the initial pragmatic model is limited according to the first field type data extracted from the target knowledge information to obtain the first target pragmatic model.

[0070] The first field type is used to represent the semantic boundary conditions of the requirement description text.

[0071] In some embodiments of this application, to enable the system to limit the output range of the initial pragmatic model based on the field type corresponding to the macro-semantic intent, thereby generating a first target pragmatic model with semantic boundary constraints, the output data of the initial pragmatic model is limited according to the first field type data extracted from the target knowledge information, given that the semantic information is represented as containing the macro-information extraction intent, to obtain the first target pragmatic model. This process typically uses the first field type as a semantic filtering condition to filter out output fields in the initial pragmatic model that are irrelevant to the macro-intention, retaining only the set of fields used to represent semantic boundary conditions. The first field type refers to fields in the target knowledge information used to describe the extraction range boundary, such as identity attributes, role tags, or background information. In this way, the system generates a pragmatic model with extraction range control capabilities, thereby improving the extraction accuracy and semantic consistency of the model in macro-semantic scenarios.

[0072] In a specific example, the user inputs the requirement description text "Please extract basic information from the resume." The system identifies the semantic information of this text as a macro-level information extraction intent, and has already extracted the first field types in the previous steps, including "Name," "Applied Position," "Work Experience," and "Skills." The system uses these field types as limiting conditions to filter the output fields of the initial pragmatic model, retaining only the extraction logic and output structure corresponding to the aforementioned fields. After completing this execution process, the system obtains the first target pragmatic model, which has the ability to extract basic information from a resume and can output field content consistent with the semantic boundary conditions in practical applications.

[0073] Sub-step 2042: When the semantic information is represented as containing the intention to extract specific information, the output data of the initial pragmatic model is limited according to the second field type data extracted from the target knowledge information to obtain the second target pragmatic model.

[0074] The second field type is used to represent the information extraction conditions of the requirement description text.

[0075] In some embodiments of this application, to enable the system to limit the output range of the initial pragmatic model based on the field type corresponding to the specific semantic intent, thereby generating a second target pragmatic model with extraction directionality, when the semantic information is represented as containing a specific information extraction intent, the output data of the initial pragmatic model is limited based on the second field type data extracted from the target knowledge information to obtain the second target pragmatic model. This process typically uses the second field type as a semantic filtering condition to remove output fields in the initial pragmatic model that are irrelevant to the specific extraction target, retaining only the set of fields used to represent the conditions for information extraction content. The second field type refers to the field in the target knowledge information used to describe the fine-grained extraction target, such as attribute values, behavioral content, or structured details, and its role is to limit the extraction focus and semantic depth of the pragmatic model. In this way, the system will generate a pragmatic model with extraction precision control capabilities, thereby improving the extraction accuracy and semantic consistency of the model in specific semantic scenarios.

[0076] In a specific example, the user inputs the requirement text "Please extract educational background information from resumes." The system identifies the semantic information of this text as a specific information extraction intent and, in previous steps, has extracted the second field type, including "education level," "degree major," "graduating institution," and "graduation year." The system uses these field types as limiting conditions to filter the output fields of the initial pragmatic model, retaining only the extraction logic and output structure corresponding to the aforementioned fields. Ultimately, the system obtains a second target pragmatic model, which is capable of extracting educational background information from resumes and, in practical applications, can output field content consistent with the information extraction conditions.

[0077] Optionally, in some embodiments of this application, the first target pragmatic model generated in sub-step 2041 can further assist in the generation of the second target pragmatic model. In this case, the automatic pragmatic model construction method based on a large model of this application further includes the following steps:

[0078] Step 205: Extract the first information field of the target data based on the first target pragmatic model generated by the first requirement description text.

[0079] In some embodiments of this application, in order for the system to utilize the generated pragmatic model to perform semantic parsing on the target data, thereby identifying the first batch of information fields related to user needs, the first information fields of the target data are extracted based on the first target pragmatic model generated from the first requirement description text. This process relies on the semantic recognition capability of the pragmatic model, which performs field location and semantic matching on the target data according to its internal structure. A pragmatic model refers to a structured model with semantic understanding and field extraction capabilities; its construction process has been completed in previous steps, and it possesses the ability to extract data based on specific semantic boundary conditions. Thus, the system can extract information fields corresponding to the first requirement description text from the original data, providing a semantic foundation and extraction samples for subsequent requirement generation and model construction.

[0080] In a specific example, such as Figure 3 As shown, the user wants to build a pragmatic model from resume data to identify basic information about job applicants. The system generates a first target pragmatic model based on the semantic boundary condition of "basic information about job applicants" and applies it to extract information from the resume data. This resume data contains multiple fields, such as name, applied position, work experience, and skills. The system uses this pragmatic model to perform semantic parsing on the resume text, identifying fields such as "Name: Zhang San," "Applied Position: Data Analyst," and "Skills: Python, SQL." Thus, the system obtains the first set of information fields.

[0081] Step 206: Determine the second requirement description text for generating the second target pragmatic model based on the first information field and the target data.

[0082] In some embodiments of this application, in order for the system to deduce new semantic requirements based on the extracted information fields and the original data content, thereby generating text input for constructing a second pragmatic model, a second requirement description text for generating the second target pragmatic model is determined based on the first information fields and the target data. This process typically involves analyzing the relationship between the semantic scope covered by the first information fields and the content in the target data that has not yet been extracted, identifying potential extraction targets, and converting them into a requirement expression in natural language form. The second requirement description text refers to the semantic input generated based on the first semantic extraction, used to guide the construction of the second pragmatic model; its content should reflect new information extraction intentions or supplementary semantic goals. In this way, the system can achieve dynamic expansion and hierarchical expression of semantic requirements, providing a semantically driven input foundation for subsequent model generation.

[0083] In a specific example, such as Figure 3As shown, the system has extracted primary information fields such as "Name: Zhang San," "Applied Position: Data Analyst," and "Skills: Python, SQL" from the resume data using the first target pragmatic model. Further relying on the "resume dataset pragmatic model," the system identifies "applicant's educational background" as an unexplored area for extraction. The system analyzes educational background-related fields in the target data, such as "degree," "university," and "graduation year," and generates a second requirement description text, such as "Please extract the applicant's educational background information, including degree, major, and graduation date." In this way, the system obtains semantic input text for constructing the second target pragmatic model, thereby achieving hierarchical extraction and semantic completion of the resume data.

[0084] Optionally, between steps 205 and 206, iterative adjustments to the first target pragmatic model can be further added to enhance the information extraction accuracy of the two target pragmatic models. In this case, the automatic pragmatic model construction method based on the large model also includes the following steps:

[0085] Step 207: Update the first target pragmatic model based on the correction of the first requirement description text in the first information field, and return to step 205.

[0086] In some embodiments of this application, in order to enable the system to modify the original semantic requirements based on the extracted information fields, thereby iteratively updating the first target pragmatic model to improve its extraction accuracy and semantic adaptability, the first target pragmatic model is updated based on the modification of the first requirement description text according to the first information fields, and the process returns to the previous step of extracting the first information fields of the target data based on the first target pragmatic model generated from the first requirement description text. This process typically involves semantic analysis of the first information fields to identify ambiguities, omissions, or extraction biases in the original requirement expression, and adjusting the content of the requirement description text accordingly, such as supplementing the field range, clarifying semantic boundaries, or modifying the extraction target. Subsequently, the system regenerates or fine-tunes the first target pragmatic model based on the modified text to make it more consistent with the actual extraction task. The pragmatic model update process may rely on a Large Language Model (LLM) for parameter adjustment or semantic reconstruction to enhance the model's ability to extract target data. In this way, the system will be able to dynamically optimize the pragmatic model, improve the accuracy and completeness of subsequent extraction results, and provide a more reliable semantic foundation for the construction of the second pragmatic model.

[0087] In a specific example, such as Figure 3As shown, the system has extracted information fields such as "Name: Zhang San," "Applied Position: Data Analyst," and "Skills: Python, SQL" from the resume data using the first target pragmatic model. System analysis revealed that the "Work Experience" field was not accurately identified because the original job description text did not explicitly require its extraction. Relying on the "Resume Dataset Pragmatic Model," the system added "Work Experience" as a missing field to the first job description text, forming a corrected semantic input, such as "Please extract the applicant's basic information, including name, position, skills, and work experience." The system then updated the first target pragmatic model to enable it to recognize the "Work Experience" field. The system then re-executed the extraction steps, obtaining a complete set of information fields including "Work Experience," providing a more comprehensive semantic context for the subsequent generation of the second job description text.

[0088] like Figure 4 The diagram illustrates a complete pragmatic model generation process based on embodiments of this application. It relies on a domain determiner and a pragmatic model generator to ultimately generate a two-tiered pragmatic model:

[0089] Broad pragmatic model (S G ): Used to cover the broad domain to which the user intent belongs, providing the semantic boundaries of the extraction task;

[0090] Specific pragmatic model (S S ): Used to refine the extraction target and focus on the specific information content that users care about.

[0091] Users express their information extraction needs using natural language. The system first receives this input and passes it to a domain determiner to understand the user's semantic intent and determine which semantic domain the need belongs to, such as contract analysis, resume processing, or medical record extraction. Once the domain is determined, the system retrieves relevant knowledge resources from its content database. These resources include field definitions, semantic tags, and hierarchical relationships, providing semantic material for the subsequent pragmatic model generator's pragmatic model construction. Next, the system processes data based on two types of conditions: semantic boundary conditions, which limit the extraction scope; and content extraction conditions, which focus on specific extraction targets. These two conditions act on the pragmatic inference model, enabling it to generate a two-tiered pragmatic model: a broad one, covering the extraction boundaries of the entire domain; and a specific one, precisely identifying the information fields that the user ultimately cares about.

[0092] Ultimately, the system outputs these reasoning results as a set of structured fields, forming an executable extraction template. This process not only achieves automatic conversion from natural language to a structured model but also ensures the semantic consistency and accuracy of the extraction task. The entire process embodies the close collaboration between semantic understanding, knowledge-driven approaches, and model generation, representing a typical paradigm for building pragmatic intelligence systems.

[0093] In summary, in this embodiment, by acquiring the requirement description text for the target pragmatic model, the system can accurately understand the user's semantic intent, thereby avoiding the manual analysis of semantic requirements in traditional methods, lowering the development threshold and improving the flexibility of model construction. This lays a semantic foundation for subsequent knowledge extraction and model generation, contributing to improved overall automation. Furthermore, by mapping the requirement description text to the target information extraction domain and extracting corresponding target knowledge information from the knowledge base, multiple information field types in that domain can be automatically identified and summarized. This effectively utilizes structured knowledge resources, enhances the model's ability to understand complex semantic structures, and thus improves the accuracy and coverage of information extraction. Finally, based on the extracted target knowledge information, a pragmatic model is generated, enabling the model to accurately extract information fields for specific domains. This improves the model's generalization ability and extraction performance, avoiding the poor extraction results caused by fixed rule templates or limited model size in traditional methods. Therefore, the method based on the embodiments of this application, by combining knowledge base-driven semantic understanding and model generation mechanisms, not only simplifies the pragmatic model construction process, but also improves the model's extraction capability and adaptability, reduces the cost of manual intervention, enhances the system's processing efficiency and intelligence level in diverse semantic scenarios, and solves the problems of complex pragmatic model design process, weak model extraction capability, and inability to automate.

[0094] refer to Figure 5 It illustrates an automatic pragmatic model construction apparatus 30 based on a large model provided in an embodiment of this application, comprising:

[0095] The requirement acquisition module 301 is used to acquire the requirement description text for the target pragmatic model;

[0096] The knowledge extraction module 302 is used to extract target knowledge information from the knowledge base according to the target information extraction domain represented by the requirement description text; the target knowledge information is used to represent multiple information field types under the target information extraction domain;

[0097] The model generation module 303 is used to generate a target pragmatic model based on target knowledge information for extracting information fields in the target information extraction domain.

[0098] Optionally, the knowledge extraction module 302 includes:

[0099] The extraction submodule is used to extract multiple pieces of information from the knowledge base based on the required description text.

[0100] The selection submodule is used to extract target information knowledge from multiple pieces of information knowledge based on the semantic information represented by the requirement description text, so as to form target knowledge information; the semantic information includes at least one of the textual direct intent or textual predictive intent of the requirement description text, textual context, and textual linguistic behavior information.

[0101] Optionally, the extraction submodules include:

[0102] The matching unit is used to perform similarity matching between the semantic vector of the requirement description text and the semantic vector of each piece of information knowledge in the knowledge base;

[0103] The extraction unit is used to extract the information and knowledge corresponding to the semantic vectors in the knowledge base that match the semantic vectors of the requirement description text and reach the preset similarity conditions.

[0104] Optionally, the selected sub-modules include:

[0105] A broad selection unit is used to identify the data of the first field type among multiple pieces of information knowledge as the target information knowledge when the semantic information is represented as containing the intention to extract macro information; the first field type is used to represent the semantic boundary conditions of the requirement description text.

[0106] The specific selection unit is used to determine the data of the second field type in multiple pieces of information knowledge as the target information knowledge when the semantic information is represented as containing the intention to extract specific information; the second field type is used to represent the information extraction content conditions of the requirement description text.

[0107] Optionally, the model generation module 303 includes:

[0108] The broad model generation submodule is used to limit the output data of the initial pragmatic model based on the first field type data extracted from the target knowledge information, when the semantic information is represented as containing the intention to extract macro information, so as to obtain the first target pragmatic model; the first field type is used to represent the semantic boundary conditions of the demand description text.

[0109] The specific model generation submodule is used to limit the output data of the initial pragmatic model based on the second field type data extracted from the target knowledge information, when the semantic information is represented as containing the intention to extract specific information, so as to obtain the second target pragmatic model; the second field type is used to represent the information extraction content conditions of the demand description text.

[0110] Optionally, the automatic pragmatic model building apparatus 30 based on large models also includes:

[0111] The first information field module is used to extract the first information field of the target data based on the first target pragmatic model generated by the first requirement description text.

[0112] The second requirement description module is used to determine the second requirement description text for generating the second target pragmatic model based on the first information field and the target data.

[0113] Optionally, the automatic pragmatic model building apparatus 30 based on large models also includes:

[0114] The first requirement description module is used to update the first target pragmatic model based on the correction of the first requirement description text to the first information field, and return to the step of extracting the first information field of the target data based on the first target pragmatic model generated by the first requirement description text.

[0115] In summary, in this embodiment, by acquiring the requirement description text for the target pragmatic model, the system can accurately understand the user's semantic intent, thereby avoiding the manual analysis of semantic requirements in traditional methods, lowering the development threshold and improving the flexibility of model construction. This lays a semantic foundation for subsequent knowledge extraction and model generation, contributing to improved overall automation. Furthermore, by mapping the requirement description text to the target information extraction domain and extracting corresponding target knowledge information from the knowledge base, multiple information field types in that domain can be automatically identified and summarized. This effectively utilizes structured knowledge resources, enhances the model's ability to understand complex semantic structures, and thus improves the accuracy and coverage of information extraction. Finally, based on the extracted target knowledge information, a pragmatic model is generated, enabling the model to accurately extract information fields for specific domains. This improves the model's generalization ability and extraction performance, avoiding the poor extraction results caused by fixed rule templates or limited model size in traditional methods. Therefore, the method based on the embodiments of this application, by combining knowledge base-driven semantic understanding and model generation mechanisms, not only simplifies the pragmatic model construction process, but also improves the model's extraction capability and adaptability, reduces the cost of manual intervention, enhances the system's processing efficiency and intelligence level in diverse semantic scenarios, and solves the problems of complex pragmatic model design process, weak model extraction capability, and inability to automate.

[0116] Reference Figure 6 The electronic device 500 may include one or more of the following components: processing component 502, memory 504, power supply component 506, multimedia component 508, audio component 510, input / output (I / O) interface 512, sensor component 514, and communication component 516.

[0117] Processing component 502 typically controls the overall operation of electronic device 500, such as operations associated with display, telephone calls, data communication, camera operation, and recording. Processing component 502 may include one or more processors 520 to execute instructions to perform all or part of the steps of the methods described above. Furthermore, processing component 502 may include one or more modules to facilitate interaction between processing component 502 and other components. For example, processing component 502 may include a multimedia module to facilitate interaction between multimedia component 508 and processing component 502.

[0118] Memory 504 is used to store various types of data to support the operation of electronic device 500. Examples of this data include instructions for any application or method operating on electronic device 500, contact data, phonebook data, messages, pictures, multimedia, etc. Memory 504 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0119] Power supply component 506 provides power to various components of electronic device 500. Power supply component 506 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to electronic device 500.

[0120] Multimedia component 508 includes an interface that provides an output interface between electronic device 500 and user. In some embodiments, the interface may include a liquid crystal display (LCD) and a touch panel (TP). If the interface includes a touch panel, the interface may be implemented as a touchscreen to receive input signals from the user. The touch panel includes one or more touch sensors to sense touches, swipes, and gestures on the touch panel. The touch sensors may not only sense the boundaries of touch or swipe actions but also detect the duration and pressure associated with the touch or swipe operation. In some embodiments, multimedia component 508 includes a front-facing camera and / or a rear-facing camera. When electronic device 500 is in an operating mode, such as shooting mode or multimedia mode, the front-facing camera and / or rear-facing camera may receive external multimedia data. Each front-facing camera and rear-facing camera may be a fixed optical lens system or have focal length and optical zoom capabilities.

[0121] Audio component 510 is used to output and / or input audio signals. For example, audio component 510 includes a microphone (MIC) used to receive external audio signals when electronic device 500 is in an operating mode, such as call mode, recording mode, and voice recognition mode. The received audio signals may be further stored in memory 504 or transmitted via communication component 516. In some embodiments, audio component 510 also includes a speaker for outputting audio signals.

[0122] Input / output (I / O) interface 512 provides an interface between processing component 502 and peripheral interface modules, such as keyboards, click wheels, buttons, etc. These buttons may include, but are not limited to, home buttons, volume buttons, power buttons, and lock buttons.

[0123] Sensor assembly 514 includes one or more sensors for providing state assessments of various aspects of electronic device 500. For example, sensor assembly 514 may detect the on / off state of electronic device 500, the relative positioning of components such as the display and keypad of electronic device 500, changes in position of electronic device 500 or a component of electronic device 500, the presence or absence of user contact with electronic device 500, orientation or acceleration / deceleration of electronic device 500, and temperature changes of electronic device 500. Sensor assembly 514 may include a proximity sensor configured to detect the presence of nearby objects without any physical contact. Sensor assembly 514 may also include a light sensor, such as a CMOS or CCD image sensor, for use in imaging applications. In some embodiments, sensor assembly 514 may also include an accelerometer, gyroscope, magnetometer, pressure sensor, or temperature sensor.

[0124] Communication component 516 facilitates wired or wireless communication between electronic device 500 and other devices. Electronic device 500 can access wireless networks based on communication standards, such as WiFi, carrier networks (such as 2G, 3G, 4G, or 5G), or combinations thereof. In one exemplary embodiment, communication component 516 receives broadcast signals or broadcast-related information from an external broadcast management system via a broadcast channel. In one exemplary embodiment, communication component 516 also includes a near-field communication (NFC) module to facilitate short-range communication. For example, the NFC module may be implemented based on radio frequency identification (RFID) technology, Infrared Data Association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology, and other technologies.

[0125] In an exemplary embodiment, the electronic device 500 may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to implement the methods provided in the embodiments of this application.

[0126] In an exemplary embodiment, a non-transitory computer-readable storage medium including instructions is also provided, such as a memory 504 including instructions, which can be executed by a processor 520 of an electronic device 500 to perform the above-described method. For example, the non-transitory storage medium may be a ROM, random access memory (RAM), CD-ROM, magnetic tape, floppy disk, and optical data storage device, etc.

[0127] In an exemplary embodiment, the electronic device 500 may also be provided as a server, including a processing component 502, which further includes one or more processors, and memory resources represented by memory 504 for storing instructions, such as applications, that can be executed by the processing component 502. The applications stored in memory 504 may include one or more modules, each corresponding to a set of instructions. Furthermore, the processing component 502 is configured to execute instructions to perform the methods provided in the embodiments of this application.

[0128] Electronic device 500 may also include a power supply component 506 configured to perform power management of electronic device 500, a wired or wireless communication component 516 configured to connect electronic device 500 to a network, and an input / output (I / O) interface 512. Electronic device 500 may operate on an operating system stored in memory 504, such as Windows Server™, Mac OS X™, Unix™, Linux™, FreeBSD™, or similar.

[0129] It should be noted that, for the sake of simplicity, the method embodiments of this application are all described as a series of actions. However, those skilled in the art should understand that the embodiments of this application are not limited to the described order of actions, because according to the embodiments of this application, some steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also understand that the embodiments described in the specification are all preferred embodiments, and the actions involved are not necessarily essential to the embodiments of this application.

[0130] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the application disclosed herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this application are indicated by the following claims.

[0131] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this application is limited only by the appended claims.

Claims

1. A method for constructing an automatic pragmatic model based on a large model, characterized in that, include: Obtain the text describing the requirements for the target pragmatic model; Based on the target information extraction domain represented by the demand description text, target knowledge information is extracted from the knowledge base; The target knowledge information is used to characterize multiple information field types in the target information extraction domain; Based on the target knowledge information, a target pragmatic model is generated for extracting information fields in the target information extraction domain; The step of extracting target knowledge information from the knowledge base based on the target information extraction domain represented by the demand description text includes: Based on the requirement description text, extract multiple pieces of information knowledge from the knowledge base; Based on the semantic information represented by the demand description text, target information knowledge is extracted from multiple pieces of information knowledge to form the target knowledge information; the semantic information includes at least one of the textual direct intent or textual predictive intent of the demand description text, textual context, and textual linguistic behavior information. The step of generating a target pragmatic model based on the target knowledge information for extracting information fields in the target information extraction domain includes: When the semantic information is represented as containing the intention to extract macro-level information, the output data of the initial pragmatic model is limited according to the first field type data extracted from the target knowledge information to obtain a first target pragmatic model; the first field type is used to represent the semantic boundary conditions that limit the extraction scope of the demand description text. When the semantic information is represented as containing a specific information extraction intent, the output data of the initial pragmatic model is limited according to the second field type data extracted from the target knowledge information to obtain a second target pragmatic model; the second field type is used to represent the information extraction content conditions of the demand description text.

2. The automatic pragmatic model construction method based on a large model as described in claim 1, characterized in that, The step of extracting multiple pieces of information from the knowledge base based on the required description text includes: The semantic vector of the requirement description text is matched with the semantic vector of each piece of information knowledge in the knowledge base based on similarity. The information and knowledge corresponding to the semantic vectors in the knowledge base that match the semantic vectors of the requirement description text with a preset similarity condition are extracted.

3. The automatic pragmatic model construction method based on a large model as described in claim 1, characterized in that, The step of extracting target information knowledge from multiple pieces of information knowledge based on the semantic information represented by the demand description text to form the target knowledge information includes: When the semantic information is represented as containing the intention to extract macro-level information, the data of the first field type among the multiple pieces of information knowledge is determined as the target information knowledge; When the semantic information is represented as containing a specific information extraction intent, the data of the second field type in the multiple pieces of information knowledge are determined as the target information knowledge; the second field type is used to represent the information extraction content conditions of the requirement description text.

4. The automatic pragmatic model construction method based on a large model as described in claim 1, characterized in that, The method for constructing automatic pragmatic models based on large models also includes: The first information field of the target data is extracted based on the first target pragmatic model generated from the first demand description text; Based on the first information field and the target data, a second requirement description text is determined for generating the second target pragmatic model.

5. The automatic pragmatic model construction method based on a large model as described in claim 4, characterized in that, The method for constructing automatic pragmatic models based on large models also includes: Based on the correction of the first information field to the first requirement description text, the first target pragmatic model is updated, and the process returns to the step of extracting the first information field of the target data based on the first target pragmatic model generated from the first requirement description text.

6. An automatic pragmatic model construction device based on a large model, characterized in that, include: The requirement elicitation module is used to obtain the requirement description text for the target pragmatic model; The knowledge extraction module is used to extract target knowledge information from the knowledge base based on the target information extraction domain represented by the requirement description text. The target knowledge information is used to characterize multiple information field types in the target information extraction domain; The model generation module is used to generate a target pragmatic model based on the target knowledge information for extracting information fields in the target information extraction domain. The knowledge extraction module includes: The extraction submodule is used to extract multiple pieces of information knowledge from the knowledge base according to the required description text. A selection submodule is used to extract target information knowledge from multiple pieces of information knowledge based on the semantic information represented by the requirement description text, so as to form the target knowledge information; the semantic information includes at least one of the textual direct intent or textual predictive intent of the requirement description text, textual context, and textual language behavior information; The model generation module includes: A broad model generation submodule is used to limit the output data of the initial pragmatic model based on the first field type data extracted from the target knowledge information, when the semantic information is represented as containing the intention to extract macro information, so as to obtain a first target pragmatic model; the first field type is used to represent the semantic boundary conditions that limit the extraction scope of the demand description text. The specific model generation submodule is used to limit the output data of the initial pragmatic model based on the second field type data extracted from the target knowledge information when the semantic information is represented as containing the intention to extract specific information, so as to obtain a second target pragmatic model; the second field type is used to represent the information extraction content conditions of the demand description text.

7. A computer-readable storage medium, characterized in that, A computer program is stored on the computer-readable storage medium, which, when executed by a processor, implements the automatic pragmatic model construction method based on a large model as described in any one of claims 1 to 5.

8. An electronic device, characterized in that, It includes a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the automatic pragmatic model construction method based on a large model as described in any one of claims 1 to 5.

Images