Material information processing method and device, equipment and medium

Abstract

The present application relates to the technical field of data processing, and provides a kind of material information processing method, device, equipment and medium, the method comprises: according to user request data and semantic feature, determine material knowledge vectorization segment;According to semantic feature, obtain corresponding prompt word template;According to prompt word template, user request data, semantic feature and material knowledge vectorization segment are encapsulated into the input information of material management intelligent agent, and the output result is generated according to the input information by material management intelligent agent;Material management intelligent agent is the model that is constructed by adopting the model parameter fine-tuning of preset AI big model to fine-tune instruction set under the scene of material management.This application can realize knowledge accurate retrieval, knowledge accurate regeneration specification reliable, can replace a large amount of manual work, improve the intelligent level and operation efficiency of whole process of material management.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method, apparatus, equipment and medium for processing material information. Background Technology

[0002] In the field of materials management, it is necessary to manage information about materials throughout their entire lifecycle. Materials information management is reflected in the retrieval of materials information and the generation of materials knowledge documents. Currently, there are several shortcomings in materials information management, specifically: Regarding basic materials information retrieval, existing methods mainly rely on traditional document retrieval systems, simple database query functions, or manual retrieval. This approach is not only cumbersome and inefficient, but also limited by search criteria, making it difficult to quickly locate accurate information and prone to query errors. Regarding the generation of materials knowledge documents, existing methods often generate documents manually based on staff experience and limited, scattered data. The content lacks comprehensiveness, real-time updates, and innovation, failing to adapt to the complex and ever-changing business needs in materials management, such as policy updates and the expansion of materials categories. Summary of the Invention

[0003] To address the problems existing in the prior art, the present invention provides a method, apparatus, equipment, and medium for processing material information.

[0004] This invention provides a method for processing material information, comprising: Obtain user request data and obtain semantic features based on the user request data; Based on the user request data and the semantic features, determine the vectorized fragments of material knowledge; Based on the semantic features, obtain the corresponding prompt word template; According to the prompt word template, the user request data, the semantic features, and the vectorized fragments of material knowledge are encapsulated into input information for the material management intelligent agent, and the material management intelligent agent generates output results based on the input information; the material management intelligent agent is a model constructed by fine-tuning the model parameters of a preset AI large model using a fine-tuning instruction set in the material management scenario.

[0005] According to a material information processing method provided by the present invention, the step of determining a vectorized fragment of material knowledge based on the user request data and the semantic features includes: Based on the semantic features, a set of knowledge vectorized fragments that match the semantic features is selected from a preset material knowledge vector library; the material knowledge vector library is constructed from the feature vectors corresponding to the knowledge fragments obtained by extracting multi-source data from the field of material management. Based on the keyword information in the user request data, the material knowledge vectorization fragment is selected from the knowledge vectorization fragment set.

[0006] According to a material information processing method provided by the present invention, the semantic features represent knowledge retrieval categories, and correspondingly, the step of filtering the material knowledge vectorized fragments from the knowledge vectorized fragment set based on keyword information in the user request data includes: Determine the feature vector corresponding to the keyword information; The feature vectors are matched for similarity in the set of knowledge vectorized fragments to obtain the feature vectors with the highest similarity. The knowledge vectorization fragments corresponding to the matched feature vectors are used as the material knowledge vectorization fragments.

[0007] According to a material information processing method provided by the present invention, the semantic features represent knowledge generation categories, and correspondingly, the step of filtering the material knowledge vectorization fragments from the knowledge vectorization fragment set based on keyword information in the user request data includes: Determine the feature vector corresponding to the keyword information; The feature vectors are matched for similarity in the knowledge vectorized fragment set to obtain the top N feature vectors in similarity ranking. The knowledge vectorization fragments corresponding to the top N feature vectors obtained from the similarity ranking are used as the knowledge vectorization fragments of the material.

[0008] According to a material information processing method provided by the present invention, the fine-tuning instruction set is an instruction sample set constructed for text prediction tasks and entity association tasks in material management scenarios, and the instruction sample set consists of multiple sets of material management instruction samples; Each set of materials management instruction samples includes: The instruction field is used to define the task objectives of text prediction or entity association in the context of materials management. The input field is the text to be predicted or the entity-related query text in the context of materials management. The output fields are the prediction results of text prediction or the relationship results of entity association in the context of material management.

[0009] According to a material information processing method provided by the present invention, the step of constructing the fine-tuning instruction set includes: Acquire multi-source data in the field of materials management; The multi-source data in the field of materials management is standardized to obtain standardized multi-source data. Extract material entities, knowledge categories, and relationships between entities from standardized multi-source data; Based on the extracted material entities, knowledge categories, and relationships between entities, a fine-tuning instruction set is constructed.

[0010] According to a material information processing method provided by the present invention, the construction steps of the knowledge vector base include: Acquire multi-source data in the field of materials management; The multi-source data in the field of materials management is standardized to obtain standardized multi-source data. Extract material entities, knowledge categories, and relationships between entities from standardized multi-source data; The extracted material entities, knowledge categories, and relationships between entities are vectorized to obtain the corresponding feature vectors; A knowledge vector base is constructed based on the feature vectors.

[0011] According to a material information processing method provided by the present invention, the numerical data of the feature vector are arranged in the order of material entity, time series, and material category.

[0012] According to a material information processing method provided by the present invention, before fine-tuning the model parameters of a preset AI large model, an appropriate model version is selected based on the data scale, task complexity, and hardware conditions in the field of material management.

[0013] The present invention also provides a material information processing device, comprising: The acquisition module is used to acquire user request data and obtain semantic features based on the user request data; The filtering module is used to determine the vectorized fragments of material knowledge based on the user request data and the semantic features; The calling module is used to obtain the corresponding prompt word template based on the semantic features; The processing module is used to encapsulate the user request data, the semantic features, and the vectorized fragments of material knowledge into input information for a material management intelligent agent based on the prompt word template. The material management intelligent agent generates output results based on the input information. The material management intelligent agent is a model constructed by fine-tuning the model parameters of a preset AI large model using a fine-tuning instruction set in the material management scenario.

[0014] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement any of the above-described material information processing methods.

[0015] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the above-described material information processing methods.

[0016] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements any of the above-described material information processing methods.

[0017] This invention provides a material information processing method, apparatus, equipment, and medium. By determining vectorized fragments of material knowledge based on user request data and semantic features, and encapsulating the user request data, semantic features, and vectorized fragments of material knowledge into input information for a material management intelligent agent based on prompt word templates, the material management intelligent agent, constructed by fine-tuning the model parameters of a preset AI model using a fine-tuning instruction set for the material management scenario, generates output results based on the input information. This enables accurate knowledge retrieval and accurate and reliable knowledge regeneration, replacing a large amount of manual work and improving the intelligence level and operational efficiency of the entire material management process. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0019] Figure 1 This is a flowchart illustrating the material information processing method provided by the present invention.

[0020] Figure 2 This is a flowchart of the overall process framework of the material information processing method of the present invention.

[0021] Figure 3 This is a schematic diagram of the material information processing device provided by the present invention.

[0022] Figure 4 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0024] Figure 1 This invention provides a schematic flowchart of a material information processing method. (See attached diagram.) Figure 1 The method includes the following steps: Step 11: Obtain user request data and obtain semantic features based on the user request data.

[0025] Step 12: Determine the vectorized fragments of material knowledge based on the user request data and semantic features.

[0026] Step 13: Obtain the corresponding prompt word template based on semantic features.

[0027] Step 14: Based on the prompt word template, encapsulate the user request data, semantic features, and material knowledge vectorized fragments into input information for the material management intelligent agent. The material management intelligent agent generates the output result based on the input information. The material management intelligent agent is a model constructed by fine-tuning the model parameters of a preset AI large model using a fine-tuning instruction set in the material management scenario.

[0028] Regarding steps 11-14, it should be noted that in the field of materials management, it is necessary to manage information about materials throughout their entire lifecycle. Materials information management is reflected in the retrieval of materials information and the generation of materials knowledge documents. Currently, there are several shortcomings in materials information management, specifically: In terms of basic materials information retrieval, existing methods mainly rely on traditional document retrieval systems, simple database query functions, or manual retrieval. This method is not only cumbersome and inefficient, but also limited by search criteria, making it difficult to quickly locate accurate information and prone to query errors. Regarding the generation of materials knowledge documents, the content generated by existing methods is mostly based on the personal experience of staff and a small amount of scattered data, resulting in a lack of comprehensiveness, real-time updates, and innovation. This makes it unable to adapt to the complex and ever-changing business needs in the field of materials management, such as policy updates and the expansion of materials categories.

[0029] To address this, the present invention provides a material information processing method. The design concept of this method aims to construct an intelligent generation / retrieval system for material management scenarios. In this system, user request data is first acquired and semantic features are extracted, and material knowledge vectorization fragments are matched accordingly. Then, the corresponding prompt word template is called to encapsulate the request data, semantic features, and knowledge fragments into input information, which is then input into an AI large-scale intelligent agent optimized by a material management fine-tuning instruction set. This accurately outputs material management-related results, improving response efficiency and professionalism.

[0030] In this invention, user request data is acquired, and semantic features are obtained based on this data. Acquiring user request data here is equivalent to obtaining the user's questions and needs regarding material information management. Obtaining semantic features is equivalent to understanding from this user request data what the user "wants to do, what they are asking, and what category they belong to." Specifically, this invention involves: receiving and parsing user-inputted material management-related request data; performing preprocessing operations such as text cleaning, word segmentation, entity recognition, and intent parsing on the request data; and using a pre-set semantic extraction model to accurately capture user intent, key entities, and contextual information from the preprocessed request data, generating a structured semantic feature vector that represents the specific meaning of the request, providing a foundation for subsequent knowledge matching and model input.

[0031] For example, a user inputs the request data "What are the acceptance standards for fire protection materials in 2025?". Then, semantic features are obtained from the user request data. This involves converting the sentence into "meaning features" that a materials management intelligent agent, constructed by fine-tuning the parameters of a pre-set AI model using a fine-tuning instruction set for a materials management scenario, can understand. The intelligent agent understands these semantic features: query type: query for standards / specifications; domain: fire protection materials; time: 2025; action: acceptance; intent: to obtain official / current acceptance requirements. Finally, through semantic encoding, the user request is transformed into a semantic feature vector containing domain type, time attribute, query intent, and entity.

[0032] In this invention, the amount of material knowledge is enormous. To quickly filter out results that highly match the user's intent, reduce interference from irrelevant information, improve retrieval efficiency and accuracy, avoid the model generating false information, and ensure that the answers are compliant, professional, and reliable, it is necessary to determine vectorized fragments of material knowledge based on the user's request data and semantic features. These vectorized fragments can include information related to the user's request, such as policies, standards, processes, cases, data, rules, systems, plans, and terminology in the field of material management.

[0033] For example, if a user asks, "What are the acceptance standards for fire protection materials?", the corresponding knowledge vectorization fragments could be: the provisions of the fire protection material acceptance specifications; acceptance technical indicators; warehousing and acceptance procedures; sampling inspection requirements; and relevant policy documents.

[0034] In this invention, the intelligent agent sets prompt word templates to standardize the output format of the large model, constrain the scope of responses, and guide the model to generate professional, compliant, and accurate content based on the retrieved vectorized fragments of material knowledge. This avoids generating false information and ensures that the output conforms to the standards and business requirements of the material management field. To this end, corresponding prompt word templates are obtained based on semantic features. For example, for the retrieval function and the document generation function of this invention, the retrieval function is configured with a dedicated retrieval prompt word template to guide the model in parsing the query intent and accurately matching relevant knowledge fragments; the document generation function is configured with a dedicated generation prompt word template to constrain the model to output professional content that conforms to the standards of the material management field based on the retrieved fragments. The two functions are independent but complementary.

[0035] In this invention, the prompt word template can package user request data, semantic features, and vectorized fragments of material knowledge to form input data that the intelligent agent can directly process. This input data is then sent to the material management intelligent agent. The material management intelligent agent generates output results based on the input information. The material management intelligent agent is a model constructed by fine-tuning the parameters of a pre-set AI model using a fine-tuning instruction set specific to the material management scenario. By fine-tuning the parameters of the pre-set AI model using the material management scenario fine-tuning instruction set, the pre-set AI model is adapted to the material management domain, endowing it with professional semantic understanding, knowledge matching, and content generation capabilities in the material management domain. This addresses the technical shortcomings of general-purpose large models in material management scenarios, such as insufficient professionalism, poor accuracy, and susceptibility to generating false information, providing core model support for subsequent knowledge retrieval, extraction, and generation functions.

[0036] The intelligent agent for material management in this invention is obtained by fine-tuning the instruction set of material management scenarios, and has strong domain adaptability. It combines input information encapsulated with prompt word templates to carry out reasoning output, resulting in accurate retrieval and standardized and reliable results. It can replace a large amount of manual work and improve the intelligence level and operational efficiency of the entire material management process.

[0037] The material information processing method provided by this invention determines vectorized fragments of material knowledge based on user request data and semantic features. Then, according to a prompt word template, it encapsulates the user request data, semantic features, and vectorized fragments of material knowledge into input information for a material management intelligent agent. The material management intelligent agent, constructed by fine-tuning the model parameters of a preset AI model using a fine-tuning instruction set for the material management scenario, generates output results based on the input information. This method enables accurate knowledge retrieval and reliable, standardized knowledge regeneration, replacing a large amount of manual work and improving the intelligence level and operational efficiency of the entire material management process.

[0038] A further step in the above method mainly explains the process of determining vectorized fragments of material knowledge based on user request data and semantic features, as follows: Based on semantic features, a set of knowledge vectorized fragments that match the semantic features is selected from the pre-set material knowledge vector library; the material knowledge vector library is constructed from the feature vectors corresponding to the knowledge fragments obtained by extracting multi-source data from the field of material management. Based on the keyword information in the user request data, filter out the material knowledge vectorization fragments from the knowledge vectorization fragment set.

[0039] It should be noted that, based on the semantic features obtained from the parsing, the preset material knowledge vector library is traversed, and a set of knowledge vectorized fragments that meet the preset threshold in terms of matching degree with the semantic features are selected through vector similarity calculation. The material knowledge vector library is constructed by extracting knowledge from multi-source data such as policy documents, business ledgers, operating procedures, historical cases, and technical standards in the field of material management, and then vectorizing the text. The result is a unified construction of the feature vectors corresponding to each knowledge fragment.

[0040] Further extract keywords from user request data, and based on keyword matching rules, perform a second precise screening on the above knowledge vector fragment set to finally determine the target material knowledge vector fragment.

[0041] This invention employs a two-stage screening method combining semantic feature matching and keyword verification. First, it utilizes semantic features and vector similarity for initial screening, enabling in-depth analysis of user query intent and effectively narrowing the search scope. Then, it further refines the initial results based on user-requested keywords, significantly reducing interference from irrelevant information. Simultaneously, the material knowledge vector database upon which the retrieval relies is constructed from multi-source professional data across various domains, ensuring the authenticity of the knowledge content and the security of the retrieved materials. These two screening mechanisms work together to significantly improve the accuracy of knowledge fragment matching while enhancing retrieval efficiency.

[0042] In a further step of the above method, in this invention, the main focus is on the knowledge retrieval and knowledge generation categories of material management. Therefore, the process of selecting material knowledge vectorization fragments from a set of knowledge vectorization fragments based on keyword information in user request data is explained as follows: 1) Knowledge Retrieval Determine the feature vector corresponding to the keyword information; The feature vectors are matched for similarity in the set of knowledge vectorized fragments, and the feature vector with the highest similarity is obtained. The knowledge vectorization fragments corresponding to the matched feature vectors are used as the material knowledge vectorization fragments.

[0043] 2) Knowledge Generation Determine the feature vector corresponding to the keyword information; The feature vectors are matched for similarity in the set of knowledge vectorized fragments, and the top N feature vectors with the highest similarity are obtained. The knowledge vectorization fragments corresponding to the top N feature vectors obtained from the similarity ranking are used as the knowledge vectorization fragments of the material.

[0044] It should be noted that in this invention, for knowledge retrieval and knowledge generation, the retrieval scenario focuses on accurate response, so the single fragment with the best matching degree is selected; the generation scenario requires the fusion and creation of multi-source information, so multiple fragments with the highest ranking are selected to adapt to the business needs of the two types of functions respectively.

[0045] The further method described above mainly explains the process of constructing a model by fine-tuning the parameters of a pre-set large AI model using a fine-tuning instruction set under a materials management scenario, as detailed below: The construction process of the intelligent agent for materials management in this invention is as follows: First, acquire multi-source data from the materials management domain. Second, standardize the multi-source data to obtain standardized multi-source data. Third, extract materials entities, knowledge categories, and relationships between entities from the standardized multi-source data. Fourth, construct a fine-tuning instruction set based on the extracted materials entities, knowledge categories, and relationships between entities. Fifth, fine-tune the parameters of a general preset AI model (such as the Qwen base model) based on the preprocessed fine-tuning instruction set to adapt it to the semantics and professional rules of materials management business, thus obtaining a materials domain-specific large model.

[0046] This invention employs a multi-channel data collection approach that combines automation and manual methods to ensure comprehensive and real-time data collection. Web crawlers are used to periodically retrieve the latest policy documents and regulations from government material management department websites and industry platforms. A database interface is connected to the material management system to obtain structured data such as material inventory, inbound and outbound records, and management standards. Internal material management operation manuals, process specifications, and other electronic documents are collected, organized, and digitally archived. Relevant academic literature and industry reports are downloaded from academic databases such as CNKI and Wanfang. Historical emergency material allocation and material management cases are systematically compiled to form a standardized case dataset.

[0047] Precise data cleaning was achieved using Python and related toolkits. The pandas library was used to process structured material data, correcting outliers with the mean or median and removing duplicate and missing data. For text data such as policy documents and academic literature, regular expressions were used to remove garbled characters, special characters, invalid spaces, and redundant paragraphs. The data was uniformly encoded in UTF-8 and formatted as TXT to ensure it was standardized and clean.

[0048] Develop dedicated annotation tools and set annotation items such as "material category" and "management standard" according to the characteristics of the domain; organize domain professionals to carry out annotation, and adopt the "double annotation + cross-review" mode to check and correct the annotation results to ensure that the annotation is accurate and consistent, and provide support for the model to learn domain features. Standardize multi-source datasets.

[0049] By using the AllenNLP toolkit, an extraction model is built to achieve entity recognition and relationship extraction, automatically identifying entities such as material names and management standards, as well as their associated relationships.

[0050] Based on the extracted material entities, knowledge categories, and relationships between entities, a fine-tuning instruction set is constructed. This fine-tuning instruction set is a set of instruction samples constructed for text prediction and entity association tasks in the material management scenario, and the instruction sample set consists of multiple sets of material management instruction samples; Each set of materials management instruction samples includes: The instruction field is used to define the task objectives of text prediction or entity association in the context of materials management. The input field is the text to be predicted or the entity-related query text in the context of materials management. The output fields are the prediction results of text prediction or the relationship results of entity association in the context of material management.

[0051] Accordingly, constructing a knowledge vector base also requires standardized multi-source data. After obtaining the standardized multi-source data, the material entities, knowledge categories, and relationships between entities are extracted. These extracted entities, categories, and relationships are then vectorized to obtain corresponding feature vectors. The Gensim library's Word2Vec model is used to generate text word vectors, effectively capturing semantic information. The numerical data of the feature vectors are arranged in the order of material entity, time series, and material category to achieve standardized representation. Finally, the knowledge vector base is constructed based on the feature vectors.

[0052] In this invention, before fine-tuning the model parameters of the pre-set large AI model, a suitable model version is selected based on the data scale, task complexity, and hardware conditions in the field of materials management. Downloading a model version compatible with the current hardware resources from official channels ensures standard model structure, environmental compatibility, and stable operation, facilitating subsequent parameter adjustments and training.

[0053] See Figure 2 This is a flowchart illustrating the overall process framework of the material information processing method of the present invention.

[0054] The material information processing device provided by the present invention is described below. The material information processing device described below and the material information processing method described above can be referred to in correspondence.

[0055] Figure 3 A schematic diagram of the structure of a material information processing device provided by the present invention is shown below. Figure 3 The device includes an acquisition module 31, a filtering module 32, a calling module 33, and a processing module 34, wherein: The acquisition module is used to acquire user request data and obtain semantic features based on the user request data. The filtering module is used to determine vectorized fragments of material knowledge based on user request data and semantic features; The calling module is used to obtain the corresponding prompt word template based on semantic features; The processing module is used to encapsulate user request data, semantic features, and vectorized fragments of material knowledge into input information for the material management intelligent agent based on the prompt word template. The material management intelligent agent generates output results based on the input information. The material management intelligent agent is a model constructed by fine-tuning the model parameters of a preset AI large model using a fine-tuning instruction set in the material management scenario.

[0056] Since the apparatus of this embodiment is based on the same principle as the method of the above embodiment, more detailed explanations will not be repeated here.

[0057] It should be noted that, in the embodiments of the present invention, the relevant functional modules can be implemented by a hardware processor.

[0058] The material information processing device provided by this invention determines vectorized fragments of material knowledge based on user request data and semantic features. Then, according to a prompt word template, it encapsulates the user request data, semantic features, and vectorized fragments of material knowledge into input information for a material management intelligent agent. The material management intelligent agent, constructed by fine-tuning the model parameters of a preset AI model using a fine-tuning instruction set for material management scenarios, generates output results based on the input information. This enables accurate knowledge retrieval and reliable, standardized knowledge regeneration, replacing a large amount of manual work and improving the intelligence level and operational efficiency of the entire material management process.

[0059] Figure 4 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 4 As shown, the electronic device may include: a processor 41, a communication interface 42, a memory 43, and a communication bus 44, wherein the processor 41, the communication interface 42, and the memory 43 communicate with each other through the communication bus 44. The processor 41 can call logical instructions in the memory 43 to execute a material information processing method, which includes: The process involves acquiring user request data and obtaining semantic features based on it; determining vectorized fragments of material knowledge based on the user request data and semantic features; obtaining corresponding prompt word templates based on the semantic features; and encapsulating the user request data, semantic features, and vectorized fragments of material knowledge into input information for a material management intelligent agent based on the prompt word templates. The material management intelligent agent generates output results based on the input information. The material management intelligent agent is a model constructed by fine-tuning the parameters of a pre-set large AI model using a fine-tuning instruction set for the material management scenario.

[0060] Furthermore, the logical instructions in the aforementioned memory 43 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0061] On the other hand, the present invention also provides a computer program product, the computer program product comprising a computer program that can be stored on a non-transitory computer-readable storage medium, wherein when the computer program is executed by a processor, the computer is able to execute a material information processing method, the method comprising: The process involves acquiring user request data and obtaining semantic features based on it; determining vectorized fragments of material knowledge based on the user request data and semantic features; obtaining corresponding prompt word templates based on the semantic features; and encapsulating the user request data, semantic features, and vectorized fragments of material knowledge into input information for a material management intelligent agent based on the prompt word templates. The material management intelligent agent generates output results based on the input information. The material management intelligent agent is a model constructed by fine-tuning the parameters of a pre-set large AI model using a fine-tuning instruction set for the material management scenario.

[0062] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to perform a material information processing method, the method comprising: The process involves acquiring user request data and obtaining semantic features based on it; determining vectorized fragments of material knowledge based on the user request data and semantic features; obtaining corresponding prompt word templates based on the semantic features; and encapsulating the user request data, semantic features, and vectorized fragments of material knowledge into input information for a material management intelligent agent based on the prompt word templates. The material management intelligent agent generates output results based on the input information. The material management intelligent agent is a model constructed by fine-tuning the parameters of a pre-set large AI model using a fine-tuning instruction set for the material management scenario.

[0063] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0064] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0065] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for processing material information, characterized in that, include: Obtain user request data and obtain semantic features based on the user request data; Based on the user request data and the semantic features, determine the vectorized fragments of material knowledge; Based on the semantic features, obtain the corresponding prompt word template; According to the prompt word template, the user request data, the semantic features, and the vectorized fragments of material knowledge are encapsulated into input information for the material management intelligent agent, and the material management intelligent agent generates output results based on the input information; the material management intelligent agent is a model constructed by fine-tuning the model parameters of a preset AI large model using a fine-tuning instruction set in the material management scenario.

2. The material information processing method according to claim 1, characterized in that, The step of determining the vectorized fragment of material knowledge based on the user request data and the semantic features includes: Based on the semantic features, a set of knowledge vectorized fragments that match the semantic features is selected from a preset material knowledge vector library; the material knowledge vector library is constructed from the feature vectors corresponding to the knowledge fragments obtained by extracting multi-source data from the field of material management. Based on the keyword information in the user request data, the material knowledge vectorization fragment is selected from the knowledge vectorization fragment set.

3. The material information processing method according to claim 2, characterized in that, The semantic features represent knowledge retrieval categories. Correspondingly, the step of filtering the material knowledge vectorization fragments from the knowledge vectorization fragment set based on keyword information in the user request data includes: Determine the feature vector corresponding to the keyword information; The feature vectors are matched for similarity in the set of knowledge vectorized fragments to obtain the feature vectors with the highest similarity. The knowledge vectorization fragments corresponding to the matched feature vectors are used as the material knowledge vectorization fragments.

4. The material information processing method according to claim 2, characterized in that, The semantic features represent the knowledge generation category. Correspondingly, the step of selecting the material knowledge vectorization fragment from the knowledge vectorization fragment set based on the keyword information in the user request data includes: Determine the feature vector corresponding to the keyword information; The feature vectors are matched for similarity in the knowledge vectorized fragment set to obtain the top N feature vectors in similarity ranking. The knowledge vectorization fragments corresponding to the top N feature vectors obtained from the similarity ranking are used as the knowledge vectorization fragments of the material.

5. The material information processing method according to claim 1, characterized in that, The fine-tuning instruction set is a set of instruction samples constructed for text prediction tasks and entity association tasks in the material management scenario. The instruction sample set consists of multiple sets of material management instruction samples. Each set of materials management instruction samples includes: The instruction field is used to define the task objectives of text prediction or entity association in the context of materials management. The input field is the text to be predicted or the entity-related query text in the context of materials management. The output fields are the prediction results of text prediction or the relationship results of entity association in the context of material management.

6. The material information processing method according to claim 5, characterized in that, The steps for constructing the fine-tuning instruction set include: Acquire multi-source data in the field of materials management; The multi-source data in the field of materials management is standardized to obtain standardized multi-source data. Extract material entities, knowledge categories, and relationships between entities from standardized multi-source data; Based on the extracted material entities, knowledge categories, and relationships between entities, a fine-tuning instruction set is constructed.

7. The material information processing method according to claim 2, characterized in that, The steps for constructing the knowledge vector base include: Acquire multi-source data in the field of materials management; The multi-source data in the field of materials management is standardized to obtain standardized multi-source data. Extract material entities, knowledge categories, and relationships between entities from standardized multi-source data; The extracted material entities, knowledge categories, and relationships between entities are vectorized to obtain the corresponding feature vectors; A knowledge vector base is constructed based on the feature vectors.

8. A material information processing device, characterized in that, include: The acquisition module is used to acquire user request data and obtain semantic features based on the user request data; The filtering module is used to determine the vectorized fragments of material knowledge based on the user request data and the semantic features; The calling module is used to obtain the corresponding prompt word template based on the semantic features; The processing module is used to encapsulate the user request data, the semantic features, and the vectorized fragments of material knowledge into input information for a material management intelligent agent based on the prompt word template. The material management intelligent agent generates output results based on the input information. The material management intelligent agent is a model constructed by fine-tuning the model parameters of a preset AI large model using a fine-tuning instruction set in the material management scenario.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the material information processing method as described in any one of claims 1-7.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the material information processing method as described in any one of claims 1-7.

Images