Private network data generation method and device, electronic equipment and storage medium

Abstract

The invention provides a private network data generation method and device, electronic equipment and a storage medium, and relates to the field of artificial intelligence, and the main technical characteristics comprise the following steps: in response to a received user request, determining a target vector matched with the user request from a local private network vector library; obtaining a target knowledge fragment corresponding to the target vector; generating prompt information according to the target knowledge fragment and the user request; and processing the prompt information through the target large language model to obtain the target private network data, thereby improving the efficiency and accuracy of private network data generation.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to the field of artificial intelligence, and in particular to a private network data generation method and device, electronic equipment and storage medium. BACKGROUND

[0002] With the development of 5G communication technology, the demand for enterprise private network construction is increasing, and the generation of private network data such as pre-sales solutions or reports for private networks has become a heavy task.

[0003] Currently, the generation of private network data mainly relies on two methods: one is manually written by technical personnel according to personal experience, and the other is based on pre-defined templates for modification and filling. The former relies on the accumulation of professional knowledge of technical personnel, and has the problems of efficiency limitation and consistency difficulty to guarantee; the latter improves the generation speed, but the homogeneity of the output content is high, and it is difficult to adapt to the individual needs of different industries.

[0004] In addition, general large language models show potential in private network data generation, but there are problems such as insufficient coverage of professional knowledge, inaccurate use of terminology, and imperfect description of technical details, making it difficult to generate private network data that meets professional requirements. SUMMARY

[0005] The present disclosure provides a private network data generation method, device, electronic equipment and storage medium. Its main purpose is to solve the problem that the generation of private network data in the prior art relies on manual work, resulting in limited efficiency, and the general large language model is difficult to generate accurate and professional private network data due to the lack of field professional knowledge.

[0006] According to a first aspect of the present disclosure, a private network data generation method is provided, comprising: determining a target vector matching the user request from a local private network vector library in response to the received user request; obtaining a target knowledge fragment corresponding to the target vector; generating prompt information according to the target knowledge fragment and the user request; processing the prompt information through a target large language model to obtain target private network data.

[0007] In some embodiments, before the step of determining a target vector matching the user request from a local private network vector library in response to the received user request, the method further comprises: obtaining related data of private network data; processing the related data to obtain structured data; performing vector conversion on the structured data to obtain vector data, and constructing the local private network vector library based on the vector data.

[0008] In some embodiments, the processing of the related data to obtain the structured data comprises: cleaning the related data of the special network data to obtain cleaned related data; classifying the cleaned related data according to a preset classification rule to obtain classified related data; wherein the preset classification rule comprises at least one of an industry type, a business scenario, a terminal application type, and a professional technical field; segmenting the classified related data to obtain data segments in a preset unit; integrating the data segments to obtain the structured data.

[0009] In some embodiments, the determining of the target vector matching the user request from the local special network vector library in response to the received user request comprises: retrieving candidate vectors from the local special network vector library based on keywords in the user request; converting the user request into a query vector; calculating the semantic similarity between the query vector and the candidate vector; determining the target vector based on the semantic similarity.

[0010] In some embodiments, the generating of the prompt information according to the target knowledge segment and the user request comprises: normalizing the user request to obtain demand information in the user request; filtering the target knowledge segment based on the demand information to determine knowledge content related to the demand information; integrating the demand information and the filtered knowledge content according to a preset logic, and adding instructions for generating the target special network data to obtain the prompt information.

[0011] In some embodiments, before the processing of the prompt information by the target large language model to obtain the target special network data, the method further comprises: adjusting the large language model in a lightweight fine-tuning manner to obtain the target large language model.

[0012] According to a second aspect of the present disclosure, a special network data generation apparatus is provided, comprising: a determination unit configured to determine a target vector matching a user request from a local special network vector library in response to the received user request; a first acquisition unit configured to acquire a target knowledge segment corresponding to the target vector; A generation unit is used to generate prompt information based on the target knowledge fragment and the user request; The first processing unit is used to process the prompt information through the target large language model to obtain target private network data.

[0013] In some embodiments, the apparatus further includes: The second acquisition unit is used to acquire relevant data of the private network data before the determining unit determines the target vector matching the user request from the local private network vector library in response to the received user request. The second processing unit is used to process the relevant data to obtain structured data; The construction unit is used to perform vector transformation on the structured data to obtain vector data, and to construct the local private network vector library based on the vector data.

[0014] In some embodiments, the second processing unit includes: The cleaning module is used to clean the relevant data of the private network data to obtain cleaned relevant data. The classification module is used to classify the cleaned relevant data according to preset classification rules to obtain classified relevant data; wherein, the preset classification rules include at least one of industry type, business scenario, terminal application type and professional technical field. The segmentation module is used to segment the classified related data to obtain data segments with preset units as units; An integration module is used to integrate the data fragments to obtain the structured data.

[0015] In some embodiments, the determining unit includes: The retrieval module is used to retrieve candidate vectors from the local private network vector library based on the keywords in the user request; A conversion module is used to convert the user request into a query vector; A calculation module is used to calculate the semantic similarity between the query vector and the candidate vector; The first determining module is used to determine the target vector based on the semantic similarity.

[0016] In some embodiments, the generating unit includes: The processing module is used to standardize the user request and obtain the demand information in the user request; The second determining module is used to filter the target knowledge fragments based on the demand information and determine the knowledge content related to the demand information; The third determining module is used to integrate the demand information and the filtered knowledge content according to a preset logic, and add an instruction to generate the target private network data to obtain the prompt information.

[0017] In some embodiments, the apparatus further includes: The adjustment unit is used to adjust the large language model in a lightweight fine-tuning manner before the first processing unit processes the prompt information through the target large language model to obtain the target private network data, so as to obtain the target large language model.

[0018] According to a third aspect of this disclosure, an electronic device is provided, comprising: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method described in the first aspect above.

[0019] According to a fourth aspect of this disclosure, a non-transitory computer-readable storage medium is provided storing computer instructions, wherein the computer instructions are configured to cause the computer to perform the method described in the first aspect above.

[0020] According to a fifth aspect of this disclosure, a computer program product is provided, comprising a computer program that, when executed by a processor, implements the method described in the first aspect above.

[0021] In summary, the private network data generation method, apparatus, electronic device, and storage medium provided in this disclosure include: responding to a received user request by determining a target vector matching the user request from a local private network vector library; obtaining a target knowledge fragment corresponding to the target vector; generating prompt information based on the target knowledge fragment and the user request; and processing the prompt information using a target large language model to obtain target private network data. Compared with related technologies, the solution of this disclosure improves the efficiency and accuracy of private network data generation by retrieving a target vector matching the user request from a local private network vector library, generating prompt information by processing the target knowledge fragment corresponding to the target vector and the user request, and generating target private network data through large language model processing.

[0022] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this application, nor is it intended to limit the scope of this application. Other features of this application will become readily apparent from the following description. Attached Figure Description

[0023] The accompanying drawings are provided to better understand this solution and do not constitute a limitation of this disclosure. Wherein: Figure 1 A flowchart illustrating a private network data generation method provided in an embodiment of this disclosure; Figure 2 A flowchart illustrating another private network data generation method provided in this embodiment of the present disclosure; Figure 3 A flowchart illustrating another private network data generation method provided in this embodiment of the present disclosure; Figure 4 A flowchart illustrating another private network data generation method provided in this embodiment of the present disclosure; Figure 5 A flowchart illustrating another private network data generation method provided in this embodiment of the present disclosure; Figure 6 A flowchart illustrating another private network data generation method provided in this embodiment of the present disclosure; Figure 7 A schematic diagram of the overall architecture of a private network data generation method provided in an embodiment of this disclosure; Figure 8 This is a schematic diagram of the structure of a private network data generation device provided in an embodiment of the present disclosure; Figure 9 A schematic diagram of another private network data generation device provided in an embodiment of this disclosure; Figure 10 This is a schematic block diagram of an example electronic device provided in an embodiment of this disclosure. Detailed Implementation

[0024] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0025] The following description, with reference to the accompanying drawings, outlines a method, apparatus, electronic device, and storage medium for generating private network data according to embodiments of this disclosure.

[0026] Figure 1 This is a flowchart illustrating a private network data generation method provided in an embodiment of this disclosure.

[0027] like Figure 1 As shown, the method includes steps 101-104.

[0028] Step 101: In response to the received user request, determine the target vector that matches the user request from the local private network vector library.

[0029] In some embodiments, a user request is a user-input description of their needs for private network data, which may take the form of text, including but not limited to text (such as "generating an equipment list for a smart mine 5G private network" or "providing key points of a security protection solution for a power private network"). The local private network vector library is a pre-built local database storing vector data related to the private network domain. The vector data in this library is obtained by vectorizing knowledge fragments from the private network domain. Determining the target vector matching the user request involves converting the user request into a corresponding vector form, then searching the local private network vector library based on the vector from the user request. By calculating the similarity between vectors, vectors whose similarity to the vectors in the user request meets preset conditions are selected as the target vectors.

[0030] Using the above method, vector data related to the user's request can be located from the local private network vector library.

[0031] Step 102: Obtain the target knowledge fragment corresponding to the target vector.

[0032] In some embodiments, each vector data stored in the local private network vector library is associated with its corresponding original knowledge fragment, which is professional information in the private network field (such as private network networking specifications, equipment parameters, industry application cases, etc.). Obtaining the target knowledge fragment corresponding to the target vector means searching for its associated original knowledge fragment in the local private network vector library based on the target vector; this original knowledge fragment is the target knowledge fragment.

[0033] Using the above method, the private network expertise corresponding to the target vector can be obtained.

[0034] Step 103: Generate prompt information based on the target knowledge fragment and the user request.

[0035] In some embodiments, generating prompt information based on target knowledge fragments and user requests refers to integrating the content of the user request with the target knowledge fragments to form input content that can be processed by the target large language model. The integration process includes, but is not limited to, using the user request as a guide for generating target private network data and using the target knowledge fragments as the basis for generating target private network data, for example, forming the form of "Based on the following knowledge: [target knowledge fragment content], satisfying user needs: [user request content], generating corresponding private network data".

[0036] Using the above method, input information containing user needs and related knowledge can be generated to drive the target large language model to generate target private network data.

[0037] Step 104: Process the prompt information using the target large language model to obtain the target private network data.

[0038] In some embodiments, the target large language model is an artificial intelligence model with text generation capabilities, which can understand and process the input prompts and output qualified private network data based on the user request and target knowledge fragments in the prompts. The target private network data can take various forms, such as private network solution documents, equipment configuration lists, technical parameter descriptions, etc.

[0039] The above methods can generate target private network data that meets user needs.

[0040] In summary, the private network data generation method provided in this disclosure can improve the efficiency and accuracy of private network data generation by retrieving a target vector matching the user request from a local private network vector library, processing the target knowledge fragment corresponding to the target vector and the user request to generate prompt information, and then processing the target private network data through a large language model.

[0041] Figure 2 A flowchart illustrating a private network data generation method provided in this disclosure embodiment is further shown, such as... Figure 2 As shown, the method includes steps 201-203.

[0042] Step 201: Obtain relevant data from the private network.

[0043] In some embodiments, the relevant data for private network data refers to raw data in the private network field, including but not limited to industry standard documents, technical specification documents, historical project plans, equipment parameter manuals, and other carriers of private network professional knowledge. Acquisition methods may include, but are not limited to, extraction from internal enterprise knowledge bases, collection from publicly available information from industry standardization organizations, or obtaining textual materials in the private network field through other compliant channels.

[0044] Step 202: Process the relevant data to obtain structured data.

[0045] In some embodiments, processing relevant data refers to organizing the acquired relevant data to conform to a preset structured format. The processing may include, but is not limited to, removing duplicate content from the relevant data, correcting errors, and standardizing the data representation, so that data that may originally be unstructured or semi-structured is converted into structured data with a fixed format, such as an entry format of "knowledge category - content details". The above is only an illustrative example and is not a limitation on a specific format.

[0046] Step 203: Perform vector transformation on the structured data to obtain vector data, and construct the local private network vector library based on the vector data.

[0047] In some embodiments, vector transformation of structured data refers to converting structured data into a computer-recognizable vector form using vectorization tools. These tools include, but are not limited to, open-source text embedding models. Building a local private network vector library based on vector data involves storing the transformed vector data in a local database and establishing an index linking the vector data with corresponding knowledge fragments. This local database can utilize vector storage tools to manage and retrieve the vector data.

[0048] Using the methods described above, a local private network vector library containing professional knowledge in the private network field can be constructed, ensuring the security of private network data and providing a data foundation for subsequent retrieval of matching target vectors from this vector library.

[0049] Figure 3 A flowchart illustrating a private network data generation method provided in this embodiment is further illustrated. Based on Figure 2 The illustrated embodiment further explains step 202. Figure 3 This may include the following steps: Step 301: Clean the relevant data of the private network data to obtain cleaned relevant data.

[0050] In some embodiments, cleaning the relevant data refers to preprocessing the acquired private network-related data to remove redundant information (such as repeated identical paragraphs), error information (such as incorrect device parameter values), and content unrelated to the private network field (such as irrelevant industry advertising text). At the same time, the format of the relevant data is standardized to obtain cleaned relevant data with accurate content and standardized format.

[0051] Step 302: Classify the cleaned relevant data according to preset classification rules to obtain classified relevant data; wherein, the preset classification rules include at least one of industry type, business scenario, terminal application type and professional technical field.

[0052] In some embodiments, classification according to preset classification rules refers to dividing the cleaned relevant data into corresponding categories based on the content attributes of the relevant data. For example, by industry type, it can be divided into categories such as mining, power, and manufacturing; by business scenario, it can be divided into categories such as production monitoring scenarios, remote equipment control scenarios, and safety inspection scenarios; by terminal application type, it can be divided into categories such as AGV terminals, sensor terminals, and surveillance camera terminals; and by professional technical field, it can be divided into categories such as wireless communication, wired transmission, and network security.

[0053] In actual classification, one or more of the above classification rules can be used to obtain the relevant data after classification.

[0054] Step 303: The classified data is segmented to obtain data segments with preset units as units.

[0055] In some embodiments, segmenting the categorized related data refers to dividing each category of related data into independent segments according to a preset granularity unit. The preset unit includes, but is not limited to, sentences, paragraphs, or combinations of short sentences with complete semantics. For example, a text about "the networking steps of 5G private network in mines" can be segmented into descriptive sentences for each step to obtain multiple data segments based on sentences. Each data segment corresponds to a specific networking operation description. The above is only an illustrative example and is not a limitation on the specific segmentation unit.

[0056] Step 304: Integrate the data fragments to obtain the structured data.

[0057] In some embodiments, integrating data fragments refers to systematically organizing the segmented data fragments according to their respective classification rules and preset units to form structured data with clear hierarchy and relationships. For example, integrating data fragments according to a hierarchical structure of "industry type - business scenario - data fragment" allows each data fragment to be uniquely identified by its respective classification label, forming structured data that facilitates subsequent processing and retrieval.

[0058] The above methods can convert raw private network-related data into structured data, providing a standardized and orderly data foundation for subsequent vector conversion and the construction of a local private network vector library.

[0059] Figure 4 A flowchart illustrating a private network data generation method provided in this embodiment is further illustrated. Based on Figure 1 The illustrated embodiment further explains step 101. Figure 4 This may include the following steps: Step 401: Based on the keywords in the user request, retrieve candidate vectors from the local private network vector library.

[0060] In some embodiments, user requests may take the form of, but are not limited to, text, voice, images, and videos. For non-text user requests, they must first be converted into an information format from which keywords can be extracted. For example, voice is converted to text using speech-to-text technology, and images or videos are extracted using image recognition and video frame parsing. Keywords are key information terms extracted from the information content of the user request, including but not limited to industry names, business types, technical terms, and demand-related terms related to the private network. Keyword extraction can be performed using keyword extraction algorithms in natural language processing (such as word frequency-inverse document frequency extraction methods, semantic analysis-based extraction methods, etc.). Each vector data stored in the local private network vector library is associated with a corresponding keyword tag, which matches the knowledge fragment content associated with the vector data. Keyword-based retrieval refers to matching the keywords extracted from the user request with the keyword tags of the vector data in the local private network vector library, and filtering out the vector data with successfully matched tags, which are the candidate vectors.

[0061] Step 402: Convert the user request into a query vector.

[0062] In some embodiments, when converting a user request into a query vector, if the user request is in text format, it can be directly converted into a vector using a text vectorization tool; if it is in non-text format such as speech, image, or video, it needs to be converted into corresponding text information or feature data first (e.g., speech-to-text, extracting text descriptions from images / videos), and then converted into a computer-recognizable vector format using a vectorization tool. The vectorization tools used include, but are not limited to, open-source text embedding models and vector conversion interfaces in deep learning frameworks, and the final output vector is the query vector.

[0063] Step 403: Calculate the semantic similarity between the query vector and the candidate vector.

[0064] In some embodiments, semantic similarity is used to characterize the degree of semantic association between the text content corresponding to the query vector and the candidate vector. Algorithms for calculating semantic similarity include, but are not limited to, cosine similarity algorithm, Euclidean distance algorithm, and Pearson correlation coefficient algorithm. Specifically, the calculation process involves inputting the query vector and each candidate vector into a selected similarity algorithm, and then using the algorithm to obtain the corresponding similarity value, which is the semantic similarity between the query vector and the candidate vector.

[0065] Step 404: Determine the target vector based on the semantic similarity.

[0066] In some embodiments, determining the target vector based on semantic similarity requires pre-setting similarity conditions. These pre-setting conditions may include, but are not limited to, setting a similarity threshold or setting the number of candidate vectors to be screened. For example, setting a similarity threshold of 0.8, candidate vectors with semantic similarity values ​​greater than or equal to 0.8 are determined as target vectors; or, candidate vectors are sorted from high to low semantic similarity values, and the top N candidate vectors (N is a pre-set positive integer) are selected as target vectors. The specific setting of the pre-setting conditions can be adjusted according to the actual application scenario, and this disclosure does not limit this.

[0067] The above method can adapt to various forms of user requests, narrow the search scope by keywords, and then accurately filter by semantic similarity, so as to achieve efficient and accurate determination of target vectors and provide reliable support for subsequent acquisition of target knowledge fragments.

[0068] Figure 5 A flowchart illustrating a private network data generation method provided in this disclosure embodiment is further illustrated. Figure 1 The illustrated embodiment further explains step 103. Figure 5 This may include the following steps: Step 501: Standardize the user request to obtain the demand information in the user request.

[0069] In some embodiments, standardizing user requests refers to converting user-input requests into standardized expressions. Standardization includes, but is not limited to: correcting spelling errors in the request, standardizing the expression of technical terms, completing omitted sentence components to achieve semantic completeness, and converting colloquial expressions into written expressions. Through standardization, clearly structured and semantically unambiguous requirement information can be obtained.

[0070] Step 502: Based on the demand information, filter the target knowledge fragments to determine the knowledge content related to the demand information.

[0071] In some embodiments, filtering target knowledge fragments based on demand information refers to selecting the most relevant content from all target knowledge fragments according to the key elements in the demand information. The filtering process includes, but is not limited to: calculating the semantic relevance between the demand information and each target knowledge fragment, and selecting target knowledge fragments with a relevance exceeding a preset threshold; or selecting target knowledge fragments that match the industry type, technical field, or other classification tags in the demand information. Filtering can determine the knowledge content most relevant to the demand information.

[0072] Step 503: Integrate the demand information and the filtered knowledge content according to a preset logic, and add an instruction to generate the target private network data to obtain the prompt information.

[0073] In some embodiments, preset logic integration refers to organizing requirement information and knowledge content according to a fixed template structure. Integration methods include, but are not limited to: placing the filtered knowledge content as background information at the beginning of the prompt message, placing the requirement information as a specific task description in the middle, and finally adding explicit generation instructions. These instructions include, but are not limited to, guiding statements such as "Please generate a private network solution based on the above knowledge" and "Output device configuration according to the above content." Through integration and the addition of instructions, a complete prompt message is formed.

[0074] The above methods can effectively combine user needs with professional knowledge to generate well-structured and clearly defined prompts, providing a solid foundation for generating accurate target private network data for the target large language model.

[0075] Figure 6 A flowchart illustrating a private network data generation method provided in this disclosure embodiment is further shown, such as... Figure 6 As shown, the method includes steps 601-605.

[0076] Step 601: In response to the received user request, determine the target vector that matches the user request from the local private network vector library.

[0077] Step 602: Obtain the target knowledge fragment corresponding to the target vector.

[0078] Step 603: Generate prompt information based on the target knowledge fragment and the user request.

[0079] For explanations of steps 601 to 603, please refer to [link / reference needed]. Figure 1 The detailed descriptions of the relevant embodiments are not repeated here.

[0080] Step 604: Adjust the large language model using a lightweight fine-tuning method to obtain the target large language model.

[0081] In some embodiments, lightweight fine-tuning refers to techniques for efficiently adjusting the parameters of a large language model, including but not limited to low-rank adaptive methods, adapter methods, or prefix tuning methods. The adjustment process involves introducing a small number of trainable parameters into the large language model while keeping most of its parameters unchanged. Training data from the private network domain is then used to train the newly added parameters, enabling the large language model to acquire knowledge understanding and generation capabilities within the private network domain, thereby obtaining a target large language model adapted for the private network domain.

[0082] Step 605: Process the prompt information using the target large language model to obtain the target private network data.

[0083] For instructions on step 605, please refer to [link / reference needed].Figure 1 The detailed descriptions of the relevant embodiments are not repeated here.

[0084] By employing the above methods, while maintaining the general capabilities of the basic large language model, it is possible to enable it to process professional knowledge in the private network field, thereby improving the professionalism and accuracy of the target private network data generation, and reducing the computing resources and time costs required for adjusting the large language model.

[0085] In some possible ways, Figure 7 This is a schematic diagram of the overall architecture of a private network data generation method provided in an embodiment of this disclosure, as shown below. Figure 7 As shown, the architecture is divided into three modules: private network solution data preparation, data retrieval, and large language model (LLM) generation. Each module generates private network data through data flow.

[0086] In the private network solution data preparation module, after obtaining the relevant data of the private network solution, the relevant data of the private network solution is converted into private network text vectors through text data mapping, and then the private network text vectors are stored in the local private network vector library to complete the vectorized storage of private network domain knowledge.

[0087] In the data retrieval module, after receiving a user request, the user request is also converted into a query vector through text data mapping; the query vector is matched and interacted with the local private network vector library to retrieve the knowledge fragment corresponding to the query vector from the local private network vector library.

[0088] In the LLM generation module, prompt information is generated based on the retrieved knowledge fragments. The prompt information is then input into the target large language model. After processing by the target large language model, the target private network data in the form of a private network solution report is output.

[0089] Corresponding to the above-described method for generating private network data, this invention also proposes a device for generating private network data. Since the device embodiments of this invention correspond to the method embodiments described above, details not disclosed in the device embodiments can be referred to in the method embodiments, and will not be repeated here.

[0090] Figure 8 This is a schematic diagram of the structure of a private network data generation device provided in an embodiment of this disclosure, as shown below. Figure 8 As shown, the device includes: The determining unit 71 is configured to, in response to a received user request, determine a target vector matching the user request from a local private network vector library; The first acquisition unit 72 is used to acquire the target knowledge fragment corresponding to the target vector; Generation unit 73 is used to generate prompt information based on the target knowledge fragment and the user request; The first processing unit 74 is used to process the prompt information through the target large language model to obtain target private network data.

[0091] The private network data generation device provided in this embodiment can retrieve target vectors matching user requests from a local private network vector library, generate prompt information by processing the target knowledge fragments corresponding to the target vectors and the user requests, and generate target private network data through large language model processing, thereby improving the efficiency and accuracy of private network data generation.

[0092] Furthermore, in one possible implementation of the embodiments of this disclosure, such as Figure 9 As shown, the device further includes: The second acquisition unit 75 is used to acquire relevant data of the private network data before the determination unit 71 determines the target vector matching the user request from the local private network vector library in response to the received user request. The second processing unit 76 is used to process the relevant data to obtain structured data; The construction unit 77 is used to perform vector transformation on the structured data to obtain vector data, and to construct the local private network vector library based on the vector data.

[0093] Furthermore, in one possible implementation of the embodiments of this disclosure, such as Figure 9 As shown, the second processing unit 76 includes: The cleaning module 761 is used to clean the relevant data of the private network data to obtain cleaned relevant data. The classification module 762 is used to classify the cleaned relevant data according to a preset classification rule to obtain the classified relevant data; wherein, the preset classification rule includes at least one of industry type, business scenario, terminal application type and professional technical field. The segmentation module 763 is used to segment the classified related data to obtain data segments with preset units as units; Integration module 764 is used to integrate the data fragments to obtain the structured data.

[0094] Furthermore, in one possible implementation of the embodiments of this disclosure, such as Figure 9 As shown, the determining unit 71 includes: The retrieval module 711 is used to retrieve candidate vectors from the local private network vector library based on the keywords in the user request; Conversion module 712 is used to convert the user request into a query vector; The calculation module 713 is used to calculate the semantic similarity between the query vector and the candidate vector; The first determining module 714 is used to determine the target vector based on the semantic similarity.

[0095] Furthermore, in one possible implementation of the embodiments of this disclosure, such as Figure 9 As shown, the generation unit 73 includes: Processing module 731 is used to perform normalization processing on the user request to obtain the demand information in the user request; The second determining module 732 is used to filter the target knowledge fragment based on the demand information and determine the knowledge content related to the demand information; The third determining module 733 is used to integrate the demand information and the filtered knowledge content according to a preset logic, and add an instruction to generate the target private network data to obtain the prompt information.

[0096] Furthermore, in one possible implementation of the embodiments of this disclosure, such as Figure 9 As shown, the device further includes: The adjustment unit 78 is used to adjust the large language model in a lightweight fine-tuning manner before the first processing unit 74 processes the prompt information through the target large language model to obtain the target private network data, so as to obtain the target large language model.

[0097] It should be noted that the foregoing explanation of the method embodiments also applies to the apparatus of the embodiments of this disclosure, and the principle is the same. Therefore, the embodiments of this disclosure are not limited thereto.

[0098] According to embodiments of this disclosure, this disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0099] Figure 10 A schematic block diagram of an example electronic device 900 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0100] like Figure 10As shown, the electronic device 900 includes a computing unit 901, which can perform various appropriate actions and processes based on a computer program stored in ROM (Read-Only Memory) 902 or a computer program loaded from storage unit 908 into RAM (Random Access Memory) 903. The RAM 903 can also store various programs and data required for the operation of the electronic device 900. The computing unit 901, ROM 902, and RAM 903 are interconnected via bus 904. An I / O (Input / Output) interface 905 is also connected to bus 904.

[0101] Multiple components in electronic device 900 are connected to I / O interface 905, including: input unit 906, such as keyboard, mouse, etc.; output unit 907, such as various types of displays, speakers, etc.; storage unit 908, such as disk, optical disk, etc.; and communication unit 909, such as network card, modem, wireless transceiver, etc. Communication unit 909 allows electronic device 900 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0102] The computing unit 901 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 901 include, but are not limited to, CPUs (Central Processing Units), GPUs (Graphics Processing Units), various special-purpose AI (Artificial Intelligence) computing chips, various computing units running machine learning model algorithms, DSPs (Digital Signal Processors), and any suitable processor, controller, microcontroller, etc. The computing unit 901 performs the various methods and processes described above, such as the private network data generation method. For example, in some embodiments, the private network data generation method can be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 908. In some embodiments, part or all of the computer program can be loaded and / or installed on the electronic device 900 via ROM 902 and / or communication unit 909. When the computer program is loaded into RAM 903 and executed by the computing unit 901, one or more steps of the methods described above can be performed. Alternatively, in other embodiments, the computing unit 901 may be configured to perform the aforementioned private network data generation method by any other suitable means (e.g., by means of firmware).

[0103] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, FPGAs (Field Programmable Gate Arrays), ASICs (Application-Specific Integrated Circuits), ASSPs (Application-Specific Standard Products), SOCs (System-on-Chips), CPLDs (Complex Programmable Logic Devices), computer hardware, firmware, software, and / or combinations thereof. These various implementations may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0104] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0105] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, RAM, ROM, EPROM (Electrically Programmable Read-Only Memory) or flash memory, optical fiber, CD-ROM (Compact Disc Read-Only Memory), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0106] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (Cathode-Ray Tube) or LCD (Liquid Crystal Display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0107] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or middleware components (e.g., application servers), or frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include LANs (Local Area Networks), WANs (Wide Area Networks), the Internet, and blockchain networks.

[0108] Computer systems can include clients and servers. Clients and servers are generally geographically separated and typically interact via communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. A server can be a cloud server, also known as a cloud computing server or cloud host, a hosting product within the cloud computing service system that addresses the shortcomings of traditional physical hosts and VPS (Virtual Private Server) services, such as high management difficulty and weak business scalability. Servers can also be servers for distributed systems or servers incorporating blockchain technology.

[0109] It's important to note that artificial intelligence (AI) is the study of enabling computers to simulate certain human thought processes and intelligent behaviors (such as learning, reasoning, thinking, and planning). It encompasses both hardware and software technologies. AI hardware technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, and big data processing. AI software technologies primarily include computer vision, speech recognition, natural language processing, machine learning / deep learning, big data processing, and knowledge graph technologies.

[0110] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.

[0111] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A method for generating private network data, characterized in that, include: In response to a received user request, a target vector matching the user request is determined from the local private network vector library; Obtain the target knowledge fragment corresponding to the target vector; Generate prompt information based on the target knowledge fragment and the user request; The target private network data is obtained by processing the prompt information using the target large language model.

2. The method according to claim 1, characterized in that, Before determining the target vector matching the user request from the local private network vector library in response to the received user request, the method further includes: Obtain relevant data from the private network; The relevant data is processed to obtain structured data; The structured data is transformed into vector data to obtain vector data, and the local private network vector library is constructed based on the vector data.

3. The method according to claim 2, characterized in that, The process of processing the relevant data to obtain structured data includes: The relevant data of the private network data is cleaned to obtain cleaned relevant data; The cleaned data is classified according to a preset classification rule to obtain the classified data; wherein the preset classification rule includes at least one of industry type, business scenario, terminal application type and professional technical field. The classified data is segmented to obtain data fragments with preset units; The data fragments are integrated to obtain the structured data.

4. The method according to claim 1, characterized in that, The step of responding to a received user request by determining a target vector matching the user request from a local private network vector library includes: Candidate vectors are obtained by retrieving them from the local private network vector database based on the keywords in the user request; Convert the user request into a query vector; Calculate the semantic similarity between the query vector and the candidate vector; The target vector is determined based on the semantic similarity.

5. The method according to claim 1, characterized in that, The step of generating prompt information based on the target knowledge fragment and the user request includes: The user request is normalized to obtain the demand information in the user request; Based on the demand information, the target knowledge fragments are filtered to determine the knowledge content related to the demand information; The required information and the filtered knowledge content are integrated according to a preset logic, and an instruction to generate the target private network data is added to obtain the prompt information.

6. The method according to any one of claims 1-5, characterized in that, Before processing the prompt information using the target large language model to obtain the target private network data, the method further includes: The target large language model is obtained by adjusting the large language model using a lightweight fine-tuning method.

7. A private network data generation device, characterized in that, include: The determining unit is used to determine, in response to a received user request, a target vector matching the user request from a local private network vector library; The first acquisition unit is used to acquire the target knowledge fragment corresponding to the target vector; A generation unit is used to generate prompt information based on the target knowledge fragment and the user request; The first processing unit is used to process the prompt information through the target large language model to obtain target private network data.

8. An electronic device, characterized in that, include: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-6.

9. A non-transitory computer-readable storage medium storing computer instructions, characterized in that, The computer instructions are used to cause the computer to perform the method according to any one of claims 1-6.

10. A computer program product, characterized in that, Includes a computer program that, when executed by a processor, implements the method according to any one of claims 1-6.

Images