Question and answer method and system, server and storage medium
By introducing question and answer verification into a question-and-answer system that integrates local and cloud-based large-scale models, security and privacy issues during data transmission are resolved, thereby improving the security and privacy of data transmission.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-04-27
- Publication Date
- 2026-03-31
AI Technical Summary
When local large models and cloud-based large models work together, they are vulnerable to tampering, theft, or forgery attacks during data transmission, which threatens the security and privacy of model computation.
During data transmission, question verification information and answer verification information are introduced between the end-side server and the cloud server to verify the target question feature data and the target answer data respectively, ensuring the security and privacy of data transmission.
It improves the security and privacy of data transmission between local and cloud-based large models, ensuring data security for users during the question-and-answer process.
Smart Images
Figure CN121765756A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of data security technology, and in particular to a question-and-answer method, system, server, and storage medium. Background Technology
[0002] With the rapid development of artificial intelligence technology, large-scale models are being used more and more widely in various fields. Traditional centralized cloud service models face challenges such as high latency, bandwidth consumption, and privacy and security, leading to the adoption of architectures that combine local and cloud-based large-scale models. Local large-scale models process data on user devices, enhancing control over data privacy, while cloud-based large-scale models provide powerful computing and storage capabilities to handle large-scale and complex data processing tasks.
[0003] However, the collaboration between local large models and cloud large models inevitably requires data transmission between the edge server and the cloud server. During the data transmission process, the model may be subject to tampering, theft, or forgery attacks, which could threaten the security and privacy of the model computation. Summary of the Invention
[0004] This application provides a question-answering method, system, server, and storage medium to address the security and privacy issues inherent in existing question-answering methods that utilize end-side servers and cloud servers.
[0005] In a first aspect, embodiments of this application provide a question-answering method applied to an end-side server deploying a large local model, the method comprising:
[0006] The user's question content is obtained, and the user's question content is parsed and processed based on the local large model. Question verification information is introduced into the parsing result to obtain the target question feature data.
[0007] The target question feature data is sent to the cloud server, so that the cloud server generates target answer data containing answer verification information based on the target question feature data, and returns the target answer data to the end server.
[0008] The target answer data is validated based on the answer validation information, and a question answer is generated based on the validated target answer data and the user question content using the local large model.
[0009] Secondly, embodiments of this application provide a question-and-answer method applied to a cloud server deploying a large cloud model, the method comprising:
[0010] The target problem feature data containing problem verification information sent by the receiving end-side server;
[0011] The target question feature data is verified based on the question verification information, and the target question feature data that has passed the verification is answered based on the cloud big model, and the answer verification information is introduced to obtain the target answer data.
[0012] The target answer data is returned to the client-side server, so that the client-side server can verify the target answer data according to the answer verification information, and generate a question answer based on the verified target answer data and the user question content based on the local large model.
[0013] Thirdly, embodiments of this application provide a question-answering system, including: an end-side server deploying a local large model and a cloud server deploying a cloud-based large model;
[0014] The edge server obtains the user's question content, parses and processes the user's question content based on the local large model, and introduces question verification information into the parsing result to obtain target question feature data; the target question feature data is then sent to the cloud server.
[0015] The cloud server verifies the target question feature data based on the question verification information, answers the verified target question feature data based on the cloud big model to obtain initial answer data, and introduces answer verification information into the initial answer data to obtain target answer data; the target answer data is then returned to the end-side server.
[0016] The endpoint server verifies the target answer data based on the answer verification information, and generates a question answer based on the verified target answer data and the user question content using a local large model.
[0017] Fourthly, embodiments of this application provide a server, the server comprising:
[0018] At least one processor; and
[0019] A memory communicatively connected to the at least one processor; wherein,
[0020] The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the question-and-answer method described in any embodiment of this application.
[0021] Fourthly, embodiments of this application provide a computer program product that stores computer instructions, which are used to cause a processor to execute and implement the question-and-answer method described in any embodiment of this application.
[0022] The technical solution of this application embodiment obtains target question feature data by acquiring user question content, parsing and processing the user question content, and introducing question verification information. The target question feature data is then sent to a cloud-based large model, which generates target answer data containing answer verification information based on the target question feature data. This target answer data is then returned to the local large model. The cloud-based large model is deployed on a cloud server. The target answer data is verified based on the answer verification information, and a question answer is generated based on the verified target answer data and the user question content from the local large model. By generating question feature data containing question verification information from the local large model for verification by the cloud-based large model, and by verifying the target answer data containing answer verification information generated by the cloud-based large model, the security and privacy issues of existing question-and-answer methods using local and cloud-based large models are resolved. This improves the security and privacy of data transmission between the local and cloud-based large models, ensuring user data security during the question-and-answer process.
[0023] 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
[0024] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0025] Figure 1 A flowchart of a question-and-answer method provided in Embodiment 1 of this application;
[0026] Figure 2 A flowchart of a question-and-answer method provided in Embodiment 2 of this application;
[0027] Figure 3 This is a schematic diagram of the structure of a question-and-answer system provided in Embodiment 2 of this application;
[0028] Figure 4 A schematic diagram of the server structure for implementing the question-and-answer method of this application embodiment. Detailed Implementation
[0029] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.
[0030] It should be noted that the terms "initial," "target," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.
[0031] Example 1
[0032] Figure 1 This is a flowchart of a question-answering method provided in Embodiment 1 of this application. This embodiment is applicable to situations where a local large model on the endpoint server and a cloud large model on the cloud server work together to answer user questions. This method can be executed by the local large model deployed on the endpoint server. Figure 1 As shown, the method includes:
[0033] S110. Obtain the user's question content, parse and process the user's question content based on the local large model, and introduce question verification information into the parsing result to obtain the target question feature data.
[0034] The user question content can be understood as the text content of the question raised by the user, generally described in natural language. User questions can be in text or voice format, and the user question content can be obtained by performing speech recognition on the voice-based user questions.
[0035] The local large model can be understood as a large model deployed locally on the client-side server. It is used for feature processing of user question content and for processing answer data to generate question answers. It can also be used for data verification during transmission. Target question feature data can be understood as question feature data containing question verification information. Question verification information can be understood as information used to verify the question feature data of the user question content, ensuring the security and privacy of question feature data transmission.
[0036] Specifically, the client-side server obtains the user's question content and parses and processes it using a local large model deployed on the client-side server to obtain question feature data. To ensure the security and privacy of the transmitted question feature data, question verification information is introduced into the question feature data to obtain the target question feature data.
[0037] For example, parsing and processing user question content using a local large model may include: performing intent recognition, semantic parsing, and data anonymization processing on the user question content using a local large model.
[0038] S120. Send the target question feature data to the cloud server so that the cloud server can generate target answer data containing answer verification information based on the target question feature data, and return the target answer data to the end server.
[0039] The target answer data can be understood as the content of the answer to the user's question, which includes answer verification information. This verification information is used to validate the answer data, ensuring the security and privacy of the answer data transmission.
[0040] Specifically, the client-side server sends the processed target question feature data, including question verification information, to the cloud server. Upon receiving the target question feature data, the cloud server verifies the verification information contained within it and generates answer data based on the verified target question feature data. Similarly, to ensure the security and privacy of answer data transmission, the cloud server incorporates answer verification information into the answer data to obtain target answer data, which is then returned to the client-side server. For example, the cloud server can generate answer data based on the verified target question feature data using a large cloud-based model deployed on the cloud server.
[0041] Optionally, before the client-side server and the cloud server interact with each other, the process further includes: the client-side server sending an initial connection request message to the cloud server, so that the cloud server responds to the client-side server's request by generating an initial connection response message, and calling the cloud server to send the initial connection response message to the client-side server where the local large model is deployed, thus completing the initialization process between the client-side server and the cloud server and providing a foundation for question answering using the local large model and the cloud large model.
[0042] S130. Verify the target answer data based on the answer verification information, and generate the question answer based on the verified target answer data and the user question content using the local large model.
[0043] In this context, "question answer" can be understood as the answer to a user's question.
[0044] Specifically, when the client-side server receives the target answer data returned by the cloud server, it verifies the target answer data according to the answer verification information, and generates the question answer based on the verified target answer data and the user's question content using the local large model.
[0045] For example, generating a question answer based on validated target answer data and user question content using a local large-scale model could involve inputting the target answer data and user question content into a prompt information template, generating prompt information, and then processing the target answer data based on the prompt information using the local large-scale model to generate the question answer. Processing the target answer data could include, for example, outputting a question answer that meets the user's needs according to the output format and content requirements of the user question content.
[0046] The technical solution of this application embodiment obtains user question content, parses and processes the user question content based on a local large model, and introduces question verification information into the parsing result to obtain target question feature data. The target question feature data is then sent to a cloud server, which generates target answer data containing answer verification information based on the target question feature data and returns the target answer data to the client server. The target answer data is verified based on the answer verification information, and a question answer is generated based on the verified target answer data and the user question content using the local large model. By generating question feature data containing question verification information on the client server for verification by the cloud server, and by verifying the target answer data containing answer verification information generated by the cloud server, the security and privacy of data transmission between the client server and the cloud server are improved, ensuring the data security of the user during the question-and-answer process.
[0047] As an optional embodiment of this application, the step of parsing the user question content based on the local large model and introducing question verification information into the parsing result to obtain target question feature data includes:
[0048] S111. Based on the local large model, the user's question content is parsed and processed to obtain initial question feature data.
[0049] The initial problem feature data can be understood as the feature data obtained by parsing the user's problem content using a large local model deployed on the client-side server.
[0050] Specifically, the local large model parses and processes the user's question content to obtain initial question feature data. The client-side server combines the initial question feature data with the verification information and sends it to the cloud server to generate question answer data. This eliminates the need to send the original user question content directly to the cloud server, thereby reducing security risks during the transmission of user question content and enhancing control over data privacy.
[0051] S112. Generate problem verification information for the initial problem feature data, and add the problem verification information to the initial problem feature data to obtain target problem feature data.
[0052] Specifically, although the original user problem content does not need to be transmitted to the cloud server, reducing security risks during transmission, the transmission of initial problem feature data still presents security concerns. Therefore, problem verification information is generated from the initial problem feature data. This verification information is then added to the initial problem feature data to obtain the target problem feature data. This allows the cloud server to verify the target problem feature data, thereby ensuring the security of problem feature data transmission.
[0053] This embodiment improves the security of problem feature data transmission by adding problem verification information to the problem feature data, thus preventing the data from being compromised by tampering, theft, or forgery attacks during transmission.
[0054] As an optional embodiment of this application, the generation of problem verification information includes at least one of steps A1 to A3:
[0055] A1. Calculate the hash digest based on the initial problem feature data and the custom hash parameters to obtain the problem data hash digest.
[0056] Custom hash parameters can be understood as parameters used to customize the hash function, enhancing security, enabling personalized data processing, or constructing complex tree-like hash structures. Custom hash parameters include, but are not limited to, the key, personalization string, salt, and digest size. The salt is generally a randomly generated data fragment, primarily intended to prevent attackers from cracking passwords using rainbow tables (pre-calculated hash value tables), because even if two users use the same password, the calculated hash values will differ due to different salt values. The digest size of the hash digest algorithm can generally be set between 8 bytes (64 bits) and 64 bytes (512 bits). Since custom hash parameters affect the calculated hash digest, and thus the verification result of the cloud server, the client-side server needs to negotiate the custom hash parameters with the cloud server. This application does not limit the hash digest algorithm used, such as MD5, BLAKE2, and Whirlpool.
[0057] Specifically, based on the large model, a preset hash digest algorithm is used to calculate the hash digest of the initial problem feature data according to the custom hash parameters, so as to obtain the problem data hash digest, which is used to verify the integrity of the problem feature data and prevent the problem data from being tampered with.
[0058] For example, choosing the BLAKE2b hash digest algorithm, the formula can be expressed as: H(m) = BLAKE2(m), where H(m) represents the hash digest of message m, and BLAKE2(m) represents the result calculated by inputting message m into the BLAKE2 algorithm. Initial problem feature data and custom hash parameters are passed to the BLAKE2 algorithm for hash digest calculation. The BLAKE2 algorithm processes the message and generates a hash digest of the specified length.
[0059] A2. Generate a problem transmission timestamp based on the time when the initial problem feature data is obtained, and set the validity period of the problem transmission timestamp.
[0060] The issue transmission timestamp can be understood as a timestamp used to mark the time when issue characteristic data was transmitted. The issue transmission timestamp can be in seconds or milliseconds based on Unix time. The issue transmission timestamp is valid within its validity period.
[0061] For example, the endpoint server and cloud server are configured with a time synchronization service. When processing the initial problem feature data obtained from the local large model, the endpoint server generates a timestamp to approximate the time the problem feature data was sent, and sets a validity period for the problem transmission timestamp. The cloud server considers the target problem feature data received within the validity period as valid problem feature data; otherwise, it considers it invalid. By setting the problem transmission timestamp and validity period, replay attacks during the transmission of problem feature data can be prevented.
[0062] A3. Generate a random number of a preset length as a one-time question token.
[0063] A one-time issue token can be understood as a string of random numbers or alphanumeric combinations generated based on the initial issue characteristic data, and can only be used once. The length of a one-time issue token can be set as needed, such as 32 bits or 64 bits.
[0064] For example, the client-side server uses a random number generator to generate a random number as a one-time question token. The one-time question token can only be used once, thus possessing uniqueness, which can be ensured through a predefined question token verification mechanism. Furthermore, in this embodiment, the one-time question token can serve as a unique identifier for the user's question content during the question-answering process.
[0065] Based on the above embodiments, the cloud server may verify the target problem feature data according to the problem verification information in the following ways: calculate the hash digest of the data contained in the target problem feature data according to a preset hash digest algorithm, and determine whether the calculated hash digest is the same as the hash digest contained in the problem verification information; determine the time difference between the time when the target problem feature data is received and the time corresponding to the problem transmission timestamp in the problem verification information, and determine whether the time difference is within the validity period corresponding to the problem transmission timestamp; and determine whether the one-time problem token is received repeatedly.
[0066] This embodiment improves the security of problem data transmission by adding at least one of the following to the initial problem feature data transmitted between the end server and the cloud server: a problem data hash digest, a problem transmission timestamp, and a one-time problem token.
[0067] As an optional embodiment of this application, after sending the target problem feature data to the cloud server, the method further includes:
[0068] If a challenge is received from the cloud server, the challenge is answered to obtain a challenge response, which is then returned to the cloud server.
[0069] Specifically, a challenge-response protocol-based authentication mechanism is added to the data transmission between the client-side server and the cloud server. This authentication mechanism mainly includes a challenge and a challenge response. The authenticator sends a challenge to the authenticee, who must provide a correct challenge response to pass the authentication.
[0070] After the client-side server sends the target question feature data to the cloud server, the cloud server, acting as the authenticator, sends a challenge to the client-side server (actually the authenticated party) upon receiving the target question feature data. Upon receiving the challenge from the cloud server, the client-side server responds to the challenge, obtains a challenge response, and returns the challenge response to the cloud server, enabling the cloud server to authenticate the legitimacy of the client-side server's identity.
[0071] This embodiment can prevent replay attacks by setting up a challenge-response based authentication mechanism, because the challenge for each authentication is randomly generated, making previous responses unusable in subsequent authentications; it can also perform identity authentication on the end-side server to prevent the theft or forgery of target problem feature data, thereby further improving the security of data transmission, and can flexibly adapt to communication environments of different scales and complexities, with strong scalability.
[0072] Example 2
[0073] Figure 2 This is a flowchart of a question-and-answer method provided in Embodiment 2 of this application. This embodiment is applicable to situations where a local large-scale model is deployed on an end-side server, and a cloud server is deployed on a cloud-based large-scale model to answer user questions. This method can be executed by the cloud server deploying the cloud-based large-scale model. Figure 2 As shown, the method includes:
[0074] S210, Target problem feature data containing problem verification information sent by the receiving end-side server.
[0075] Specifically, the client-side server parses and processes the user's question content based on the local large model, and incorporates question verification information into the parsing results to obtain the target question feature data.
[0076] S220. Verify the target problem feature data based on the problem verification information.
[0077] Specifically, after receiving the target problem feature data, the cloud server verifies the target problem feature data based on the problem verification information contained within it. For example, the verification of the problem feature data can be performed using a verification algorithm, verification model, or verification tool, based on predefined verification data and rules; this embodiment of the invention does not impose any limitations on this.
[0078] S230. Based on the cloud-based large model, answer the target question feature data that has passed the verification to obtain initial answer data; and introduce answer verification information into the initial answer data to obtain target answer data.
[0079] Here, the cloud-based large model can be understood as a large model deployed on a cloud server, used to answer user questions. Initial answer data can be understood as the cloud-based large model's response to the target question's feature data. Target answer data can be understood as data incorporating answer verification information. Answer verification information can be understood as information used to verify the answer.
[0080] Specifically, if the verification passes, the target question feature data is input into the prompt information template using the cloud-based big model deployed on the cloud server to obtain prompt information. Based on the prompt information, the cloud-based big model deployed on the cloud server retrieves relevant knowledge to obtain initial answer data.
[0081] To ensure the security of transmitting initial response data from the cloud server to the client server, response verification information needs to be included in the initial response data. Therefore, response verification information is added to the initial response data to obtain the target response data, which is then verified by the client server to ensure the security of response data transmission.
[0082] This embodiment improves the security of response data transmission by adding response verification information to the response data, thus preventing the data from being compromised or stolen during transmission due to tampering, theft, or forgery attacks.
[0083] S240. Return the target answer data to the end-side server so that the end-side server can verify the target answer data according to the answer verification information, and generate the question answer based on the verified target answer data and the user question content based on the local large model.
[0084] Specifically, the cloud server returns the generated target answer data containing answer verification information to the local large model, so that when the client server receives the target answer data, it can verify the target answer data according to the answer verification information, and generate the question answer based on the verified target answer data and the user question content based on the local large model.
[0085] The technical solution of this invention involves a cloud server receiving target question feature data containing question verification information from an end-server. The cloud server verifies the target question feature data based on the verification information, and then uses a cloud-based large model deployed on the cloud server to answer the verified target question feature data, obtaining initial answer data. Answer verification information is then incorporated into the initial answer data to obtain target answer data. The target answer data is returned to the end-server, allowing the end-server to verify the target answer data based on the answer verification information and generate a question answer based on the verified target answer data and the user's question content using a local large model. By deploying a cloud-based large model on the cloud server to verify and answer the target question feature data containing question verification information generated by the end-server, and generating target answer data containing answer verification information for the end-server to verify, the security and privacy of data transmission between the end-server and the cloud server are improved, ensuring user data security during the question-and-answer process.
[0086] As an optional embodiment of this application, after the target problem feature data containing problem verification information sent by the receiving end-side server, it further includes:
[0087] Send a challenge to the endpoint server so that the endpoint server can respond to the challenge, obtain the challenge response, and return the challenge response to the cloud server;
[0088] Accordingly, the target problem feature data is validated based on the problem validation information, including:
[0089] Verify the challenge response returned by the cloud server;
[0090] If the challenge response is verified to be correct, the target response data will be verified based on the response verification information.
[0091] Specifically, a challenge-response protocol-based authentication mechanism is added to the data transmission between the endpoint server and the cloud server. This authentication mechanism mainly includes a challenge and a challenge response. After receiving the target question feature data, the cloud server, acting as the authenticator, sends a challenge to the endpoint server, which is the authenticated party. Upon receiving the challenge from the cloud server, the endpoint server responds to it, obtaining a challenge response, and returns the response to the cloud server. The cloud server verifies the correctness of the challenge response to authenticate the endpoint server's identity. If the challenge response is correct, the endpoint server's authentication is successful, and the target question feature data is then verified based on the question verification information.
[0092] This embodiment can prevent replay attacks by setting up a challenge-response based authentication mechanism, because the challenge for each authentication is randomly generated, making previous responses unusable in subsequent authentications; it can also perform identity authentication on the end-side server to prevent the theft or forgery of problem feature data, thereby further improving the security of data transmission, and can flexibly adapt to communication environments of different scales and complexities, with strong scalability.
[0093] As an optional embodiment of this application, the step of generating the answer verification information includes at least one of steps B1 to B3:
[0094] B1. Calculate the hash digest based on the initial answer data and the custom hash parameters to obtain the answer data hash digest.
[0095] Specifically, based on the large model, a preset hash digest algorithm is used to calculate the hash digest of the initial answer data according to the custom hash parameters, so as to obtain the answer data hash digest, which is used to verify the integrity of the answer data and prevent the answer data from being tampered with.
[0096] B2. Generate a response transmission timestamp based on the time the target response data is received, and set the validity period of the response transmission timestamp.
[0097] The response transmission timestamp can be understood as a timestamp used to mark the time when the response data was transmitted. The response transmission timestamp can be in seconds or milliseconds based on Unix time. The response transmission timestamp is valid within its validity period.
[0098] For example, the client-side server and the cloud server are configured with a time synchronization service. When generating initial response data, the cloud server generates a timestamp to approximate the time the response data was sent, and sets a validity period for the response transmission timestamp. Target response data received by the client-side server within the validity period is considered valid; otherwise, it is considered invalid. Setting the response transmission timestamp and validity period can prevent replay attacks during response data transmission.
[0099] B3. Generate a random number of a preset length as a one-time answer token.
[0100] A one-time response token can be understood as a random string of numbers or alphanumeric combinations generated for the initial response data, and can only be used once. The length of the one-time response token can be set as needed, such as 32 bits or 64 bits.
[0101] For example, the cloud server uses a random number generator to generate a random number as a one-time answer token. The one-time answer token can only be used once, and therefore is unique. This uniqueness can be ensured through a predefined answer token verification mechanism. Furthermore, in this embodiment, the one-time answer token can serve as a unique identifier for the answer content during the question-and-answer process.
[0102] Based on the above embodiments, the method by which the end-side server verifies the target answer data according to the answer verification information may include: calculating the hash digest of the data contained in the target answer data according to a preset hash digest algorithm, and determining whether the calculated hash digest is the same as the hash digest contained in the answer verification information; determining the time difference between the time when the target answer data is received and the time corresponding to the answer transmission timestamp in the answer verification information, and determining whether the time difference is within the validity period corresponding to the answer transmission timestamp; and determining whether the one-time answer token is received repeatedly.
[0103] This embodiment improves the security of response data transmission by adding at least one of the following to the initial response data transmitted between the client server and the cloud server: response data hash digest, response transmission timestamp, and one-time response token.
[0104] Example 3
[0105] Figure 3 This is a schematic diagram of the structure of a question-and-answer system provided in Embodiment 3 of this application. Figure 3 As shown, the question-and-answer system includes: an end-side server 310 that deploys a local large model and a cloud server 320 that deploys a cloud large model;
[0106] The edge server 310 obtains the user's question content, parses and processes the user's question content based on the local large model, and introduces question verification information into the parsing result to obtain the target question feature data; and sends the target question feature data to the cloud server.
[0107] The cloud server 320 verifies the target question feature data based on the question verification information; it answers the target question feature data that has passed the verification based on the cloud big model to obtain initial answer data, and introduces answer verification information into the initial answer data to obtain target answer data; the target answer data is then returned to the end server.
[0108] The client-side server 310 verifies the target answer data based on the answer verification information, and generates the question answer based on the verified target answer data and the user question content using the local large model.
[0109] The question-and-answer system provided in this application can execute the question-and-answer method provided in any embodiment of this application, and has the corresponding functional modules and beneficial effects of the execution method.
[0110] Example 4
[0111] Figure 4 A schematic diagram of the structure of a server 10 that can be used to implement embodiments of this application is shown. The server is intended to represent various forms of servers. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of this application described and / or claimed herein.
[0112] like Figure 4 As shown, server 10 includes at least one processor 11 and memory, such as read-only memory (ROM) 12 and random access memory (RAM) 13, communicatively connected to at least one processor 11. The memory stores computer programs executable by at least one processor. Processor 11 can perform various appropriate actions and processes based on the computer program stored in ROM 12 or loaded into RAM 13 from storage unit 18. RAM 13 can also store various programs and data required for the operation of server 10. Processor 11, ROM 12, and RAM 13 are interconnected via bus 14. Input / output (I / O) interface 15 is also connected to bus 14.
[0113] Multiple components in server 10 are connected to I / O interface 15, including: input unit 16, such as keyboard, mouse, etc.; output unit 17, such as various types of monitors, speakers, etc.; storage unit 18, such as disk, optical disk, etc.; and communication unit 19, such as network card, modem, wireless transceiver, etc. Communication unit 19 allows server 10 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.
[0114] Processor 11 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. Processor 11 performs the various methods and processes described above, such as question-answering methods.
[0115] In some embodiments, the question-and-answer method may be implemented as a computer program tangibly contained in a computer program product, such as storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or installed on server 10 via ROM 12 and / or communication unit 19. When the computer program is loaded into RAM 13 and executed by processor 11, one or more steps of the question-and-answer method described above may be performed. Alternatively, in other embodiments, processor 11 may be configured to execute the question-and-answer method by any other suitable means (e.g., by means of firmware).
[0116] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments 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.
[0117] In some embodiments, the question-answering method may be implemented as a computer program, which is implicitly included in a computer program product. When executed by a processor, the computer program implements the question-answering method of this application. The computer program product can be understood as a software product that primarily implements its solution through a computer program. The computer program used to implement the method of this application may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the functions / operations specified in the flowcharts and / or block diagrams are implemented. The computer program may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.
[0118] In the context of this application, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage 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. Alternatively, a computer-readable storage medium can be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.
[0119] To provide interaction with the user, the systems and techniques described herein can be implemented on a server 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 server. 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).
[0120] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or computing systems that include middleware components (e.g., application servers), or computing systems that include 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 local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.
[0121] A computing system can include endpoint servers and servers. Endpoint servers and servers are generally geographically separated and typically interact via a communication network. The relationship between endpoint servers and servers is created by computer programs running on the respective computers and having endpoint server-server relationships with each other. Servers can be cloud servers, also known as cloud computing servers or cloud hosts, which are hosting products within the cloud computing service ecosystem to address the shortcomings of traditional physical hosting and VPS services, such as high management difficulty and weak business scalability.
[0122] 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 application can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this application can be achieved, and this is not limited herein.
[0123] The specific embodiments described above do not constitute a limitation on the scope of protection of this application. 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 application should be included within the scope of protection of this application.
Claims
1. A question and answer method, characterized by, The method applied to a terminal-side server deploying a local large model comprises: Obtaining user question content, performing parsing processing on the user question content based on the local large model, and introducing question verification information in the parsing result to obtain target question feature data; Sending the target question feature data to a cloud-side server to enable the cloud-side server to generate target answer data containing answer verification information according to the target question feature data and return the target answer data to the terminal-side server; According to the answer verification information, the target answer data is verified, and based on the local large model, a question answer is generated according to the target answer data that passes the verification and the user question content.
2. The method of claim 1, wherein, The method applied to a terminal-side server deploying a local large model comprises: Based on the local large model, the initial question feature data is obtained by performing parsing processing on the user question content; The question verification information of the initial question feature data is generated, and the target question feature data is obtained by adding the question verification information to the initial question feature data.
3. The method of claim 2, wherein, The question verification information of the initial question feature data includes at least one of the following: According to the initial question feature data and a self-defined hash parameter, a hash digest is calculated to obtain question data hash digest; According to the time of obtaining the initial question feature data, a question transmission timestamp is generated, and an effective period of the question transmission timestamp is set; A random number of a preset length is generated as a one-time question token.
4. The method of claim 1, wherein, After sending the target question feature data to the cloud-side server, the method further comprises: If a challenge sent by the cloud-side server is received, the challenge is answered to obtain a challenge response, and the challenge response is returned to the cloud-side server.
5. A question and answer method characterized by, The method applied to a terminal-side server deploying a local large model comprises: Receiving target question feature data containing question verification information sent by a terminal-side server; According to the question verification information, the target question feature data is verified; Based on the cloud-side large model, the initial answer data is obtained by answering the target question feature data that passes the verification, and the answer verification information is introduced into the initial answer data to obtain the target answer data; The target answer data is returned to the terminal-side server to enable the terminal-side server to verify the target answer data according to the answer verification information, and based on the local large model, a question answer is generated according to the target answer data that passes the verification and user question content.
6. The method of claim 5, wherein, After receiving the target question feature data containing question verification information sent by the terminal-side server, the method further comprises: Sending a challenge to the terminal-side server to enable the terminal-side server to answer the challenge to obtain a challenge response, and returning the challenge response to the cloud-side server; Correspondingly, the verification of the target question feature data according to the question verification information comprises: Verifying the challenge response returned by the cloud-side server; If the challenge response is verified to be correct, the target answer data is verified according to the answer verification information.
7. The method of claim 5, wherein, The generation of the answer verification information includes at least one of the following: Hash digest calculation is performed according to the initial answer data and a self-defined hash parameter to obtain answer data hash digest; An answer transmission timestamp is generated according to the time when the target answer data is received, and an effective period of the answer transmission timestamp is set; A random number of a preset length is generated as a one-time answer token.
8. A question answering system, characterized by The server includes an end-side server and a cloud-side server; The end-side server obtains user question content, performs analysis and processing on the user question content based on a local large model, and introduces question verification information in the analysis result to obtain target question feature data; The target question feature data is sent to the cloud-side server; The cloud-side server verifies the target question feature data according to the question verification information, obtains initial answer data by answering the target question feature data that passes the verification based on a cloud-side large model, and introduces answer verification information in the initial answer data to obtain target answer data; The target answer data is returned to the end-side server; The end-side server verifies the target answer data according to the answer verification information, and generates question answers based on a local large model according to the target answer data that passes the verification and user question content.
9. A server, characterized by The server includes: at least one processor; and a memory connected in communication with the at least one processor; wherein The memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor to enable the at least one processor to execute the question and answer method of any one of claims 1-7.
10. A computer program product, characterised in that, The computer program product stores computer instructions for enabling the processor to execute the question and answer method of any one of claims 1-7 when executed.