Information query method, device and system

Abstract

The invention discloses an information query method, device and system, and relates to the technical field of artificial intelligence, first knowledge information matched with knowledge information carried in a query request can be obtained from a knowledge base in response to the query request, the first knowledge information is knowledge information related to the query request, and the first knowledge information is matched with the knowledge information carried in the query request. According to the method and the device, prompt information related to the query request is generated according to the first knowledge information, the prompt information is used for supplementing the query request, and a query result corresponding to the query request is output according to the query request and the prompt information, so that the query request can be supplemented and optimized through the knowledge information related to the task; and the accuracy of the query result during information query according to the query request is improved.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and particularly relates to an information query method, device, and system. Background Art

[0002] With the development of artificial intelligence, the application scenarios of large network models are becoming increasingly rich. Therefore, higher requirements are placed on the performance of large network models in various fields. Due to the huge number of parameters in large network models, the training cost is expensive during the training process of large network models, and it is difficult to ensure the stability of the model effect. Based on this, in related technologies, a pre-training method is used to train large network models and provide services externally. When a large network model is pre-trained, due to the limitation of the word segments allowed in the single session control of the large network model, the large network model cannot combine the relevant information of the task execution environment when performing tasks, resulting in inaccurate execution results of the large network model. For the same task, it needs to be executed repeatedly many times, reducing the efficiency and accuracy of the large network model in the application process. Summary of the Invention

[0003] This application provides an information query method, device, and system, which can supplement and optimize a query request through task-related knowledge information, improving the accuracy of the query result.

[0004] To achieve the above object, this application adopts the following technical solutions.

[0005] In a first aspect, this application provides an information query method, which is applied to an information query system. The information query system includes a knowledge base, and the knowledge base includes knowledge information related to one or more query tasks. The method may include:

[0006] First, in response to a query request, obtain first knowledge information that matches the query request from the knowledge base; this step is mainly used to, when receiving a query request, according to the knowledge information carried in the query request, obtain the first knowledge information related to the query request in the knowledge base. The first knowledge information is one or more knowledge information in the knowledge base, and the first knowledge information can expand the knowledge information carried in the query request.

[0007] Then, generate prompt information related to the query request according to the first knowledge information; this step is mainly used to determine the prompt information related to the first knowledge information according to the first knowledge information obtained in the previous step. The prompt information may include one or more, and the prompt information is used to supplement the query content carried in the query request to make the query content carried in the query request clearer.

[0008] Finally, output the query result corresponding to the query request according to the query request and the prompt information. This step is mainly used to combine the query request and the prompt information related to the query request to obtain the query result corresponding to the query request and output it.

[0009] As a possible implementation manner, obtaining the first knowledge information matching the query request from the knowledge base as described above may include: obtaining the first knowledge information matching the query request from the knowledge base according to a first preset condition, where the first preset condition is related to one or more of the following: knowledge relevance, time relevance, and type relevance. Based on this, the relevance level of the knowledge information in the knowledge base to the query request can be judged from different dimensions through the knowledge information carried in the query request, and the knowledge information with high relevance to the knowledge information carried in the query request is used as the first knowledge information. Among them, when judging the relevance between the knowledge information carried in the query request and the knowledge information in the knowledge base, it can be judged by combining one or more of knowledge relevance, time relevance, and type relevance, or it can be adjusted according to the specific business. It should be noted that it can also be judged by other attributes, and no limitation is made here.

[0010] As a possible implementation manner, the query request carries query content, and obtaining the first knowledge information matching the query request from the knowledge base according to the first preset condition may include: using the knowledge information whose update content has a knowledge relevance greater than or equal to a preset threshold to the query content as the first knowledge information. Based on this, the first knowledge information can be determined according to the knowledge relevance between the knowledge information in the query request and the knowledge information in the knowledge base. Specifically, the knowledge information with a knowledge relevance greater than or equal to the preset threshold can be used as the first knowledge information, where the knowledge relevance indicates the semantic similarity degree between the update content in the knowledge information in the knowledge base and the query content in the knowledge information in the query request.

[0011] As a possible implementation manner, the query request carries a query time, and obtaining the first knowledge information matching the query request from the knowledge base according to the first preset condition includes: using the knowledge information whose update time has a time relevance greater than or equal to a preset threshold to the query time as the first knowledge information. Based on this, the first knowledge information can be determined according to the time relevance between the knowledge information in the query request and the knowledge information in the knowledge base. Specifically, the knowledge information with a time relevance greater than or equal to the preset threshold can be used as the first knowledge information, where the time relevance indicates the time proximity degree between the update time in the knowledge information in the knowledge base and the query time in the knowledge information in the query request.

[0012] As a possible implementation manner, the query request carries a query type, and obtaining the first knowledge information matching the query request from the knowledge base according to the first preset condition includes: using the knowledge information whose type correlation between the update type of the knowledge information and the query type is greater than or equal to a preset threshold as the first knowledge information. Based on this, the first knowledge information can be determined according to the type correlation between the knowledge information in the query request and the knowledge information in the knowledge base, and the knowledge information whose type correlation is greater than or equal to the preset threshold is used as the first knowledge information. Herein, the knowledge correlation indicates the type similarity degree between the update type in the knowledge information in the knowledge base and the query type in the knowledge information in the query request.

[0013] As a possible implementation manner, the query request includes a task identifier, and the knowledge base includes multiple sub-knowledge bases corresponding to different task identifiers. Obtaining the first knowledge information matching the query request from the knowledge base includes: according to the task identifier, obtaining the sub-knowledge base matching the query request from the multiple sub-knowledge bases; and obtaining the first knowledge information matching the query request from the sub-knowledge base. Based on this, the knowledge base may include multiple sub-knowledge bases. The multiple sub-knowledge bases may be set in the same knowledge base set or distributed at different positions according to task requirements, and the multiple sub-knowledge bases correspond to different task identifiers. The sub-knowledge base corresponding to the query request can be determined according to the task identifier in the query request, and the first knowledge information is obtained from the sub-knowledge base corresponding to the task identifier carried in the query request. The correlation degree between the first knowledge information and the knowledge information carried in the query request is greater than or equal to a preset threshold. Specifically, the standard for obtaining the first knowledge information is the same as that in the foregoing embodiments and will not be elaborated herein.

[0014] As a possible implementation manner, generating the prompt information related to the query request according to the first knowledge information includes: based on a prompt logic algorithm, generating the prompt information related to the query request according to the first knowledge information, and the prompt information is used to supplement the query content carried in the query request. Based on this, one or more first knowledge information obtained from the knowledge base have a high correlation degree with the knowledge information carried in the query request. The relevant content of the first knowledge information can be converted into prompt information through the prompt logic algorithm. Herein, the prompt logic algorithm is used to convert the first knowledge information into simpler prompt information. The prompt logic algorithm includes a prompt logic function, which makes supplements or restrictions to the query content carried in the query request, so that the query content is clearer and more complete, facilitating the understanding and query of the query content. For a query request, there may be one or more prompt information, which perfect the query content from different levels and are not limited herein.

[0015] As a possible implementation method, the above-mentioned prompt information related to the query request is generated according to the first knowledge information based on the prompt logic algorithm, including: based on the prompt logic algorithm, obtaining multiple candidate prompt information corresponding to the first knowledge information; determining the prompt information from the one or more candidate prompt information, and the prompt information is the candidate prompt information whose similarity with the query content is greater than or equal to a preset threshold value among the multiple candidate prompt information. Based on this, the first knowledge information can be converted into one or more candidate prompt information according to the logical prompt algorithm. When there is one candidate prompt information, the candidate prompt information can be directly used as the prompt information. When the candidate prompt information includes multiple candidates, it is necessary to screen among the multiple candidate prompt information to determine the candidate prompt information with a high degree of matching with the query request as the prompt information. In some examples, the corresponding candidate prompt information can be determined as the prompt information by the similarity between the candidate prompt information and the query content in the query request being greater than a preset threshold. The prompt information can be one or more, and different screening criteria for the prompt information can be set according to specific needs, and no limitation is made here.

[0016] As a possible implementation method, the query request carries the device identification of the network device that initiated the query request, and the query result corresponding to the query request is output according to the query request and the prompt information, including: obtaining the query result corresponding to the query request according to the query request and the prompt information; and outputting the query result to the network device according to the device identification. Based on this, the query request carries the device identification of the network device that sent the query request, wherein the network device may include a terminal device and a network element device, etc. that are capable of executing the query request, which are not limited here. The query result corresponding to the query request is obtained according to the query content in the knowledge information carried in the query request combined with the supplement and limitation of the query content by the prompt information, and the query result is returned to the network device that issued the query request according to the device identification carried in the query request.

[0017] As a possible implementation, the query result corresponding to the query request according to the query request and the prompt information is obtained, including: based on the information query model, obtaining the query result corresponding to the query request according to the query request and the prompt information. Based on this, the information query model includes a knowledge question and answer model, which is used to obtain the query result corresponding to the query request according to the input query content and prompt information, wherein the prompt information is used to supplement and limit the query content, so that the query content is easier to be understood by the information query model, and the query result is more accurate.

[0018] As a possible implementation manner, the method further includes: in response to a knowledge update request, updating the knowledge base according to the knowledge information carried in the knowledge update request. Based on this, the information in the knowledge base in the above embodiments is derived from the received knowledge update request, and the knowledge update request may come from a network device. Specifically, it may include a terminal device, a network element device, etc., which is not limited herein. The network device stores the relevant knowledge information of its own environment, the knowledge information obtained by interacting with other network devices, etc. in the knowledge base by sending a knowledge update request. Specifically, by obtaining the knowledge information carried in the knowledge update request, the knowledge base is updated, so that the relevant knowledge information of the network device's own environment and the knowledge information of interacting with other network devices can be stored in the knowledge base.

[0019] As a possible implementation manner, the above knowledge information includes update content, and the updating the knowledge base according to the knowledge information carried in the knowledge update request includes: updating the knowledge base according to the update content. Based on this, the knowledge information carried in the knowledge update request may include update content. Specifically, the update content includes the environment of the network device or the knowledge information generated with other network devices. When the above update content is included in the knowledge update request, the update content is saved to the knowledge base to update the knowledge information in the knowledge base.

[0020] As a possible implementation manner, the above knowledge information includes an update time, and the updating the knowledge base according to the knowledge information carried in the knowledge update request includes: updating the knowledge base according to the update time. Based on this, in addition to the update content, the knowledge update request may further include an update time. In such a case, the update content and the update time can be stored in the knowledge base together, where there is a corresponding relationship between the update content and the update time, and the update time is used to indicate the time when the update content is stored in the knowledge base.

[0021] As a possible implementation manner, the above knowledge information includes an update type, and the updating the knowledge base according to the knowledge information carried in the knowledge update request includes: updating the knowledge base according to the update type. Based on this, in addition to the update content, the knowledge update request may further include an update type. In such a case, the update content and the update type can be stored in the knowledge base together, where there is a corresponding relationship between the update content and the update type, and the update type is used to indicate the corresponding type of the update content. In some examples, the update content corresponding to different update types is stored in different locations for subsequent searching and classification of knowledge information.

[0022] As a possible implementation method, the above-mentioned knowledge update request includes a task identifier, and the knowledge base includes multiple sub-knowledge bases, and the multiple sub-knowledge bases correspond to different task identifiers. The updating of the knowledge base according to the knowledge information carried in the knowledge update request includes: obtaining a sub-knowledge base that matches the knowledge update request from the multiple sub-knowledge bases according to the task identifier; and updating the sub-knowledge base according to the knowledge information carried in the knowledge update request. Based on this, the knowledge base may include multiple sub-knowledge bases, and the multiple sub-knowledge bases may belong to the same knowledge base and be centrally set, or they may be distributed in different locations according to task requirements, and the multiple sub-knowledge bases correspond to different task identifiers. The sub-knowledge base corresponding to the task identifier in the knowledge update request can be determined, and the knowledge information carried in the knowledge update request can be stored in the sub-knowledge base, wherein the knowledge information carried in the knowledge update request may include update content, update time, update type, etc., and may also carry other information related to the update request as needed, without any limitation here.

[0023] As a possible implementation, the method further includes: in response to a model optimization request, optimizing one or more model parameters in the information query model related to the model optimization request according to the knowledge information carried in the model optimization request. Based on this, for a model optimization request, one or more model parameters in the information query model can be optimized according to the knowledge information in the model optimization request. The information query model can be a knowledge question and answer model. In some examples, when optimizing the information query model according to the knowledge information carried in the model optimization request, the model parameters in the model related to the model optimization request can be optimized, while other parameters remain unchanged, thereby increasing the model training efficiency and improving the stability of the model.

[0024] As a possible implementation method, the above-mentioned optimization of one or more model parameters related to the model optimization request in the information query model according to the knowledge information carried in the model optimization request includes: matching and obtaining second knowledge information from the knowledge base according to the knowledge information carried in the model optimization request; optimizing one or more model parameters related to the model optimization request in the information query model according to the second knowledge information. Based on this, knowledge information in the knowledge base with a high degree of relevance to the knowledge information can be determined as the second knowledge information according to the knowledge information carried in the model optimization request. Specifically, the standard in the process of determining the second knowledge information can adopt the relevant standard for determining the first knowledge information in the above-mentioned embodiment, or the standard for selecting the second knowledge information can be adjusted according to the specific model training requirements, and no limitation is made here.

[0025] As a possible implementation, the above information query system includes a knowledge Q&A system. The query request includes a question request, and the query result includes the answer result corresponding to the question request. It should be noted that the information query system may include a knowledge Q&A system. The knowledge Q&A system responds to the question request, determines the first knowledge information in the knowledge base according to the knowledge information carried in the question request, determines the prompt information corresponding to the question request based on the first knowledge information, obtains the answer result corresponding to the question request according to the question request and the prompt information, and outputs it.

[0026] In a second aspect, the present application provides an information query method applied to an information query system. The information query system includes a knowledge base. The method includes: responding to a knowledge update request and updating the knowledge base according to the knowledge information carried in the knowledge update request. Based on this, the information in the knowledge base in the above embodiments is derived from the received knowledge update request. The knowledge update request may come from a network device. Specifically, it may include a terminal device, a network element device, etc., which is not limited herein. The network device stores the relevant knowledge information of its own environment, the knowledge information obtained by interacting with other network devices, etc. in the knowledge base by sending a knowledge update request. Specifically, by obtaining the knowledge information carried in the knowledge update request, the knowledge base is updated, so that the relevant knowledge information of the network device's own environment and the knowledge information of interacting with other network devices can be stored in the knowledge base.

[0027] As a possible implementation, the above knowledge information includes update content. The step of updating the knowledge base according to the knowledge information carried in the knowledge update request includes: updating the knowledge base according to the update content. Based on this, the knowledge information carried in the knowledge update request may include update content. Specifically, the update content includes the environment of the network device or the knowledge information generated with other network devices. When the update content is included in the knowledge update request, the update content is saved to the knowledge base to update the knowledge information in the knowledge base.

[0028] As a possible implementation, the above knowledge information includes an update time. The step of updating the knowledge base according to the knowledge information carried in the knowledge update request includes: updating the knowledge base according to the update time. Based on this, in addition to the update content, the knowledge update request may also include an update time. In such a case, the update content and the update time can be stored in the knowledge base together. Among them, there is a corresponding relationship between the update content and the update time, and the update time is used to indicate the time when the update content is stored in the knowledge base.

[0029] As a possible implementation manner, the above knowledge information includes an update type. Updating the knowledge base according to the knowledge information carried in the knowledge update request includes: updating the knowledge base according to the update type. Based on this, in addition to the update content, the knowledge update request may further include an update type. In such a case, the update content and the update type can be stored in the knowledge base together. Among them, there is a corresponding relationship between the update content and the update type, and the update type is used to indicate that the update content is of the corresponding type. In some examples, the update content corresponding to different update types is stored in different locations to facilitate the subsequent search and classification of knowledge information.

[0030] As a possible implementation manner, the above knowledge update request includes a task identifier. The knowledge base includes multiple sub-knowledge bases, and the multiple sub-knowledge bases correspond to different task identifiers. Updating the knowledge base according to the knowledge information carried in the knowledge update request includes: obtaining, according to the task identifier, a sub-knowledge base that matches the knowledge update request from the multiple sub-knowledge bases; updating the sub-knowledge base according to the knowledge information carried in the knowledge update request. Based on this, the knowledge base may include multiple sub-knowledge bases. The multiple sub-knowledge bases can be set in the same knowledge base set or distributed in different locations according to task requirements. Moreover, the multiple sub-knowledge bases correspond to different task identifiers. The sub-knowledge base corresponding to the knowledge update request can be determined according to the task identifier in the knowledge update request, and the knowledge information carried in the knowledge update request is stored in the sub-knowledge base. Among them, the knowledge information carried in the knowledge update request may include update content, update time, update type, etc., and other information related to the update request may also be carried as needed, which is not limited here.

[0031] As a possible implementation manner, the above information query system includes a knowledge Q&A system. It should be noted that the information query system may include a knowledge Q&A system. The knowledge Q&A system can respond to the knowledge update request and store the knowledge information carried in the knowledge update request in the knowledge base. For details, refer to the response of the above information query system to the knowledge update request.

[0032] Thirdly, the present application provides an information query method, which further includes: in response to a model optimization request, optimizing one or more model parameters related to the model optimization request in the information query model according to the knowledge information carried in the model optimization request. Based on this, for a model optimization request, one or more model parameters in the information query model can be optimized according to the knowledge information in the model optimization request. The information query model can be a knowledge Q&A model. In some examples, when optimizing the information query model according to the knowledge information carried in the model optimization request, the model parameters related to the model optimization request in the model can be optimized while other parameters remain unchanged, which can improve the model training efficiency and the stability of the model at the same time.

[0033] As a possible implementation manner, the above-mentioned optimizing one or more model parameters related to the model optimization request in the information query model according to the knowledge information carried in the model optimization request includes: matching second knowledge information from the knowledge base according to the knowledge information carried in the model optimization request; optimizing one or more model parameters related to the model optimization request in the information query model according to the second knowledge information. Based on this, according to the knowledge information carried in the model optimization request, the knowledge information with a high relevance to this knowledge information in the knowledge base can be determined as the second knowledge information. Specifically, the criteria in the process of determining the second knowledge information can adopt the relevant criteria for determining the first knowledge information in the above-mentioned embodiments, or the criteria for selecting the second knowledge information can be adjusted according to specific model training requirements, which are not limited here.

[0034] As a possible implementation manner, the above-mentioned information query system includes a knowledge Q&A system. It should be noted that the information query system can include a knowledge Q&A system. The knowledge Q&A system can, in response to a model optimization request, optimize the knowledge Q&A model in the knowledge question system according to the knowledge information carried in the model optimization request. In some examples, the knowledge information carried in the model optimization request can be screened with the knowledge information in the knowledge base to obtain the second knowledge information, and then the parameters of the model can be updated according to the second knowledge information. For the specific response of the above-mentioned information query system to the model optimization request, please refer to the above.

[0035] Fourthly, the present application provides a network device, which includes: a transceiver for sending and receiving signals; a memory for storing computer program instructions; and a processor for executing the computer program instructions to support the network device in implementing the method in any possible implementation manner of the first aspect, the second aspect, and the third aspect.

[0036] Fifth aspect, the present application provides an information query system, including one or more network devices, where the one or more network devices are used to support the information query system to implement the method in any possible implementation manner of the first aspect, the second aspect, or the third aspect.

[0037] Sixth aspect, the present application provides a computer-readable storage medium, on which computer program instructions are stored, and when the computer program instructions are executed by a processing circuit, the method in any possible implementation manner of the first aspect, the second aspect, or the third aspect is implemented.

[0038] Seventh aspect, the present application provides a computer program product containing instructions, and when the computer program product runs on a computer, the computer is caused to execute the method in any possible implementation manner of the first aspect, the second aspect, or the third aspect.

[0039] Eighth aspect, the present application provides a chip system, the chip system includes a processing circuit and a storage medium, and computer program instructions are stored in the storage medium; when the computer program instructions are executed by the processing circuit, the method in any possible implementation manner of the first aspect, the second aspect, or the third aspect is implemented. Description of the Drawings

[0040] Figure 1 A schematic diagram of the information query effect provided for the related art;

[0041] Figure 2 A schematic diagram of the model training principle provided for the related art;

[0042] Figure 3 A schematic diagram of the network system architecture provided for the related art;

[0043] Figure 4 A schematic diagram of the application system architecture of an information query method provided for an embodiment of the present application;

[0044] Figure 5 A schematic diagram of the hardware result of a network device provided for an embodiment of the present application;

[0045] Figure 6 A schematic diagram of the process of an information query method provided for an embodiment of the present application;

[0046] Figure 7 A schematic diagram of the process of another information query method provided for an embodiment of the present application;

[0047] Figure 8 A schematic diagram of the process of yet another information query method provided for an embodiment of the present application;

[0048] Figure 9 Schematic diagram of the query effect of an information query method provided by an embodiment of the present application;

[0049] Figure 10 Schematic diagram of the information writing process of an information query method provided by an embodiment of the present application;

[0050] Figure 11 Schematic diagram of the parameter optimization process of an information query method provided by an embodiment of the present application;

[0051] Figure 12 Schematic diagram of the parameter optimization process of another information query method provided by an embodiment of the present application. Detailed implementation manners

[0052] Next, the technical solutions in the embodiments of the present application will be described in conjunction with the accompanying drawings in the embodiments of the present application. Among them, in the description of the embodiments of the present application, unless otherwise specified, " / " means "or". For example, A / B may mean A or B; herein, "and / or" is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B may mean: A exists alone, A and B exist simultaneously, and B exists alone. In addition, in the description of the embodiments of the present application, "a plurality of" means two or more than two.

[0053] Hereinafter, terms such as "first" and "second" are only used to distinguish different described objects, and do not limit the position, order, priority, quantity or content of the described objects. For example, if the described object is "field", the ordinal numbers before "field" in "the first field" and "the second field" do not limit the position or order between "fields", and "first" and "second" do not limit whether the "fields" they modify are in the same message, nor do they limit the order of "the first field" and "the second field". For another example, if the described object is "level", the ordinal numbers before "level" in "the first level" and "the second level" do not limit the priority between "levels". For another example, the quantity of the described object is not limited by the ordinal number and can be one or more. Taking "the first device" as an example, the quantity of "device" therein can be one or more. In addition, the objects modified by different prefix words can be the same or different. For example, if the described object is "device", "the first device" and "the second device" can be the same type of device or different types of devices. For another example, if the described object is "information", "the first information" and "the second information" can be information with the same content or different content. In short, the use of ordinal numbers and other prefix words for distinguishing described objects in the embodiments of the present application does not constitute a limitation on the described objects, and the description of the described objects refers to the description in the claims or the context of the embodiments, and should not constitute unnecessary limitations because of the use of such prefix words.

[0054] In addition, in the embodiments of the present application, "connection" can be a direct connection or an indirect connection; furthermore, it can refer to an electrical connection or a communication connection; for example, when two electrical components A and B are connected, it can mean that A is directly connected to B, or it can mean that A and B are indirectly connected through other electrical components or connection media, or it can mean that A and B are indirectly connected through other communication devices or communication media, as long as communication can be carried out between A and B.

[0055] Currently, with the continuous development of network large models, one of the visions of 6G communication is "intelligent inclusiveness", and typical 6G scenarios include "the integration of artificial intelligence and communication". With the development of network large models, it has triggered an era change in network large models, and network large models are expected to become the underlying technology for intelligent inclusiveness. It is expected that in the future, there will be network large models dedicated to wireless networks deployed within the network.

[0056] Due to the huge number of parameters and high training costs of network large models, they usually provide services externally in a pre-trained manner. Since network large models are pre-trained, they do not have knowledge information about the task execution environment (such as network elements, terminals, etc.) when performing tasks, such as relevant context information for task execution and historical memory information for task execution. Therefore, it is very easy for a complex task to be manually split into multiple executions. Splitting into multiple executions will result in inaccurate execution results and waste resources. As a solution, network large models can answer questions based on a combination of user questions and an externally attached (vector) knowledge base. For specific effects, see Figure 1 , which shows a schematic diagram of an information query effect provided by the related art, as Figure 1 shown. Before querying the question, the question is not processed. The question "What is NAMO in the 6G network architecture?" is directly input into the network large model, and the network large model answers the input question, obtains the answer corresponding to the question and outputs it. The explanation of the keyword NAMO obtained has poor relevance to the 6G network architecture scenario in the original question. Therefore, the answer to the question based on this keyword explanation is also inaccurate and cannot meet the needs.

[0057] In the related art, during the training process of network large models, the process of Pre-Training-->STF (supervised fine-tuning)-->RLHF (reinforcement learning from human feedback) is usually followed. Constructing a Memory experience pool during the reinforcement learning process is a typical practice. See Figure 2 , which shows a schematic diagram of a model training principle provided by the related art, asFigure 2 As shown, during the model training process, the following aspects can be included:

[0058] Random exploration: The neural network model interacts with the simulator. For example, simulator A sends the current state s to the neural network, and the neural network model randomly selects an action a through the action selection module and outputs the action a to simulator B. Simulator B obtains many sets of environmental states and environmental feedback, etc., and stores the corresponding state transition probability tuples (state, action, reward, next state) in the Memory.

[0059] Experience replay: The model performs a replay operation on the Memory with a probability of (1 - ε), samples and updates the parameters of the neural network model based on the historical information stored in the Memory; at the same time, it continues to perform random exploration with a probability of ε. The value of ε is decreased for each training epoch.

[0060] Model convergence: When ε is very small or no longer changes, the model parameters tend to converge, and the training of the neural network model is completed.

[0061] On the other hand, referring to Figure 3 , which shows a schematic diagram of a network system architecture provided by the related art. As Figure 3 shown, in the sixth-generation mobile communication technology, the evolution direction of 6G network elements can be that each network element is one role (agent). One role can include communication between near-field wireless communications (NF) and NF, communication between terminal devices, and communication between network devices. When training the network large model through service data, the role itself also needs to be trained using the reinforcement learning algorithm. During this process, the interaction between the role and the environment and the interaction between roles will generate rich knowledge information. Exemplarily, the interaction between the role and the environment can include the role obtaining environmental parameters, and the interaction between roles can include the role sending and receiving data packets. The knowledge information can include task execution context, task execution historical memory, task execution industry background, etc. These are the knowledge information of the agent. These knowledge information is essential for the model parameter training of the role, but cannot be recorded and stored to assist the parameter training of the network large model, resulting in inaccurate answers from the network large model. And according to the current storage of the vector database, it is stored in the cloud, which cannot ensure that the network element device agent can be in a state of privacy protection. If the user asks the question: "Check the current connection status of the network element", since the network large model cannot retrieve the surrounding environment information, configuration command information, etc. of the network element, the execution result of the network large model will be very inaccurate.

[0062] Based on this, the present application provides an information query method. When there is a query request, it can first obtain knowledge information related to the query request from the knowledge base, then generate prompt information related to the query request according to the knowledge information to supplement the query request, and finally output the query result corresponding to the query request according to the query request and the prompt information. Based on this, it is possible to supplement and optimize the query request with task-related knowledge information, improving the accuracy of the query result when querying information according to the query request. Among them, the knowledge information of the terminal device or network element device is stored in the core network or wireless network. Based on the knowledge base, tasks such as model parameter training experience replay and knowledge Q&A can be completed to improve the accuracy of the model's answer.

[0063] The following will specifically introduce the information query method provided by the embodiments of the present application in conjunction with the accompanying drawings.

[0064] In some examples, referring to Figure 4 , which shows a schematic diagram of the application system architecture of an information query method provided by an embodiment of the present application. As shown in Figure 4 , a task knowledge function (TKF), hereinafter referred to as the knowledge base, can be provided between the network large model and the terminal device or network element device. Among them, there can be one or more TKFs. TKF is a logical function of the network data analytics function (NWDAF), which can perform data storage and logical judgment. In some examples, TKF can be used as a tool plug-in for the network large model, and the parameters corresponding to TKF can be used as parameter plug-ins for the network large model.

[0065] In the embodiments of the present application, TKF can support knowledge information update and knowledge information query operations. For example, in response to a query request, knowledge information matching the query request is obtained in TKF.

[0066] As an example, TKF can open an interface to the terminal device / network element device. The interface can enable information interaction with the terminal device / network element device, and can receive request messages sent by the terminal device / network element device, such as query requests, knowledge update requests, model optimization requests, etc.

[0067] As an example, TKF and the programmable logic controller PLF cooperate with each other to convert knowledge information into prompt information based on a preset function according to the request message.

[0068] As an example, the network large model application programming interface APIs cooperate with each other to complete the information interaction task with the terminal device / network element device.

[0069] As an example, TKF can be deployed on the core network or the radio access network (RAN).

[0070] It should be noted that the embodiments of the present application can be applied to a communication system. Exemplarily, it can include a 5G communication system, a 6G communication system, or a future evolved communication system, or other communication systems, etc., and there is no limitation in this regard. In addition, the embodiments of the present application can also be applicable to an IT (Information Technology System). For example, it can be applied to a network architecture including one or more of the following:

[0071] Terminal device: The user equipment, i.e., the mobile station, can be vehicle-mounted, portable, or handheld, etc. The physical device and the mobile user can be completely independent. All information related to the user can be stored in the smart card (SIM) card, which can be used on the mobile station. The terminal can complete the interaction of the air interface directly with the base station. The terminal can send signals and / or receive signals.

[0072] For example, the terminal device can be used to send a query request, receive a query result corresponding to the query request, and can also be used to initiate a knowledge update request and initiate a model optimization request.

[0073] Base station (xNB): The radio base station in the network, which is also a kind of network element device of the radio access network, responsible for all functions related to the air interface:

[0074] (1) Wireless link maintenance function, maintaining the wireless link with the terminal, and at the same time responsible for the protocol conversion between wireless link data and IP data quality supervision;

[0075] (2) Wireless resource management function, including the establishment and release of wireless links, the scheduling and allocation of wireless resources, etc.;

[0076] (3) Part of the mobility management function, including configuring the terminal to perform measurements, evaluating the quality of the terminal's wireless link, and making decisions on the handover of the terminal between cells, etc.

[0077] For example, the base station can include a knowledge base for storing and processing knowledge information, or a network large model for information query, and can also be used as a relay node in the communication process to assist other devices in information transmission.

[0078] Core network device (CN): The function of the core network device is mainly to provide user connection, user management, and bearer for services, and provide an interface to the external network as a bearer network.

[0079] For example, the core network device may include a knowledge base for storing and processing knowledge information, or a network large model for information query.

[0080] General-purpose computer: It can be used as a terminal device in the network to send query requests, receive query results corresponding to the query requests, and can also be used to initiate knowledge update requests and model optimization requests.

[0081] In some examples, refer to Figure 5 , which shows a schematic diagram of the hardware structure of a network node (for example, the terminal device or network element device in the foregoing embodiments) provided by an embodiment of the present application. As Figure 5 shown, the network device may include a processor 501, a communication line 502, a memory 503, and at least one communication interface ( Figure 5 only an example is given here to include the communication interface 504 for illustration).

[0082] The processor 501 may be a general-purpose central processing unit (CPU), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits for controlling the execution of the program of the solution of the present application.

[0083] The communication line 502 may include a path for transmitting information between the above components.

[0084] The communication interface 504 uses any device such as a transceiver to communicate with other devices or communication networks, such as Ethernet, RAN, WLAN, etc.

[0085] In the embodiment of the present application, the communication line 502 and the communication interface 504 can be used to support the creation / modification / deletion of a secure protocol tunnel between the network device and other network devices (such as the first network device and the second network device), the negotiation of the association relationship between the secure protocol tunnel and the service instance (such as adding / deleting / updating keys, etc.), the transmission of service data, etc.

[0086] The memory 503 can be a read-only memory (ROM) or other types of static storage devices that can store static information and instructions, a random access memory (RAM) or other types of dynamic storage devices that can store information and instructions, or can also be an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM), or other optical disc storage, optical disc storage (including compact discs, laser discs, optical discs, digital versatile discs, Blu-ray discs, etc.), magnetic disk storage media, or other magnetic storage devices, or any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto. The memory can exist independently and be connected to the processor through the communication line 502. The memory can also be integrated with the processor.

[0087] Among them, the memory 503 is used to store computer execution instructions for implementing the solution of this application. Among them, the memory 503 can store instructions for implementing two modular functions: sending instructions, receiving instructions, and processing instructions, and is controlled by the processor 501 to execute. The processor 501 is used to execute the computer execution instructions stored in the memory 503, so as to implement the method provided in the following embodiments of this application. Figure 5 The memory 503 shown in is only a schematic diagram, and this memory can also include other functionalized instructions. In this regard, the present invention does not make any limitations thereto.

[0088] Optionally, the computer execution instructions in this application can also be referred to as application code, and this application does not make specific limitations thereto.

[0089] In a specific implementation, as an embodiment, the processor 501 can include one or more CPUs, such as Figure 5 the CPU0 and CPU1 in.

[0090] It should be noted that Figure 5 only as an example of a network device, does not limit the specific structure of the network device. For example, the network device can also include other functional modules.

[0091] In some examples, refer to Figure 6 , which shows a schematic flowchart of an information query method provided by an embodiment of this application. This method can be applied to an information query system, and the information query system includes a knowledge base, and the knowledge base includes knowledge information related to one or more query tasks, such as Figure 6As shown, the method may include:

[0092] S601: In response to a query request, obtain first knowledge information that matches the query request from the knowledge base.

[0093] It should be noted that when a query request is received, according to the knowledge information carried in the query request, first knowledge information related to the query request is obtained from the knowledge base. The first knowledge information is one or more knowledge information in the knowledge base. The first knowledge information may come from a terminal device or a network element device, and may also include other network devices that have information interaction with this information query system. Specifically, the first information may include the knowledge information generated by the interaction between the terminal device or the network element device and the environment, as well as the knowledge information generated by the interaction between each device. These knowledge information can be matched with the knowledge information carried in the query request during the information query process to obtain the first knowledge information, and the first knowledge information can be used to supplement and perfect the query request.

[0094] In some embodiments, obtaining the first knowledge information that matches the query request from the above knowledge base may include: obtaining the first knowledge information that matches the query request from the knowledge base according to a first preset condition, and the first preset condition is related to one or more of the following: knowledge relevance, time relevance, type relevance. It should be noted that during the process of querying the first knowledge information, the relevance level between the knowledge information carried in the query request and the knowledge information in the knowledge base can be judged through different dimensions, and the knowledge information with a high relevance to the knowledge information carried in the query request is used as the first knowledge information. Among them, when judging the relevance between the knowledge information carried in the query request and the knowledge information in the knowledge base, it can be judged by combining one or more of knowledge relevance, time relevance, and type relevance, or other dimensions that can reflect the relevance between the two can be adopted according to the specific task, and no limitation is made here.

[0095] In some embodiments, the query request carries query content, and obtaining the first knowledge information that matches the query request from the knowledge base according to the first preset condition may include: using the knowledge information whose updated content has a knowledge relevance greater than or equal to a preset threshold with the query content as the first knowledge information. Based on this, the first knowledge information can be determined according to the knowledge relevance between the knowledge information in the query request and the knowledge information in the knowledge base. Specifically, the knowledge information with a knowledge relevance greater than or equal to the preset threshold can be used as the first knowledge information. Exemplarily, the knowledge relevance may indicate the semantic similarity between the updated content in the knowledge information in the knowledge base and the query content in the knowledge information in the query request.

[0096] In some embodiments, the query request carries a query time. Obtaining first knowledge information that matches the query request from the knowledge base according to the first preset condition includes: using knowledge information whose time correlation between the update time of the knowledge information and the query time is greater than or equal to a preset threshold as the first knowledge information. Based on this, the first knowledge information can be determined according to the time correlation between the knowledge information in the query request and the knowledge information in the knowledge base. Specifically, knowledge information with a time correlation greater than or equal to the preset threshold can be used as the first knowledge information. Exemplarily, the time correlation can indicate the degree of temporal proximity between the update time in the knowledge information in the knowledge base and the query time in the knowledge information in the query request.

[0097] In some embodiments, the query request carries a query type. Obtaining first knowledge information that matches the query request from the knowledge base according to the first preset condition includes: using knowledge information whose type correlation between the update type of the knowledge information and the query type is greater than or equal to a preset threshold as the first knowledge information. Based on this, the first knowledge information can be determined according to the knowledge correlation between the knowledge information in the query request and the knowledge information in the knowledge base. Specifically, knowledge information with a knowledge correlation greater than or equal to the preset threshold can be used as the first knowledge information. Exemplarily, the knowledge correlation can indicate the degree of type similarity between the update type in the knowledge information in the knowledge base and the query type in the knowledge information in the query request. For example, the type similarity between the two can be determined by judging the cosine correlation between the update type corresponding to the knowledge information in the knowledge base and the query type corresponding to the query request.

[0098] In some embodiments, the query request includes a task identifier, and the knowledge base includes multiple sub-knowledge bases corresponding to different task identifiers. Obtaining first knowledge information that matches the query request from the knowledge base includes: according to the task identifier, obtaining a sub-knowledge base that matches the query request from the multiple sub-knowledge bases; obtaining the first knowledge information that matches the query request from the sub-knowledge base. It should be noted that the knowledge base may include multiple sub-knowledge bases. The multiple sub-knowledge bases can be set in the same knowledge base set or distributed at different locations according to task requirements, and the multiple sub-knowledge bases correspond to different task identifiers. The sub-knowledge base corresponding to the query request can be determined according to the task identifier in the query request, and the first knowledge information can be obtained from the sub-knowledge base corresponding to the task identifier carried in the query request. The correlation between the first knowledge information and the knowledge information carried in the query request meets the first preset condition. Specifically, the standard for obtaining the first knowledge information is the same as the method for obtaining the first knowledge information from the knowledge base in the foregoing embodiments, and will not be elaborated here.

[0099] Exemplarily, refer toFigure 7 , which shows a schematic flowchart of another information query method provided by an embodiment of the present application. As Figure 7 shown, when the information query method is specifically a knowledge Q&A method, the method may include: a terminal device or a network element device sends a task request based on task network attached storage (T-NAS). The TCF obtains a sub-knowledge base (equivalent to the knowledge base TKF in the foregoing embodiment) that matches the task request from the multiple sub-knowledge bases according to the task identifier carried in the task request; Exemplarily, based on a task control function (TCF), the first knowledge information that matches the task request is obtained from the sub-knowledge base according to the task identifier carried in the task request of the T-NAS. In some examples, a task session is further established based on the matching task request and sub-knowledge base, and a corresponding task identifier (task ID) is determined. The knowledge base may include multiple sub-knowledge bases. The multiple sub-knowledge bases may be set in the same knowledge base set, or may be distributed at different locations according to task requirements, and the multiple sub-knowledge bases correspond to different task identifiers. The sub-knowledge base that matches the task identifier in the task request of the T-NAS can be determined, and the first knowledge information is obtained from the sub-knowledge base corresponding to the task identifier carried in the query request. The standard for obtaining the first knowledge information is the same as that in the foregoing embodiment and will not be elaborated here.

[0100] S602: Generate prompt information related to the query request according to the first knowledge information;

[0101] It should be noted that, according to the first knowledge information obtained in the foregoing steps, the prompt information related to the first knowledge information is determined. The prompt information may include one or more. The prompt information is used to supplement and perfect the query content carried in the query request to make the query content carried in the query request clearer.

[0102] In some embodiments, generating the hint information related to the query request according to the first knowledge information described above includes: generating the hint information related to the query request based on a hint logic algorithm according to the first knowledge information, where the hint information is used to supplement the query content carried by the query request. It should be noted that one or more pieces of first knowledge information obtained from the knowledge base are highly relevant to the knowledge information carried by the query request. Exemplarily, the relevant content of the first knowledge information can be converted into hint information through a hint logic algorithm. Among them, the hint logic algorithm includes a hint logic function, and the hint logic function is used to convert the first knowledge information into a simpler hint information to supplement or limit the query content carried in the query request, making the query content clearer and more complete, facilitating the understanding and query of the query content. For a query request, there may be one or more hint information, which can improve the query content from different levels, and no limitation is made here.

[0103] In some embodiments, generating the hint information related to the query request based on the hint logic algorithm according to the first knowledge information described above includes: obtaining multiple candidate hint information corresponding to the first knowledge information based on the hint logic algorithm; determining the hint information from the one or more candidate hint information, where the hint information is the candidate hint information whose similarity to the query content among the multiple candidate hint information is greater than or equal to a preset threshold. Based on this, the first knowledge information can be converted into one or more candidate hint information according to the logic hint algorithm. Exemplarily, when there is one candidate hint information, the candidate hint information can be directly used as the hint information. When there are multiple candidate hint information, it is necessary to screen among the multiple candidate hint information to determine the candidate hint information with a high matching degree with the query request as the hint information. In some examples, the hint information in the candidate hint information can be determined by the similarity between the candidate hint information and the query content in the query request. The hint information can be one or more, and different screening criteria for the hint information can be set according to specific requirements, and no limitation is made here.

[0104] In some examples, the information query system may include a knowledge Q&A system. The TKF assists the large network model to improve the accuracy of knowledge Q&A. In this process, the TKF needs to interact with the prompt logic function (PLF) to obtain the candidate hint information for knowledge Q&A. Among them, there may be one or more candidate hint information, and the hint information that can improve the query request can be determined from the one or more hint information.

[0105] In some embodiments, refer to Figure 7, during the process of selecting prompt information, the PLF can directly select appropriate prompt information from one or more candidate prompt information through a prompt logic function. Specifically, it can compare the candidate prompt information with the content information in the query request, select the candidate prompt information with higher similarity or more overlapping fields as the final prompt information, and send it to the knowledge base for supplementing and prompting the questions in the query request. In some examples, the method may include the following S1 - S5:

[0106] S1: The terminal device / network element device matches the corresponding TKF through the TCF and establishes a task session with the TKF.

[0107] S2: The terminal device / network element device performs a "knowledge Q&A" task. The terminal / network element device sends a question or query request to the TKF. The query request may include one or more of the following:

[0108] AgentID (i.e., the device identifier in the foregoing embodiment): The key of the Agent;

[0109] TaskID (i.e., the task identifier in the foregoing embodiment): The ID of the task;

[0110] TaskType (i.e., the query type in the foregoing embodiment): Knowledge Q&A;

[0111] TimeStamp (i.e., the query time in the foregoing embodiment): The timestamp when the task is triggered;

[0112] Content (i.e., the query content in the foregoing embodiment): The specific question.

[0113] S3: The TKF retrieves knowledge related to the Content from the knowledge base. When retrieving, the scores of Recency and Relevance need to be considered.

[0114] It should be noted that for the Recency score: The closer the knowledge information is to the present, the higher the score. The decay rate of the knowledge information over time can be set as a floating point number between *0, 1+, for example, 0.95. Exemplarily, the Recency score can be 0.95 ^ (TimeStamp in the query request - TimeStamp of the last update of the knowledge information in the knowledge base in the update request).

[0115] Relevance score: The higher the knowledge relevance, the higher the Relevance score. Specifically, the Relevance score can be expressed as semantic similarity. In some examples, to determine the semantic similarity, it can be searched precisely according to TaskType, or the cosine correlation between TaskType and the TaskType of the knowledge in the database can be calculated.

[0116] In some examples, the scoring algorithm used in knowledge information retrieval combines the Relevance score and the Recency score, specifically: semantic similarity + decay rate ** number of hours elapsed. (where ** represents squaring)

[0117] In some examples, the semantic similarity is represented by the Relevance field, and the time relevance can be represented by the Recency field.

[0118] In the calculation of the Recency score, the decay rate of knowledge information over time can be set to a floating point number between *0,1+;

[0119] Specifically, when the decay rate = 0, it means that the Recency score is only matched according to semantic similarity and has nothing to do with time; it can be used to query objective facts that have nothing to do with time. For example, query what particles are composed of?

[0120] When the decay rate = 1, it means that the Recency score is maximized and related to time. For example, query what are all the latest news?

[0121] When the decay rate is a floating point number between 0 and 1, it means that the knowledge in the knowledge base will gradually become obsolete over time. For example, query what is the profit of Company A in the past year?

[0122] S4: TKF queries the candidate hint information corresponding to the knowledge information obtained through knowledge retrieval through PLF, queries the Prompt from the candidate hint information, and returns the hint information Prompt to TKF. Exemplarily, the Prompt returned by PLF can be:

[0123] Your task is to do reading comprehension from the following content and answer the questions.

[0124] Content: ***Question:***

[0125] S5: TKF sends the "question request (i.e., the query request in the foregoing embodiment) + Prompt" to the network large model API. After the network large model generates the Q&A result, it returns the Q&A result corresponding to the question request to the terminal device / network element device.

[0126] In some examples, for the query operation of knowledge information on TKF, the control signaling for knowledge information query can include the fields: "TaskID + TaskType + Action", where both TaskID and TaskType have been implemented in the relevant technologies of TCF.

[0127] In some embodiments, see Figure 8, which shows a schematic flowchart of another information query method provided by an embodiment of the present application. As Figure 8 shown, during the process of selecting prompt information, the PLF can directly send one or more candidate prompt information to the knowledge base, and the knowledge base selects appropriate prompt information from the one or more candidate prompt information. Specifically, by comparing the candidate prompt information with the content information in the query request, the candidate prompt information with higher similarity or more overlapping fields can be selected as the final prompt information and sent to the knowledge base for supplementing and prompting the problem in the query request. Exemplarily, the method may include S1-S5, where steps S1-S3 are the same as the foregoing steps S1-S3 and will not be elaborated herein. S4 may include the following S41-S43:

[0128] S41: The TKF queries the candidate prompt information from the PLF (Prompt Logic Function);

[0129] S42: After the PLF queries the prompt information, it returns multiple candidate prompt information related to the question request to the TKF;

[0130] S43: The TKF performs a similarity match between the received multiple candidate prompt information and the Content field, and selects the most suitable Prompt.

[0131] S5 is as follows:

[0132] S5: The TKF sends the "question + Prompt" to the network large model API. After the network large model generates an answer, it returns the answer to the terminal / network element device. In some scenarios, the Prompt in this embodiment can also be expressed as prefix information (Prefix).

[0133] S603: Generate a question-and-answer result corresponding to the question request according to the question request and the prompt information, and output the query result corresponding to the question.

[0134] It should be noted that the network large model can combine the query request and the prompt information related to the query request to obtain the query result corresponding to the query request and output it.

[0135] In some embodiments, the query request sent by the terminal device or network element device carries the device identifier of the network device that initiates the query request. Outputting the query result corresponding to the query request according to the query request and the prompt information includes: obtaining the query result corresponding to the query request according to the query request and the prompt information; and outputting the query result to the network device according to the device identifier. It should be noted that the query request carries the device identifier of the network device that sends the query request. Herein, the network device may include terminal devices, network element devices, and other devices capable of executing query requests, which are not limited herein. Based on the query content in the knowledge information carried in the query request and the supplementation and limitation of the query content by the prompt information, the query result corresponding to the query request is obtained, and the query result is returned to the network device that sends the query request according to the device identifier carried in the query request.

[0136] In some embodiments, obtaining the query result corresponding to the query request according to the query request and the prompt information includes: based on an information query model, obtaining the query result corresponding to the query request according to the query request and the prompt information. Based on this, the information query model includes a knowledge Q&A model, which is used to obtain the query result corresponding to the query request according to the input query content and prompt information. The prompt information is used to supplement and limit the query content in the query request, making the query content easier to be understood by the information query model and making the query result more accurate.

[0137] Based on this, referring to Figure 9 , which shows a schematic diagram of the query effect of an information query method provided by an embodiment of the present application. As shown in Figure 9 , for the same problem request: What is NAMO in the 6G network architecture?, the prompt information "task-centric 6G network AI architecture" provided by the knowledge base supplements and prompts the original problem request, making the application scenario of NAMO by the network large model clearer. Therefore, the Q&A result corresponding to the problem request obtained based on the accurate usage scenario is very detailed and accurate. Compared with the query effect of the related technology shown in Figure 1 and Figure 1 , its accuracy fully reflects the technical effect of the embodiment of the present application.

[0138] In some examples, the network large model may include a wireless operation and maintenance expert knowledge Q&A model, a network element operation log query model, etc., as well as other network large models for information query. The method provided by the embodiment of the present application is also applicable to the above network large models, which are not limited herein.

[0139] In some embodiments, TKF can assist in the optimization of the network large model itself. On this basis, the method may further include: in response to a knowledge update request, updating the knowledge base according to the knowledge information carried in the knowledge update request. Specifically, the usage scenarios in the process of assisting the optimization of the network large model may include: the models corresponding to network elements / terminal devices may be network large models, and the knowledge base is updated according to the knowledge information carried in the knowledge update request to complete the optimization of the network large model.

[0140] It should be noted that the information in the knowledge base in the above embodiments is derived from the received knowledge update requests. The knowledge update requests may come from network devices, specifically, may include terminal devices and network elements, etc., which are not limited herein. The network devices store the relevant knowledge information about their own environments and the knowledge information obtained from interactions with other network devices, etc. in the knowledge base by sending knowledge update requests. Exemplarily, the knowledge base can be updated by obtaining the knowledge information carried in the knowledge update request, so that the relevant knowledge information about the network devices' own environments and the knowledge information about interactions with other network devices can be stored in the knowledge base.

[0141] In some embodiments, the above knowledge information includes update content. The updating of the knowledge base according to the knowledge information carried in the knowledge update request includes: updating the knowledge base according to the update content. It should be noted that the knowledge information carried in the knowledge update request may include update content, and the update content may include the environment of the network device or the knowledge information generated with other network devices. When the above update content is included in the knowledge update request, the update content is saved to the knowledge base to update the knowledge information in the knowledge base.

[0142] In some embodiments, the above knowledge information includes update time. The updating of the knowledge base according to the knowledge information carried in the knowledge update request includes: updating the knowledge base according to the update time. It should be noted that in addition to the update content, the knowledge update request may also include the update time. In such a case, the update content and the update time can be stored in the knowledge base together. Among them, there is a corresponding relationship between the update content and the update time, and the update time is used to indicate the update time when the update content is stored in the knowledge base. In some examples, the update content corresponding to the knowledge information may be updated multiple times, and the update time is used to indicate the time of the last update of the update content corresponding to the knowledge information.

[0143] In some embodiments, the above knowledge information includes an update type. Updating the knowledge base according to the knowledge information carried in the knowledge update request includes: updating the knowledge base according to the update type. It should be noted that in addition to the update content, the knowledge update request may also include an update type. In such a case, the update content and the update type can be stored in the knowledge base together. Among them, there is a corresponding relationship between the update content and the update type, and the update type is used to indicate the type corresponding to the update content. In some examples, the update content corresponding to different update types is stored in different locations to facilitate the subsequent search and classification of knowledge information.

[0144] In some embodiments, the above knowledge update request includes a task identifier. The knowledge base includes multiple sub-knowledge bases, and the multiple sub-knowledge bases correspond to different task identifiers. Updating the knowledge base according to the knowledge information carried in the knowledge update request includes: obtaining, according to the task identifier, a sub-knowledge base that matches the knowledge update request from the multiple sub-knowledge bases; updating the sub-knowledge base according to the knowledge information carried in the knowledge update request. Based on this, the knowledge base may include multiple sub-knowledge bases. The multiple sub-knowledge bases can be set in the same knowledge base set or distributed at different locations according to task requirements, and the multiple sub-knowledge bases correspond to different task identifiers. The sub-knowledge base corresponding to the knowledge update request can be determined according to the task identifier in the knowledge update request, and the knowledge information carried in the knowledge update request can be stored in the sub-knowledge base. Among them, the knowledge information carried in the knowledge update request may include update content, update time, update type, etc., and other information related to the update request can also be carried according to needs, which is not limited here.

[0145] Specifically, in some embodiments, refer to Figure 10 , which shows a schematic diagram of the information writing process of an information query method provided by an embodiment of the present application. As Figure 10 shown, the TKF may include multiple sub-knowledge bases, and the multiple sub-knowledge bases correspond to different task identifiers. Exemplarily, updating the knowledge base according to the knowledge information carried in the knowledge update request may include: a terminal or a network element device sends a task request based on T-NAS (i.e., the knowledge update request in the foregoing embodiment), and the task request carries a task identifier. The TCF obtains, according to the task identifier, a sub-knowledge base that matches the knowledge update request from the multiple sub-knowledge bases; in some examples, the TKF performs a request confirmation on the task request and establishes a task session corresponding to the terminal device or the network element device.

[0146] It should be noted that the knowledge base may include multiple sub-knowledge bases. The multiple sub-knowledge bases can be set centrally or distributed at different locations according to task requirements. Moreover, the multiple sub-knowledge bases correspond to different task identifiers. The sub-knowledge base corresponding to the task identifier in the knowledge update request can be determined, and the knowledge information carried in the knowledge update request can be stored in this sub-knowledge base. Among them, the knowledge information carried in the knowledge update request may include update content, update time, update type, etc., and other information related to the update request can also be carried as needed, which is not limited here.

[0147] In some embodiments, the above method further includes: in response to a model optimization request, optimizing one or more model parameters related to the model optimization request in the information query model according to the knowledge information carried in the model optimization request. For the model optimization request, one or more model parameters in the information query model can be optimized according to the knowledge information in the model optimization request. In some examples, the information query model can be a knowledge Q&A model. When optimizing the information query model according to the knowledge information carried in the model optimization request, the model parameters related to the model optimization request in the model can be optimized while other parameters remain unchanged, which can improve the model training efficiency and the stability of the model.

[0148] In some embodiments, optimizing one or more model parameters related to the model optimization request in the information query model according to the knowledge information carried in the model optimization request includes: matching second knowledge information from the knowledge base according to the knowledge information carried in the model optimization request; optimizing one or more model parameters related to the model optimization request in the information query model according to the second knowledge information. It should be noted that based on the model optimization request, the knowledge information with a high degree of relevance to this knowledge information in the knowledge base can be determined as the second knowledge information according to the knowledge information carried in the model optimization request. Specifically, the criteria in the process of determining the second knowledge information can adopt the relevant criteria for determining the first knowledge information in the above embodiments, or the criteria for selecting the second knowledge information can be adjusted according to specific model training requirements, which is not limited here.

[0149] Specifically, in some embodiments, refer to Figure 11 , which shows a schematic diagram of the parameter optimization process of an information query method provided by an embodiment of the present application. As Figure 11 shown, after the parameter optimization is triggered, the terminal device or the network element device sends a parameter optimization request to the TKF, retrieves the knowledge information related to the task based on the parameter optimization request, such as the second knowledge information, and sends it to the network large model for model training to optimize the model parameters. In some examples, refer to Figure 12, which shows a schematic diagram of the parameter optimization process of another information query method provided by the embodiments of the present application. As Figure 12 shown, during the process of model training, a part of the parameters that do not need to be optimized can be parameter-frozen, and only another part of the external parameters that need to be optimized are parameter-trained to improve the efficiency of model training and ensure that the parts of the model irrelevant to the task remain stable.

[0150] In some examples, when there is one or more sub-knowledge bases in the knowledge base, after receiving a parameter optimization request, the sub-knowledge base matching the parameter optimization request can be determined through TCF and then optimized. The matching method of the parameter optimization request and the sub-knowledge base by TCF can be matched through the task identifier. Refer to the matching method of the knowledge update request and the sub-knowledge base in the above embodiments, which will not be elaborated here.

[0151] In some embodiments, the above information query system includes a knowledge Q&A system, the query request includes a question request, and the query result includes the answer result corresponding to the question request. It should be noted that the information query system may include a knowledge Q&A system. The knowledge Q&A system responds to the question request, determines the first knowledge information in the knowledge base according to the knowledge information carried in the question request, determines the prompt information corresponding to the question request based on the first knowledge information, obtains the Q&A result corresponding to the question request according to the question request and the prompt information, and outputs it.

[0152] In some embodiments, the present application provides an information query method applied to an information query system. The information query system includes a knowledge base. The method includes: responding to a knowledge update request and updating the knowledge base according to the knowledge information carried in the knowledge update request. Based on this, the knowledge information in the knowledge base in the above embodiments comes from the received knowledge update request. The knowledge update request may come from a network device. In some examples, the network device may include a terminal device and a network element device, etc., which is not limited here. The network device stores the relevant knowledge information of its own environment, the knowledge information obtained by interacting with other network devices, etc. in the knowledge base by sending a knowledge update request. Exemplarily, the knowledge base can be updated by obtaining the knowledge information carried in the knowledge update request, so that the relevant knowledge information of the network device's own environment and the knowledge information of interacting with other network devices can be stored in the knowledge base.

[0153] In some embodiments, the above-mentioned knowledge information includes update content. Updating the knowledge base according to the knowledge information carried in the knowledge update request includes: updating the knowledge base according to the update content. The knowledge information carried in the knowledge update request may include update content. The specific update content includes the environment of the network device or the knowledge information generated by interacting with other network devices. When the update content is included in the knowledge update request, the update content is saved to the knowledge base. In some examples, the existing knowledge information in the knowledge base can be updated according to the knowledge update request.

[0154] In some embodiments, the above-mentioned knowledge information includes update time. Updating the knowledge base according to the knowledge information carried in the knowledge update request includes: updating the knowledge base according to the update time. Based on this, in addition to the update content, the knowledge update request may further include the update time. In such a case, the update content and the update time can be stored in the knowledge base together. Among them, there is a corresponding relationship between the update content and the update time. The update time is used to indicate the time when the update content is stored in the knowledge base. In some examples, when updating the existing knowledge information in the knowledge base according to the knowledge update request, the update time corresponding to the knowledge information is synchronously updated.

[0155] In some embodiments, the above-mentioned knowledge information includes update type. Updating the knowledge base according to the knowledge information carried in the knowledge update request includes: updating the knowledge base according to the update type. Based on this, in addition to the update content, the knowledge update request may further include the update type. In such a case, the update content and the update type can be stored in the knowledge base together. Among them, there is a corresponding relationship between the update content and the update type. The update type is used to indicate the corresponding type of the update content. In some examples, the update content corresponding to different update types is stored in different locations for subsequent searching and classification of knowledge information. In some examples, when updating the existing knowledge information in the knowledge base according to the knowledge update request, the update type corresponding to the knowledge information is synchronously updated.

[0156] In some embodiments, the above-mentioned knowledge update request includes a task identifier, the knowledge base includes multiple sub-knowledge bases, and the multiple sub-knowledge bases correspond to different task identifiers. The updating of the knowledge base according to the knowledge information carried in the knowledge update request includes: obtaining a sub-knowledge base that matches the knowledge update request from the multiple sub-knowledge bases according to the task identifier; and updating the sub-knowledge base according to the knowledge information carried in the knowledge update request. It should be noted that a knowledge base may include multiple sub-knowledge bases, and the multiple sub-knowledge bases may belong to the same knowledge base and be centrally set, or they may be distributed in different locations according to task requirements, and the multiple sub-knowledge bases correspond to different task identifiers. The sub-knowledge base corresponding to the task identifier in the knowledge update request can be determined, and the knowledge information carried in the knowledge update request can be stored in the sub-knowledge base, wherein the knowledge information carried in the knowledge update request may include update content, update time, update type, etc., and other information related to the update request may also be carried as needed, without any limitation here.

[0157] In some embodiments, the information query system may include a knowledge question answering system. The knowledge question answering system may respond to a knowledge update request and store the knowledge information carried in the knowledge update request in a knowledge base. For details, see the response of the information query system to the knowledge update request.

[0158] In some examples, TKF can be used to update knowledge information in a knowledge base and optimize a large network model, and can include the following steps S1-S7:

[0159] S1: The terminal device / network element device matches the corresponding TKF through TCF and establishes a task session with the TKF.

[0160] S2: After the terminal device / network element device collects the task-related knowledge information, it initiates a knowledge update request to the TKF. The knowledge update request includes one or more of the following:

[0161] AgentID (i.e., the device ID in the aforementioned embodiment): the key of the Agent;

[0162] TaskID (i.e., the task identifier in the aforementioned embodiment): the ID of the task;

[0163] TaskType (i.e., the update type in the aforementioned embodiment): task type;

[0164] TimeStamp (i.e., update time in the aforementioned embodiment): timestamp of knowledge update;

[0165] Content (i.e., the updated content in the aforementioned embodiment): detailed content of knowledge.

[0166] S3: The TKF persists the knowledge information into the vector database.

[0167] In some examples, when updating the knowledge information of the TKF, the control signaling of the corresponding knowledge update request may include the fields: "AgentID + TaskID + TaskType + TimeStamp + Content", where TaskID and TaskType are already implemented in the relevant technologies of the TCF.

[0168] In some examples, the method further includes:

[0169] S4: The terminal device / network element device triggers the "parameter optimization" task.

[0170] S5: The terminal device / network element device can initiate a parameter optimization request to the TKF. The parameter optimization request includes one or more of the following:

[0171] TaskID: The ID of the task;

[0172] TaskType: The type of the task;

[0173] Action: Task parameter optimization.

[0174] S6: The TKF retrieves the knowledge information related to the parameter optimization request (i.e., the second knowledge information in the foregoing embodiments). When retrieving, one or more of the following criteria can be used:

[0175] Recency score: The closer the knowledge information is to the present, the higher the score. The decay rate of the knowledge information over time can be set to a floating point number between *0, 1+, for example, 0.95. An exemplary Recency score can be 0.95^(TimeStamp in the query request - TimeStamp in the update request for the last update of the knowledge information in the knowledge base).

[0176] Relevance score: The higher the knowledge relevance, the higher the Relevance score. Specifically, the Relevance score can be represented as semantic similarity. In some examples, to determine the semantic similarity, an exact search can be performed according to TaskType, or the cosine correlation between TaskType and the TaskType of the knowledge in the database can be calculated.

[0177] In some examples, the scoring algorithm used in the knowledge information retrieval combines the Relevance score and the Recency score. Specifically: Semantic similarity + Decay rate ** Number of hours elapsed. (where ** represents squaring)

[0178] In some examples, semantic similarity is represented by the Relevance field, and temporal relevance is represented by the Recency field.

[0179] During the calculation of the Recency score, the decay rate of knowledge information over time can be set to a floating point number between *0,1+;

[0180] Specifically, when the decay rate = 0, it means that the Recency score is matched only based on semantic similarity and has nothing to do with time; it can be used to query objective facts that are independent of time, for example, what are the particles composed of?

[0181] When the decay rate = 1, it means that the maximization of the Recency score is related to time. For example, what are all the recent news?

[0182] When the decay rate is a floating point number between 0 and 1, it means that the knowledge in the knowledge base will become obsolete over time. For example, what is the profit of Company A in the most recent year?

[0183] In some examples, knowledge information may be retrieved based on the tasks involved in the parameter optimization request using other criteria for determining the relevance of knowledge information to obtain second knowledge information corresponding to the parameter optimization request, without any limitation herein.

[0184] S7: TKF sends the retrieved knowledge information to the network big model API to trigger the model training of the network big model.

[0185] In some examples, the TKF sends the retrieved knowledge information to the network big model API to trigger the model training of the network big model (ie, S7), which may include the following S71-S72:

[0186] S71: TKF sends the retrieved knowledge information to the API of the plug-in parameters.

[0187] S72: The plug-in parameters call the network big model API for repeated interactions, and continuously optimize the plug-in parameters by "freezing the network big model parameters and only training the plug-in parameters".

[0188] When TKF assists the large network model to improve the accuracy of knowledge question and answer, TKF and PLF newly add interactive signaling. The interactive signaling can include the following fields: "TaskID+TaskType+Task Category+Task Execution Parameters", where TaskID and TaskType have been implemented in TCF related technologies.

[0189] In some embodiments, the present application provides an information query method, which further includes: in response to a model optimization request, optimizing one or more model parameters related to the model optimization request in the information query model according to the knowledge information carried in the model optimization request. It should be noted that for the model optimization request, one or more model parameters in the information query model can be optimized according to the knowledge information in the model optimization request. The information query model can be a knowledge-based question answering model. In some examples, when optimizing the information query model according to the knowledge information carried in the model optimization request, the model parameters related to the model optimization request in the model can be optimized while other parameters remain unchanged, which can improve the training efficiency of the model and the stability of the model at the same time.

[0190] In some embodiments, the above-mentioned optimizing one or more model parameters related to the model optimization request in the information query model according to the knowledge information carried in the model optimization request includes: matching second knowledge information from the knowledge base according to the knowledge information carried in the model optimization request; optimizing one or more model parameters related to the model optimization request in the information query model according to the second knowledge information. In some examples, the knowledge information carried in the model optimization request can be used to determine the knowledge information with a high degree of relevance to this knowledge information in the knowledge base as the second knowledge information. The second knowledge information is used to supplement and optimize the knowledge information carried in the model optimization request. Specifically, the criteria in the process of determining the second knowledge information can adopt the relevant criteria for determining the first knowledge information in the above-mentioned embodiments, or the criteria for selecting the second knowledge information can be adjusted according to the specific model training requirements, which are not limited here.

[0191] In some embodiments, the above-mentioned information query system includes a knowledge-based question answering system. It should be noted that the information query system can include a knowledge-based question answering system. The knowledge-based question answering system can, in response to a model optimization request, optimize the knowledge-based question answering model in the question system according to the knowledge information carried in the model optimization request. In some examples, the knowledge information carried in the model optimization request can be screened with the knowledge information in the knowledge base to obtain the second knowledge information, and then the parameters of the model can be updated according to the second knowledge information. For details, refer to the response of the above-mentioned information query system to the model optimization request.

[0192] After the TKF function is provided in the embodiments of the present application, the knowledge information generated during the execution of the network element device / terminal device can be stored in the TKF; the model parameter training corresponding to the network element device / terminal device can use the knowledge information in the TKF for assistance; in the knowledge-based question answering scenario, the knowledge information in the TKF can improve the accuracy of the question answering results.

[0193] It should be understood that the various solutions of the embodiments of the present application can be reasonably combined and used, and the explanations or descriptions of the various terms appearing in the embodiments can be referred to or explained with each other in the various embodiments, which is not limited herein.

[0194] It should also be understood that in the various embodiments of the present application, the magnitudes of the serial numbers of the above processes do not mean the order of execution, and the execution order of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present application.

[0195] It can be understood that in order to implement the functions of any of the above embodiments, the controller or the resource allocation simulation device includes the corresponding hardware structure and / or software module for executing each function. Those skilled in the art should easily realize that, in combination with the units and algorithm steps of the examples described in the embodiments disclosed herein, the present application can be implemented in the form of hardware or a combination of hardware and computer software. Whether a certain function is executed in the form of hardware or computer software driving hardware depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present application.

[0196] The embodiments of the present application can perform functional module division on a controller or a resource allocation simulation device with controller resource allocation capabilities. For example, each functional module can be divided corresponding to each function, or two or more functions can be integrated into one processing module. The above integrated module can be implemented in the form of hardware or in the form of a software functional module. It should be noted that the division of modules in the embodiments of the present application is illustrative, only a logical function division, and there may be other division methods in actual implementation.

[0197] It should also be understood that the various modules in the controller or the resource allocation simulation device can be implemented in the form of software and / or hardware, which is not specifically limited herein. In other words, the electronic device is presented in the form of functional modules. Here, the "module" can refer to an application specific integrated circuit ASIC, a circuit, a processor and a memory that execute one or more software or firmware programs, an integrated logic circuit, and / or other devices that can provide the above functions.

[0198] In an alternative approach, when data transmission is implemented using software, it can be realized in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the processes or functions described in the embodiments of the present application are realized in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions may be transmitted from a website, computer, server, or data center to another website, computer, server, or data center via wired (such as coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (such as infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that can be accessed by a computer or a data storage device such as a server or data center that includes one or more integrated available media. The available medium may be a magnetic medium (such as a floppy disk, hard disk, magnetic tape), an optical medium (such as a digital video disk (DVD)), or a semiconductor medium (such as a solid state disk (SSD)), etc.

[0199] The steps of the methods or algorithms described in connection with the embodiments of the present application may be implemented in hardware or by a processor executing software instructions. The software instructions may be composed of corresponding software modules, and the software modules may be stored in a RAM, flash memory, ROM, EPROM, EEPROM, register, hard disk, removable hard disk, CD-ROM, or any other form of storage medium well known in the art. An exemplary storage medium is coupled to the processor such that the processor can read information from the storage medium and write information to the storage medium. Of course, the storage medium may also be a component of the processor. The processor and the storage medium may be located in an ASIC. Additionally, the ASIC may be located in a controller or a resource allocation emulation device. Of course, the processor and the storage medium may also exist as discrete components.

[0200] Through the description of the above embodiments, those skilled in the art can clearly understand that, for the convenience and simplicity of description, only the above division of each functional module is used as an example. In actual applications, the above functions may be allocated to different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above.

Claims

1. An information query method, characterized in that, applied to an information query system, the information query system includes a knowledge base, the knowledge base includes knowledge information related to one or more query tasks, and the method includes: responding to a query request, and obtaining first knowledge information matching the query request from the knowledge base; generating prompt information related to the query request according to the first knowledge information; outputting a query result corresponding to the query request according to the query request and the prompt information.

2. The method according to claim 1, characterized in that, the obtaining of the first knowledge information matching the query request from the knowledge base includes: obtaining the first knowledge information matching the query request from the knowledge base according to a first preset condition, and the first preset condition is related to one or more of the following: knowledge relevance, time relevance, type relevance.

3. The method according to claim 2, characterized in that, the query request carries query content, and the obtaining of the first knowledge information matching the query request from the knowledge base according to the first preset condition includes: taking the knowledge information whose knowledge relevance between the updated content of the knowledge information and the query content is greater than or equal to a preset threshold as the first knowledge information.

4. The method according to claim 2 or 3, characterized in that, the query request carries a query time, and the obtaining of the first knowledge information matching the query request from the knowledge base according to the first preset condition includes: taking the knowledge information whose time relevance between the update time of the knowledge information and the query time is greater than or equal to a preset threshold as the first knowledge information.

5. The method according to any one of claims 2-4, characterized in that, the query request carries a query type, and the obtaining of the first knowledge information matching the query request from the knowledge base according to the first preset condition includes: taking the knowledge information whose type relevance between the update type of the knowledge information and the query type is greater than or equal to a preset threshold as the first knowledge information.

6. The method according to any one of claims 1-5, characterized in that, the query request includes a task identifier, the knowledge base includes a plurality of sub-knowledge bases, the plurality of sub-knowledge bases correspond to different task identifiers, and the obtaining of the first knowledge information matching the query request from the knowledge base includes: obtaining a sub-knowledge base matching the query request from the plurality of sub-knowledge bases according to the task identifier; obtaining the first knowledge information matching the query request from the sub-knowledge base.

7. The method according to any one of claims 1-6, characterized in that, the generating of the prompt information related to the query request according to the first knowledge information includes: generating the prompt information related to the query request based on a prompt logic algorithm according to the first knowledge information, and the prompt information is used to supplement the query content carried by the query request.

8. The method according to claim 7, characterized in that, The step of generating prompt information related to the query request based on the prompt logic algorithm according to the first knowledge information includes: Based on the prompt logic algorithm, obtaining multiple candidate prompt information corresponding to the first knowledge information; The prompt information is determined from the one or more candidate prompt information, where the prompt information is candidate prompt information whose similarity with the query content is greater than or equal to a preset threshold among the multiple candidate prompt information.

9. The method according to any one of claims 1 to 8, It is characterized in that The query request carries a device identifier of the network device that initiates the query request, and outputting a query result corresponding to the query request according to the query request and the prompt information includes: Obtaining a query result corresponding to the query request according to the query request and the prompt information; The query result is output to the network device according to the device identification.

10. The method according to claim 9, It is characterized in that The obtaining, according to the query request and the prompt information, a query result corresponding to the query request includes: Based on the information query model, a query result corresponding to the query request is obtained according to the query request and the prompt information.

11. The method according to any one of claims 1 to 10, It is characterized in that The method further comprises: In response to a knowledge update request, the knowledge base is updated according to the knowledge information carried in the knowledge update request.

12. The method according to claim 11, It is characterized in that The knowledge information includes update content, and updating the knowledge base according to the knowledge information carried in the knowledge update request includes: The knowledge base is updated according to the updated content.

13. The method according to claim 12, It is characterized in that The knowledge information includes an update time, and the updating of the knowledge base according to the knowledge information carried in the knowledge update request includes: The knowledge base is updated according to the update time.

14. The method according to claim 12 or 13, It is characterized in that The knowledge information includes an update type, and updating the knowledge base according to the knowledge information carried in the knowledge update request includes: The knowledge base is updated according to the update type.

15. The method according to any one of claims 11 to 14, It is characterized in that The knowledge update request includes a task identifier, the knowledge base includes multiple sub-knowledge bases, and the multiple sub-knowledge bases correspond to different task identifiers. The updating of the knowledge base according to the knowledge information carried in the knowledge update request includes: According to the task identifier, acquiring a sub-knowledge base matching the knowledge update request from the multiple sub-knowledge bases; The sub-knowledge base is updated according to the knowledge information carried in the knowledge update request.

16. The method according to any one of claims 10 to 15, It is characterized in that The method further comprises: In response to the model optimization request, one or more model parameters related to the model optimization request in the information query model are optimized according to the knowledge information carried in the model optimization request.

17. The method according to claim 16, It is characterized in that The knowledge information includes an optimization type, and optimizing one or more model parameters related to the model optimization request in the information query model according to the knowledge information carried in the model optimization request includes: Matching second knowledge information from the knowledge base according to the knowledge information carried in the model optimization request; Optimizing one or more model parameters related to the model optimization request in the information query model according to the second knowledge information.

18. The method according to any one of claims 1-17, wherein, the information query system includes a knowledge answering system, the query request includes a question request, and the query result includes an answer result corresponding to the question request.

19. An information query method, wherein, applied to an information query system, the information query system includes a knowledge base, and the method includes: Responding to a knowledge update request, and updating the knowledge base according to the knowledge information carried in the knowledge update request.

20. The method according to claim 19, wherein, the knowledge information includes update content, and updating the knowledge base according to the knowledge information carried in the knowledge update request includes: Updating the knowledge base according to the update content.

21. The method according to claim 20, wherein, the knowledge information includes an update time, and updating the knowledge base according to the knowledge information carried in the knowledge update request includes: Updating the knowledge base according to the update time.

22. The method according to claim 20 or 21, wherein, the knowledge information includes an update type, and updating the knowledge base according to the knowledge information carried in the knowledge update request includes: Updating the knowledge base according to the update type.

23. The method according to any one of claims 19-22, wherein, the knowledge update request includes a task identifier, the knowledge base includes a plurality of sub-knowledge bases, the plurality of sub-knowledge bases correspond to different task identifiers, and updating the knowledge base according to the knowledge information carried in the knowledge update request includes: Obtaining a sub-knowledge base matching the knowledge update request from the plurality of sub-knowledge bases according to the task identifier; Updating the sub-knowledge base according to the knowledge information carried in the knowledge update request.

24. An information query method, wherein, applied to an information query system, the information query system includes a knowledge base, and the method includes: Responding to a model optimization request, and optimizing one or more model parameters related to the model optimization request in the information query model according to the knowledge information carried in the model optimization request.

25. The method according to claim 24, wherein, the knowledge information includes an optimization type, and optimizing one or more model parameters related to the model optimization request in the information query model according to the knowledge information carried in the model optimization request includes: Matching second knowledge information from the knowledge base according to the knowledge information carried in the model optimization request; Optimize one or more model parameters related to the model optimization request in the information query model according to the second knowledge information.

26. A network device Characterized in that The network device includes: A transceiver for transmitting and receiving signals; A memory for storing computer program instructions; A processor for executing the computer program instructions to enable the network device to implement the method according to any one of claims 1-18 or 19-23 or 24-25.

27. An information query system Characterized in that Includes one or more network devices, and the one or more network devices are used to support the information query system to implement the method according to any one of claims 1-18 or 19-23 or 24-25.

28. A computer-readable storage medium Characterized in that Computer program instructions are stored on the computer-readable storage medium, and when the computer program instructions are executed by a processing circuit, the method according to any one of claims 1-18 or 19-23 or 24-25 is implemented.

29. A computer program product containing instructions Characterized in that When the computer program product runs on a computer, the computer is caused to execute the method according to any one of claims 1-18 or 19-23 or 24-25.

30. A chip system Characterized in that The chip system includes a processing circuit and a storage medium, and computer program instructions are stored in the storage medium; when the computer program instructions are executed by the processing circuit, the method according to any one of claims 1-18 or 19-23 or 24-25 is implemented.

Images