Information retrieval method and apparatus

Abstract

An information retrieval method and apparatus, which are used for improving the stability of a retrieval device in obtaining an accurate retrieval result. The method comprises: on the basis of description text of a target task, a first network element first using a first retrieval policy to perform retrieval to obtain a first retrieval result, and sending the first retrieval result to a second network element; correspondingly, the second network element executing the target task on the basis of the first retrieval result, and generating feedback information for indicating the execution status of the target task; when the feedback information indicates that the execution of the target task is abnormal, the first network element determining a second retrieval policy on the basis of a cause of the abnormal execution of the target task, and using the second retrieval policy to perform retrieval to obtain a second retrieval result; and the first network element sending the second retrieval result to the second network element, such that the second network element can execute the target task on the basis of the second retrieval result.

Description

An information retrieval method and apparatus

[0001] Cross-references to related applications

[0002] This application claims priority to Chinese Patent Application No. 202510182264.1, filed on February 18, 2025, entitled "An Information Retrieval Method and Apparatus", the entire contents of which are incorporated herein by reference. Technical Field

[0003] This application relates to the field of artificial intelligence technology, and in particular to an information retrieval method and apparatus. Background Technology

[0004] With the continuous development of artificial intelligence (AI) and related technologies, large language models (LLMs) can be used to orchestrate and configure various network functions and resources in a more flexible and dynamic way. LLMs can optimize and manage communication services over the network; for example, they can parse network logs, identify potential problems in the network, and propose corresponding solutions.

[0005] Current solutions for optimizing and managing large language models primarily rely on a network architecture centered around a large language model and an agent. In this architecture, the large language model retrieves task-relevant information from a vast amount of stored data, and the agent can then orchestrate and configure network functions and resources based on this information.

[0006] Retrieval-augmented generation (RAG) provides large language models with access to internal databases within their feature domains. Since these internal databases can include historical inference results from the large model (such as question-and-answer text records) and other data (such as Wikipedia data and expert knowledge bases), no supervised fine-tuning of the large language model is required. Existing techniques typically employ RAG to optimize the output of large language models by referencing knowledge bases outside of their local training data sources, aiming to improve the accuracy and reduce the perplexity of the output results. However, RAG-based retrieval enhancement schemes often suffer from issues such as missing important information, redundant and repetitive retrieval information, and outdated retrieval information, resulting in poor stability in achieving accurate retrieval results. Summary of the Invention

[0007] This application provides an information retrieval method and apparatus to improve the stability of the retrieval results obtained by the retrieval device.

[0008] Firstly, embodiments of this application provide an information retrieval method, which can be executed by a first network element, or in other words, the method can be applied to a first network element. In this application, the first network element can be one or more network elements in a mobile communication technology network based on a service-oriented architecture. The method may include: the first network element receiving a description text of a target task from a second network element; wherein the second network element is used to execute the target task, which is a task in a network communication service; the first network element performing a search in a knowledge base according to the description text using a first search strategy to obtain a first search result corresponding to the target task; the first network element sending the first search result to the second network element and receiving feedback information from the second network element; wherein the feedback information is used to instruct the second network element on the execution status of the target task based on the first search result; when the feedback information indicates that the target task execution is abnormal, the first network element obtains the reason for the target task execution abnormality and determines a second search strategy based on the reason; wherein the first search strategy is different from the second search strategy; the first network element performing a search in a knowledge base according to the description text using the second search strategy to obtain a second search result corresponding to the target task; and the first network element sending the second search result to the second network element.

[0009] Using this method, the first network element can first use a first retrieval strategy to obtain a first retrieval result based on the description text of the target task, and then send the first retrieval result to the second network element. Correspondingly, the second network element can execute the target task based on the first retrieval result and generate feedback information indicating the execution status of the target task. When the feedback information indicates an abnormal execution of the target task, the first network element can determine a second retrieval strategy based on the reason for the abnormal execution and use the second retrieval strategy to obtain a second retrieval result. The first network element can then send the second retrieval result to the second network element, enabling the second network element to execute the target task based on the second retrieval result. Since the second retrieval strategy is determined based on the reason for the abnormal execution of the target task, accurate retrieval results can be reliably obtained based on the second retrieval strategy. Therefore, by using this method, when the target task cannot be executed normally based on the first retrieval result, a second retrieval strategy can be determined based on the reason for the abnormal execution of the target task to obtain accurate second retrieval results, thereby improving the stability of obtaining accurate retrieval results.

[0010] In one possible design, the first network element updates the knowledge base based on the acquired external data; the description text includes the execution cycle of the target task, and the first network element performs periodic searches in the knowledge base based on the execution cycle and the description text, using the corresponding search strategy to obtain the search results corresponding to the target task; when the third search result obtained in the current execution cycle is inconsistent with the search result obtained in the previous execution cycle, the first network element sends the third search result to the second network element.

[0011] Through this design, the first network element can continuously acquire external data and update the knowledge base based on the acquired external data to ensure the timeliness of the data in the knowledge base. The first network element can periodically search the knowledge base based on the description text according to the execution cycle of the target task, and compare the third search result obtained in the current execution cycle with the search result obtained in the previous execution cycle. When the third search result is inconsistent with the search result obtained in the previous execution cycle, it indicates that new data related to the target task has been updated in the knowledge base. The first network element then sends the third search result to the second network element, thereby ensuring the timeliness of the search results obtained by the second network element. Understandably, when the third search result is consistent with the search result obtained in the previous execution cycle, it indicates that the data related to the target task in the database has not changed, and the first network element does not need to send the third search result to the second network element to avoid unnecessary communication between the two network elements.

[0012] In one possible design, the feedback information includes retrieval result verification information. Before the first network element sends the second retrieval result to the second network element, the second retrieval result is verified based on the retrieval result verification information and the verification passes.

[0013] Optionally, the verification information for the search results includes at least one of the following: data format requirements for the second network element, access permission restrictions, and input character restrictions.

[0014] With this design, before the first network element sends the second search result to the second network element, it can verify the second search result based on the search result verification information and ensure that the second search result meets the requirements of the second network element to execute the target task, thus avoiding the target task being executed abnormally multiple times.

[0015] In one possible design, the first retrieval strategy and the second retrieval strategy include at least one of the following: keywords describing the text, a knowledge base, and the retrieval scope.

[0016] This design can improve the accuracy of search results obtained by using either the first or second search strategy.

[0017] In one possible design, the reasons for the abnormal execution of the target task include at least one of the following: inability to parse the search results, inability to access links in the search results, search results not matching the target task, search results not meeting the input character restrictions, and inability to execute the target task normally.

[0018] In one possible design, the first search result and / or the second search result includes multiple search data. The first network element processes the multiple search data and merges them into one search result. The processing of the multiple search data includes deleting irrelevant search data, deduplication, identifying contradictory search data, and merging multiple search data.

[0019] With this design, when the retrieval result obtained by the first network element includes multiple retrieval data, the first network element can merge the multiple retrieval data into one retrieval result to send the necessary and concise retrieval data to the second network element, thereby avoiding excessive retrieval data from interfering with the accuracy of the model and reducing the amount of information transmitted between the first network element and the second network element.

[0020] Secondly, embodiments of this application provide an information retrieval device. This device has the functionality to implement various possible methods of the first aspect described above. This functionality can be implemented by hardware or by hardware executing corresponding software. The hardware or software includes one or more units or modules corresponding to the aforementioned functionality.

[0021] Thirdly, an electronic device is provided, comprising: a memory for storing computer program instructions; and a processor for executing the computer program instructions to support the electronic device in implementing the methods provided in any of the first aspects above.

[0022] Fourthly, this application also provides a computer-readable storage medium storing instructions that, when executed on a device, cause the device to perform the method of any possible design of the first aspect.

[0023] Fifthly, this application also provides a computer program product that, when run on a device, causes the device to perform any possible design of any aspect of the first aspect of the embodiments of this application.

[0024] The technical effects brought about by the second to fifth aspects above can be found in the description of the beneficial effects of the corresponding solutions in the first aspect above, and will not be repeated here. Attached Figure Description

[0025] Figure 1A is a schematic diagram of the architecture of an information retrieval system provided in an embodiment of this application;

[0026] Figure 1B is a schematic diagram of the structure for retrieval based on the RAG method provided in an embodiment of this application;

[0027] Figure 2 is a schematic diagram of the architecture of a wireless communication system provided in an embodiment of this application;

[0028] Figure 3 is a schematic diagram of the architecture of another wireless communication system provided in an embodiment of this application;

[0029] Figure 4 is a flowchart illustrating an information retrieval method improved according to an embodiment of this application;

[0030] Figure 5 is a flowchart illustrating another information retrieval method provided in an embodiment of this application;

[0031] Figure 6 is a flowchart illustrating another information retrieval method provided in an embodiment of this application;

[0032] Figure 7 is a flowchart illustrating another information retrieval method provided in an embodiment of this application;

[0033] Figure 8 is a flowchart illustrating an information retrieval device provided in an embodiment of this application;

[0034] Figure 9 is a flowchart illustrating another information retrieval device provided in an embodiment of this application. Detailed Implementation

[0035] To make the objectives, technical solutions, and advantages of this application clearer, a further detailed description of this application will be provided below in conjunction with the accompanying drawings. The specific operating methods and functional descriptions in the method embodiments can also be applied to the device embodiments or system embodiments.

[0036] The following explanations of some terms used in the embodiments of this application are provided to facilitate understanding by those skilled in the art.

[0037] 1) Large language models (LLMs) are deep learning models trained on large amounts of text data that can generate natural language text or understand the meaning of language text. LLMs can handle various natural language tasks, such as text classification, question answering, and dialogue, and are an important pathway to artificial intelligence.

[0038] 2) Agent-based Next-Generation Core Network (A-Core): In A-Core, a system composed of network agents and traditional network functions in various core networks can orchestrate, deploy, execute, and reclaim end-to-end network slices for new services, and optimize network performance based on information fed back from the network environment.

[0039] 3) Network large model (NLM): NLM can perform analysis tasks based on input information and manage network agents and other network components.

[0040] 4) Intelligent function query (IFQ) can query the functions required by various intelligent agents based on their task and requirement descriptions.

[0041] It should be understood that in the embodiments of this application, "at least one" means one or more, and "more than one" means two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can mean: A exists alone, A and B exist simultaneously, or B exists alone, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, or c can mean: a, b, c, a and b, a and c, b and c, or a, b, and c, where a, b, and c can be single or multiple. Furthermore, the terms "first," "second," etc., used in the embodiments of this application are only for descriptive purposes and should not be construed as indicating or implying relative importance or order.

[0042] The following description, in conjunction with the accompanying drawings, illustrates the applicable scenarios for the embodiments of this application.

[0043] The information retrieval method provided in this application embodiment is applicable to the A-core system architecture. Figure 1A is a schematic diagram of an A-core system architecture. The A-core system architecture can consist of multiple agents and various common components. Based on this architecture, the network can be managed and controlled through multiple agents, network tools, capabilities, and memory components. Data and information are exchanged between the agents and common components in natural language via agent-based interfaces (ABIs). The agents in this architecture can include network agents, planning agents, assembly agents, connection agents, and execution agents.

[0044] The following describes the network element functions involved in this application:

[0045] Network agents (NetAgents) can be used to perceive and collect data in the network environment and learn from historical data to improve their decision-making capabilities, enabling flexible orchestration, deployment, creation, and recycling of generative networks for new services. Network agents can also provide data support to other agents.

[0046] A planning agent is an entity that receives customized task requests. It analyzes the user-requested business (such as a description of the requested business) and breaks it down into multiple executable tasks, outputting a list of these tasks. The task list includes a task description and dependency metrics for each task.

[0047] The assemble agent can receive a list of tasks from the planning agent, configure corresponding network functions and / or application functions for each task in the list, and determine the corresponding functional topology based on the relationships between multiple tasks.

[0048] The connection agent can determine the user equipment (UE) and / or radio access network (RAN) that make up the end-to-end slice, as well as the corresponding configuration parameters and resources, based on the output of the assembly agent. This enables the management and control of the connection topology of the end-to-end slice, as well as the management of the accessing UEs and slice selection.

[0049] Execution agents can be used to manage and control the lifecycle of network slices. For example, execution agents can deploy network and / or application function instances, monitor the operational status of network slices, dynamically update and reclaim network slices, etc.

[0050] Net generative pre-trained transformers (NetGPT) can fine-tune existing large language models by moving network-specific data, and generate inference results (such as task decomposition, function selection, etc.) related to network slices for each agent based on input prompts.

[0051] The public memory module can be used to collect, vectorize, and store network state information and expert knowledge, and to create corresponding indexes for the stored data. The public memory module can retrieve relevant information based on data query requests from other components. During the retrieval process, it can perform searches based on RAG prompts to improve the accuracy of NetGPT's inferences based on the retrieval results. Furthermore, the public memory module can have simple logical reasoning capabilities to achieve basic retrieval and summarization functions.

[0052] ABI refers to the interface between an agent and a public component. It can understand and determine the destination address of a message and forward the message to the appropriate agent or public component. In other words, when a consumer sends a message to an agent or public component, it only needs to provide a target description, and the ABI can understand the destination address of the message and forward it to the appropriate agent or public component.

[0053] Currently, when an agent performs a task, it can send a task description to the common memory module. The common memory module can then retrieve relevant historical tasks and processing methods from the corresponding knowledge base based on the task description, and send the retrieval results to the second network element so that the agent can execute the task based on the results. The common memory module typically uses the RAG (Research and Analysis Group) method for retrieval.

[0054] As shown in Figure 1B, the retrieval process using RAG can be divided into two stages. The first stage involves creating an external database (such as an expert knowledge base and document repository) and converting this data into a format conducive to efficient retrieval. An embedded language model is used to store the data in a vector database. This stage typically involves offline data updates, meaning data updates are performed when the system is not running. The second stage involves retrieving the received query information, matching the query information with data in the database to identify highly relevant information, combining this information with the query information to generate a suggestion message, and sending this suggestion message to the LLM (Local Library Management) so that the LLM can generate more accurate and detailed search results based on the suggestion message.

[0055] However, RAG-based retrieval schemes suffer from problems such as generating information that is irrelevant to the query information or containing incorrect facts, resulting in low stability in generating accurate results.

[0056] Based on this, embodiments of this application provide an information retrieval method, which can be implemented through a first network element and a second network element. The first network element can be used to perform retrieval based on the description text of a target task, and the second network element can be used to execute the target task, which can be a task in a network communication service. For example, taking Figure 1A as an example, the first network element can be a common memory module, and the second network element can be any one of a planning agent, an assembly agent, a connection agent, or an execution agent.

[0057] In this method, a first network element can use a first retrieval strategy to search the knowledge base based on the description text of the target task sent by a second network element, obtain the first retrieval result corresponding to the target task, and send the first retrieval result to the second network element. The second network element can execute the target task based on the first retrieval result and send feedback information to the first network element indicating the execution status of the target task. When the feedback information indicates an execution error in the target task, the first network element determines a second retrieval strategy based on the reason for the error and uses the second retrieval strategy to search the knowledge base based on the description text, obtaining the second retrieval result corresponding to the target task. The first network element sends the second retrieval result to the second network element. Correspondingly, the second network element can execute the target task based on the second retrieval result. In this method, the second retrieval strategy is determined based on the reason for the execution error of the target task; therefore, obtaining the second retrieval result based on the second retrieval strategy can improve the stability of generating accurate retrieval results, thereby enabling the second network element to execute the target task normally based on the second retrieval result.

[0058] Figure 2 is a schematic diagram of the network architecture of 5G (5th generation wireless systems) based on a service-oriented architecture. The 5G network architecture shown in Figure 2 may include terminal devices, access network devices, and core network (CN) devices. Terminal devices access the data network (DN) through access network devices and core network devices. The core network devices include various network elements (NFs), such as some or all of the following: UDM network elements, UDR network elements, NEF network elements (not shown in the figure), application function (AF) network elements, policy control function (PCF) network elements, access and mobility management function (AMF) network elements, session management function (SMF) network elements, user plane function (UPF) network elements, NWDAF network elements, network repository function (NRF) network elements (not shown in the figure), and location management function (LMF) network elements (not shown in the figure).

[0059] Access network equipment can be radio access network (RAN) equipment. Examples include: base stations, evolved NodeBs (eNodeBs), transmission reception points (TRPs), next-generation NodeBs (gNBs) in 5G mobile communication systems, base stations in future mobile communication systems, or access nodes in wireless fidelity (WiFi) systems. It can also be modules or units that perform some of the functions of a base station; for example, it can be a central unit (CU) or a distributed unit (DU). RAN equipment can be macro base stations, micro base stations, indoor stations, relay nodes, or donor nodes. The embodiments of this application do not limit the specific technologies or equipment forms used in the RAN equipment.

[0060] Terminal devices can be user equipment, mobile stations, mobile terminals, etc. They can be widely used in various scenarios, such as device-to-device (D2D), vehicle-to-everything (V2X) communication, machine-type communication (MTC), the Internet of Things (IoT), virtual reality, augmented reality, industrial control, autonomous driving, telemedicine, smart grids, smart furniture, smart offices, smart wearables, smart transportation, and smart cities. Terminal devices can include mobile phones, tablets, computers with wireless transceiver capabilities, wearable devices, vehicles, urban air mobility vehicles (such as drones and helicopters), ships, robots, robotic arms, and smart home devices.

[0061] Access network equipment and terminal equipment can be fixed or mobile. They can be deployed on land, including indoors or outdoors, handheld or vehicle-mounted; they can also be deployed on water; and they can be deployed in the air on aircraft, balloons, and satellites. The embodiments of this application do not limit the application scenarios of the access network equipment and terminal equipment.

[0062] The core network equipment involved in this application is described below:

[0063] The AMF (Agency Flow Management) network element includes functions such as mobility management and access authentication / authorization. It is also responsible for transmitting user policies between terminal devices and the PCF (Programmable Flow Function). The SMF (Sustainable Flow Management) network element includes functions such as session management, execution of PCF-issued control policies, UPF (User Flow Filter) selection, and Internet Protocol (IP) address allocation for terminal devices. The UPF (User Flow Filter) network element, as the interface with the data network, includes functions such as user plane data forwarding, session / flow-level billing statistics, and bandwidth limiting. The UDM (User Flow Management) network element includes functions such as managing subscription data and user access authorization. The UDR (User Flow Filter) network element includes functions such as storing and retrieving subscription data, policy data, and application data. The NEF (Network Flow Filter) network element is used to support the opening of capabilities and events. The AF (Application Flow Filter) network element transmits application-side requirements to the network side, such as Quality of Service (QoS) requirements or user state event subscriptions. AF can be a third-party functional entity or an application server deployed by the operator. The PCF (Programmable Flow Function) network element includes policy control functions such as billing at the session and service flow levels, QoS bandwidth assurance and mobility management, and terminal device policy decisions. NRF network elements can be used to provide network element discovery functionality, providing network element information corresponding to the network element type based on requests from other network elements. NRF network elements also provide network element management services, such as network element registration, updates, deregistration, and network element status subscription and push.

[0064] Network data analytics function (NWDAF) network elements are primarily used to collect data (including one or more of the following: terminal device data, access network device data, core network data, and third-party application device data). This data can be data from the terminal device, access network device, core network element, or third-party application device itself, or data from the terminal device on that access network device, core network element, or third-party application device. The collected data is then analyzed, and the analysis results are output for use by the network, network management equipment, and applications in decision-making. NWDAF can utilize machine learning (ML) models for data analysis. 3GPP separates the training and inference functions of NWDAF. An NWDAF can support only model training, only data inference, or both. An NWDAF supporting model training can also be called a training NWDAF or an NWDAF supporting MTLF (MTLF for short). A training NWDAF can train a model based on the acquired data to obtain a trained model. An NWDAF that supports data inference can also be called an inference NWDAF, or an NWDAF that supports analytics logical function (AnLF) (abbreviated as AnLF). An inference NWDAF can input input data into a trained model to obtain analysis results or inference data. In this embodiment, a training NWDAF refers to an NWDAF that at least supports model training. As a possible implementation, a training NWDAF can also support data inference. An inference NWDAF refers to an NWDAF that at least supports data inference. As a possible implementation, an inference NWDAF can also support model training. If an NWDAF supports both model training and data inference, it can be called a training NWDAF, an inference NWDAF, a training-inference NWDAF, or simply an NWDAF. In this embodiment, an NWDAF can be a separate network element or can be co-located with other network elements, such as being placed in a PCF network element or an AMF network element.

[0065] As one implementation method, the data analysis network element in the embodiments of this application can be the aforementioned NWDAF network element, or it can be a network element that has the functions of the aforementioned NWDAF network element in future communications. For ease of explanation, the following description uses an NWDAF network element as the data analysis network element, and the NWDAF network element will be abbreviated as NWDAF.

[0066] The ADRF network element, abbreviated as ADRF, can be used to store model-related data. For example, ADRF can store historical data collected by the network and historical analysis results generated by NWDAF. This data can be generated by AnLF. ADRF can provide model-related data to MTLF upon request. ADRF and AnLF are not shown in Figure 2 or Figure 3.

[0067] In this application, AnLF can be used to perform inference, derive analytical information, and disclose analytical services. Here, analysis can refer to deriving statistical information and / or predictions based on requests from analytical consumers. MTLF can be used to train machine learning (ML) models and disclose new training services, for example, by providing a trained ML model to AnLF. In this application, for the process of acquiring data related to ML models, AnLF can act as a data producer network element, denoted as NFp; MTLF can act as a data consumer network element, denoted as NFc.

[0068] The LMF (Local Position Controller) network element is used to manage the location information of terminal devices. It can calculate or verify the location of the terminal devices and / or estimate their speed, providing the accuracy of the estimation. The LMF network element can receive location requests from the AMF (Agency Position Controller) network element via the Nlmf interface. The granularity of the terminal's location calculated by the LMF network element can be one or more of the following: longitude, latitude, altitude, cell, and Global Positioning System (GPS).

[0069] A Domain Provider (DN) is a network located outside of the carrier's network. A carrier's network can connect to multiple DNs, and various services can be deployed on a DN, providing data and / or voice services to terminal devices. For example, a DN might be the private network of a smart factory. Sensors installed in the workshop can act as terminal devices, and a control server for these sensors is deployed within the DN. The control server provides services to the sensors. Sensors can communicate with the control server, receive instructions from it, and transmit the collected sensor data back to the control server accordingly. Another example is a DN serving as an internal office network for a company. Employees' mobile phones or computers can act as terminal devices, accessing information and data resources on the company's internal office network.

[0070] It is understood that the above network elements are examples of one implementation method. This application does not exclude the possibility that in future wireless communication systems, there may be network elements or devices with the above network element functions that have other names or other forms.

[0071] In Figure 2, Nudr, Npcf, Namf, Nudm, Nsmf, Naf, and Nnwdaf are the service interfaces provided by UDR, PCF, AMF, UDM, SMF, AF, and NWDAF, respectively, used to call the corresponding service operations. N1, N2, N3, N4, and N6 are interface sequence numbers, and the meanings of these interface sequence numbers are as follows:

[0072] 1) N1: The interface between the AMF network element and the terminal device, which can be used to transmit non-access stratum (NAS) signaling (such as QoS rules from the AMF network element) to the terminal device.

[0073] 2) N2: The interface between the AMF network element and the access network equipment, which can be used to transmit radio bearer control information from the core network side to the access network equipment.

[0074] 3) N3: The interface between the access network equipment and the UPF network element, mainly used to transmit uplink and downlink user plane data between the access network equipment and the UPF network element.

[0075] 4) N4: The interface between SMF network elements and UPF network elements. It can be used to transmit information between the control plane and the user plane, including the distribution of forwarding rules, QoS rules, traffic statistics rules, etc. from the control plane to the user plane, as well as the reporting of information from the user plane.

[0076] 5) N6: The interface between the UPF network element and the DN, used to transmit uplink and downlink user data streams between the UP network element F and the DN.

[0077] Additionally, in the architecture shown in Figure 2, Nadrf can serve as a service interface for ADRF.

[0078] Figure 3 is a schematic diagram of a 5G network architecture based on point-to-point interfaces. The functions of the network elements are described in Figure 2, and will not be repeated here. The main difference between Figure 3 and Figure 2 is that the interfaces between the control plane network elements in Figure 2 are service-oriented interfaces, while the interfaces between the control plane network elements in Figure 3 are point-to-point interfaces.

[0079] In the architecture shown in Figure 3, the interface names and functions between the various network elements of the core network are as follows:

[0080] 1) N5: The interface between AF network element and PCF network element, which can be used for application service request distribution and network event reporting.

[0081] 2) N7: The interface between PCF network elements and SMF network elements, which can be used to issue protocol data unit (PDU) session granularity and service data flow granularity control strategies.

[0082] 3) N8: The interface between the AMF network element and the UDM network element. It can be used by the AMF network element to obtain access and mobility management related subscription data and authentication data from the UDM network element, as well as by the AMF network element to register the current mobility management information of the terminal device with the UDM network element.

[0083] 4) N9: User plane interface between UPF network elements, used to transmit uplink and downlink user data streams between UPF network elements.

[0084] 5) N10: The interface between the SMF network element and the UDM network element. It can be used for the SMF network element to obtain session management-related subscription data from the UDM network element, and for the SMF network element to register terminal device current session-related information with the UDM network element.

[0085] 6) N11: The interface between SMF network elements and AMF network elements. It can be used to transmit PDU session tunnel information between access network devices and UPF network elements, transmit control messages sent to terminal devices, and transmit radio resource control information sent to access network devices.

[0086] 7) N15: The interface between PCF network elements and AMF network elements, which can be used to issue terminal equipment policies and access control related policies.

[0087] 8) N23: The interface between the PCF network element and the NWDAF network element. The NWDAF network element can collect data from the PCF network element through this interface. It should be noted that the NWDAF network element can also have interfaces with other devices (such as AMF network elements, UPF network elements, access network devices, terminal devices, etc.), which are not fully shown in the figure.

[0088] 9) N35: Interface between UDM network element and UDR network element, which can be used by UDM network element to obtain user subscription data information from UDR network element.

[0089] 10) N36: The interface between PCF network elements and UDR network elements, which can be used by PCF network elements to obtain policy-related contract data and application data related information from UDR network elements.

[0090] It is understood that the aforementioned network elements or functions can be network components in hardware devices, software functions running on dedicated hardware, or virtualized functions instantiated on a platform (e.g., a cloud platform). As one possible implementation method, the aforementioned network elements or functions can be implemented by a single device, multiple devices working together, or a functional module within a single device; this application embodiment does not specifically limit this. Furthermore, the structures shown in Figures 2 and 3 can be coupled with the system architecture shown in Figure 1A to implement the information retrieval method provided in this application embodiment.

[0091] To improve the stability of accurate retrieval results obtained by the first network element using a large language model, this application provides an information retrieval method. The method is described below with reference to the accompanying drawings. Referring to Figure 4, the method includes:

[0092] S401: The second network element sends a description text of the target task to the first network element. Correspondingly, the first network element receives the description text of the target task from the second network element. The target task is a task within the network communication service.

[0093] In this step, the description text of the target task can refer to natural language text used to describe the target task. For example, the description text of the target task could be "analyze the internet access behavior of users in the current community". The second network element can send the description text of the target task to the first network element. Correspondingly, the first network element receives the description text of the target task from the second network element. For example, as shown in Figure 1A, the first network element can send the description text of the target task to the first network element via the ABI. Correspondingly, the first network element can receive the description text of the target task from the first network element via the ABI.

[0094] S402: The first network element searches the knowledge base using the first retrieval strategy based on the description text to obtain the first retrieval result corresponding to the target task.

[0095] In this step, after obtaining the description text, the first network element can determine a first retrieval strategy based on the description text and use the first retrieval strategy to search the knowledge base to obtain the first retrieval result corresponding to the target task. The first retrieval strategy may include at least one of the following: keywords in the description text, the knowledge base, and the retrieval scope. The knowledge base may include description texts and retrieval results of historical tasks, a knowledge vector base (such as expert knowledge and useful document information), etc.

[0096] Optionally, the first network element can perform text analysis processing on the descriptive text to obtain keywords. The first network element can determine the corresponding knowledge base and search scope based on the descriptive text. For example, the descriptive text may include the identifier of a second network element, and the first network element can determine the corresponding knowledge base and search scope of the second network element based on the stored correspondence between network elements and knowledge bases and search scopes. Alternatively, the first network element can determine the corresponding knowledge base and search scope of the descriptive text based on its semantic information. Alternatively, the knowledge base and search scope corresponding to the descriptive text can also be the default knowledge base and search scope set in the database.

[0097] For example, the first retrieval strategy includes keywords for the descriptive text, a knowledge base, and a retrieval scope. The first network element can, according to the first retrieval strategy, search for the keywords of the descriptive text within the retrieval scope of the knowledge base corresponding to the descriptive text, determine one or more data items in the knowledge base that are relevant to the keywords of the descriptive text, and process the one or more data items to generate the first retrieval result.

[0098] Optionally, the first search result can be data obtained by processing the search data obtained by the first network element. That is, after obtaining the search data, the first network element processes the search data to obtain the first search result. The processing of the search data by the first network element can include deduplication, compression, or other processing to reduce the amount of information in the first search result, thereby ensuring that the first search result meets the input requirements of the second network element. Simultaneously, it can also reduce the communication overhead between the first and second network elements.

[0099] For example, when the first search result includes multiple search data, the first network element can process the multiple search data and merge them into one search result. The processing methods of the first network element for the multiple search data include deleting irrelevant search data, deduplication, identifying contradictory search data, and merging multiple search data.

[0100] Deleting irrelevant search data can refer to removing data from multiple search results that is not related to the descriptive text. Deduplication can refer to removing duplicate data from multiple search results. Identifying contradictory search data can refer to identifying the correct data among contradictory search results. For example, the first network element can determine the correct data among contradictory search results based on the validity period of the search data and / or the quantity and quality parameters of similar data among multiple search results. Merging multiple search results can refer to extracting and summarizing multiple search results into one. For example, the first network element can calculate the similarity between multiple search results and the descriptive text, weight the multiple search results according to the similarity, and merge the multiple search results into one search result based on the weights.

[0101] For example, when there are multiple search data in the first search result, the first network element can delete data that is not related to the description text, delete duplicate data, and identify the correct data in contradictory search data. It can also extract and summarize the remaining search data and merge the multiple search data into one search data, which is the first search result.

[0102] Optionally, the first search result can be the search data obtained by the first network element, that is, the first network element does not need to process the search data obtained by the search, so as to respond quickly and reduce the computational load of the first network element.

[0103] Furthermore, the first search result can include both structured and unstructured data. For example, the first search result could be: "An analysis of the historical network usage of all users in the current community is as follows: 50% of users use the network for more than 8 hours a day, 30% of users use the network for 2-8 hours a day, and another 20% of users use the network for less than 2 hours a day. The main usage scenarios include social networking platforms, video platforms, etc.; Cell ID: 93; Information source: https: / / dataservice / v1." Here, "An analysis of the historical network usage of all users in the current community is as follows: 50% of users use the network for more than 8 hours a day, 30% of users use the network for 2-8 hours a day, and another 20% of users use the network for less than 2 hours a day. The main usage scenarios include social networking platforms, video platforms, etc." is unstructured data, while "Cell ID: 93" is structured data.

[0104] S403: The first network element sends the first search result to the second network element. Correspondingly, the second network element receives the first search result from the first network element.

[0105] Taking Figure 1A as an example, the first network element can send the first search result to the second network element via the ABI. Correspondingly, the second network element can receive the first search result from the first network element via the ABI.

[0106] S404: The second network element executes the target task based on the first search result and generates feedback information to indicate the execution status of the target task.

[0107] In one or more embodiments, after receiving the first search result, the second network element can first determine whether the data in the first search result meets the requirements of the second network element. For example, the requirements of the second network element include data format requirements, i.e., the data format of the data in the first search result meets the data format requirements of the second network element; access permission restrictions, i.e., the second network element can access all content in the first search result (including cloud services, resource links, websites, etc.); and input character restrictions, i.e., the data in the first search result is less than or equal to the input character limit of the second network element. When the data in the first search result does not meet the requirements of the second network element, the second network element terminates the execution of the target task and generates feedback information to indicate an abnormality in the execution of the target task.

[0108] When the data in the first search result meets the requirements of the second network element, the second network element can determine the specific implementation steps of the target task based on the first search result, and execute the target task according to the specific implementation steps. For example, the second network element can combine the first search result and the task description into a prompt input, and use its internal AgentGPT to arrange the target task into multiple subtasks based on the prompt input, and execute the multiple subtasks sequentially. When the target task execution encounters an error, the second network element terminates the execution of the target task and generates feedback information indicating the error.

[0109] Optionally, the feedback information regarding the execution status of the target task may include whether the target task's execution status is normal or abnormal. For example, the feedback information may use binary data to represent the target task's execution status, where 0 indicates that the target task is executing abnormally, and 1 indicates that the target task is executing normally.

[0110] Optionally, the feedback information may include the execution status and / or execution condition information of the target task. For example, when the execution condition of the target task is normal, the feedback information may only include the execution status of the target task, i.e., the execution status of the target task is normal. As another example, when the execution condition of the target task is abnormal, the feedback information may include both the execution status and execution condition information of the target task. That is, the execution status of the target task is abnormal and a descriptive message about the abnormal execution of the target task. Alternatively, when the execution status of the target task is abnormal, the feedback information may only include the execution condition information of the target task, i.e., a descriptive message about the abnormal execution of the target task. Accordingly, the first network element can determine that the execution status of the target task is abnormal based on this descriptive information.

[0111] S405: The second network element sends feedback information to the first network element. Correspondingly, the first network element receives the feedback information from the second network element.

[0112] Taking Figure 1A as an example, the second network element can send feedback information to the first network element through the ABI. Correspondingly, the first network element can receive feedback information from the second network element through the ABI.

[0113] S406: When the feedback information indicates that the target task is being executed normally, the first network element can terminate the information retrieval task of the target task.

[0114] S407: When feedback information indicates an abnormality in the execution of the target task, the first network element obtains the reason for the abnormality and determines a second retrieval strategy based on the reason. The first retrieval strategy differs from the second retrieval strategy.

[0115] In one or more embodiments, when the execution status of the target task is abnormal, the feedback information may include the reason for the abnormal execution of the target task. Accordingly, the first network element can obtain the reason for the abnormal execution of the target task from the feedback information.

[0116] Alternatively, when the target task execution status is abnormal, the feedback information may include log information of the target task execution, and the first network element can determine the reason for the target task execution abnormality based on the log information.

[0117] Optionally, the reasons for the abnormal execution of the target task may include at least one of the following: inability to parse the search results, inability to access the links in the search results, the search results not matching the target task, the search results not meeting the input character restrictions, or inability to execute the target task normally.

[0118] For example, reasons for the abnormal execution of the target task include: "The data format in the first search result is incorrect or the necessary fields are missing, so the content in the first search result cannot be parsed correctly," "The access permissions of the second network element are insufficient, resulting in the inability to access the resource links in the first search result," "The relevance of the retrieved information is low due to keyword mismatch or inconsistent data content," "The execution of the target task exceeds the memory or computing resource limits, resulting in task failure," "The first search result does not meet the input character restrictions," "The arrangement logic is incorrect, the task execution order or condition judgment is incorrect," "The correct query conditions cannot be generated or the target data source is unavailable," etc.

[0119] In one or more embodiments, the first network element can determine a second search strategy based on the reason for the abnormal execution of the target task. Optionally, the first network element can adjust the first search strategy based on the reason for the abnormal execution of the target task to obtain a second search strategy. For example, when the reason for the abnormal execution of the target task is "low relevance of the retrieved information due to keyword mismatch or inconsistent data content," the first network element can redetermine the keywords corresponding to the description text of the target task based on this reason. As another example, when the reason for the abnormal execution of the target task is "insufficient access permissions of the second network element, resulting in inability to access resource links in the first search results," the first network element can redetermine the search scope or search database that satisfies the access permissions of the second network element based on this reason.

[0120] Understandably, the second search strategy can be obtained by adjusting the first search strategy based on the cause of the target task execution anomaly. Alternatively, the second search strategy can also be a search strategy generated based on the cause of the target task execution anomaly. The second search strategy includes at least one of the following: keywords describing the text, a knowledge base, and a search scope.

[0121] S408: The first network element uses the second retrieval strategy to search the knowledge base based on the description text of the target task, and obtains the second retrieval result corresponding to the target task.

[0122] The specific method for the first network element to perform a search in the knowledge base using the second search strategy can be found in the above description of the specific method for performing a search in the knowledge base using the first search strategy. Since the second search strategy is generated based on the cause of the target task's execution anomaly, it can improve the accuracy of the second search results.

[0123] In one or more embodiments, the first network element can verify the second search result based on the search result verification information and pass the verification. The search result verification information may include all or part of the requirements of the second network element, and may be included in the feedback information. The first network element verifies the second search result based on the search result verification information. When the second search result does not meet the search result verification information, the first network element can correct the second search result based on the search result verification information, and the corrected second search result can meet the requirements of the second network element. It is understood that this method enables the second network element to normally execute the target task based on the second search result, avoiding the second network element repeatedly reporting abnormal target task execution information to the first network element.

[0124] Optionally, the verification information for the search results may include at least one of the following: data format requirements for the second network element, access permission restrictions, and input character restrictions.

[0125] For example, the first network element can verify whether the second search result meets the data format requirements of the second network element. When the second search result does not meet the data format requirements of the second network element, the first network element can adjust the format of the second search result based on the data format requirements of the second network element to make the second search result meet the data format requirements of the second network element. As another example, the first network element can compress the second search result based on the input character restrictions of the second network element to make the second search result meet the input character restrictions of the second network element. The compression process can be a simplification of the second search result, i.e., deleting some content from the second search result or extracting key content from the second search result.

[0126] S409: The first network element sends the second search result to the second network element. Correspondingly, the second network element receives the second search result from the first network element.

[0127] Taking Figure 1A as an example, the first network element can send the second search result to the second network element through the ABI. Correspondingly, the second network element can receive the second search result from the first network element through the ABI.

[0128] Understandably, after receiving the second search result, the second network element can execute the target task based on the second search result.

[0129] Through steps S401-S409 above, the first network element can first send the first retrieval result corresponding to the target task to the second network element. Correspondingly, the second network element can execute the target task based on the first retrieval result and send feedback information to the first network element indicating the execution status of the target task. When the second network element encounters an error while executing the target task, the first network element can determine a second retrieval strategy based on the reason for the error and use the second retrieval strategy to re-retrieve and generate the second retrieval result corresponding to the target task. Understandably, this method can improve the accuracy of the retrieval results generated by the first network element using a large language model, enabling the second network element to execute the target task normally based on the second retrieval result.

[0130] In one or more embodiments, the first network element can update the knowledge base based on acquired external data. The external data can be data acquired by the first network element from the network. The description text of the target task may include the execution cycle of the target task. Based on the execution cycle of the target task and the description text, the first network element performs periodic searches in the knowledge base using a corresponding retrieval strategy to obtain the retrieval results corresponding to the target task. For example, in the embodiment shown in Figure 4, the execution cycle for obtaining the second retrieval result is the next execution cycle after obtaining the execution cycle of the first retrieval.

[0131] Understandably, the retrieval strategy corresponding to the current execution cycle can be the same as the retrieval strategy corresponding to the previous execution cycle. Alternatively, the retrieval strategy corresponding to the current execution cycle can be different from the retrieval strategy corresponding to the previous execution cycle. For example, the retrieval strategy corresponding to the current execution cycle can be the second retrieval strategy in the embodiment shown in Figure 4, and the retrieval strategy corresponding to the previous execution cycle can be the first retrieval strategy in the embodiment shown in Figure 4.

[0132] When the third search result obtained in the current execution cycle is inconsistent with the search result obtained in the previous execution cycle, the first network element sends the third search result to the second network element.

[0133] In other words, the data in the knowledge base can be continuously updated. The first network element can periodically search based on the execution cycle of the target task and the description text of the target task. When the retrieved data is inconsistent with the data retrieved in the previous execution cycle, the first network element can send the currently retrieved data to the second network element to ensure the timeliness of the search results output by the first network element, so that the second network element can obtain the latest data in a timely manner. Understandably, the search strategy used by the first network element for periodic searching based on the execution cycle of the target task can remain consistent, that is, the search strategy corresponding to the current execution cycle is the same as the search strategy of the previous execution cycle.

[0134] Furthermore, when the third search result is consistent with the search result obtained in the previous execution cycle, the first network element does not need to send the third search result to the second network element, thereby reducing the communication volume between the first network element and the second network element.

[0135] Optionally, before sending the third search result to the second network element, the first network element can verify the third search result based on the search result verification information corresponding to the second network element to ensure that the third search result meets the requirements of the second network element. The search result verification information can be stored in the first network element or sent from the second network element to the first network element.

[0136] For example, the following section, with reference to Figure 5, describes the periodic retrieval based on the execution cycle of the target task.

[0137] Figure 5 is a flowchart of an information retrieval method provided in an embodiment of this application. As shown in Figure 5, in this method, the common memory module can be used as the first network element in the embodiment shown in Figure 4, and the planning agent can be used as the second network element in the embodiment shown in Figure 4. In addition, the method also includes assembling agents, connecting agents, and executing agents. The target task is to subscribe to new information related to historical tasks. That is, periodically retrieving information related to a certain historical task. The description text of the target task includes the execution cycle of the target task. The common memory module can perform periodic retrieval according to the execution cycle of the target task. Furthermore, the common memory module can continuously acquire external data and update the knowledge base based on the external data. For example, the common memory module can continuously acquire external data related to the target task and use a top-down hierarchical clustering method to save the external data to update the knowledge base. The data related to the target task may include UE / RAN capabilities, status, availability, and location information, as well as document knowledge and available resources (such as network data throughput and network congestion information) related to historical tasks.

[0138] Referring to Figure 5, this embodiment may include the following steps:

[0139] S501: The planning agent sends a description text of the target task to the common memory module. This target task is used to request subscription to new information related to historical tasks. The description text of the target task may include keywords, time periods, and other information.

[0140] The planning agent can send the description text of the target task to the common memory module via the ABI interface. Correspondingly, the common memory module receives the description text of the target task via the ABI interface.

[0141] S502: The common memory module can periodically search the knowledge base according to the execution cycle of the target task, using the corresponding retrieval strategy, to obtain the retrieval results corresponding to the target task.

[0142] For example, the common memory module can use methods such as keyword matching and cosine similarity judgment to determine the data in the knowledge base that are relevant to the target task, and obtain the retrieval results corresponding to the target task.

[0143] Furthermore, after obtaining the retrieval result of the current execution cycle, the common memory module can compare the retrieval result with the retrieval result obtained in the previous execution cycle. If the retrieval result is inconsistent with the retrieval result obtained in the previous execution cycle, other steps are executed. If the retrieval result is consistent with the retrieval result obtained in the previous execution cycle, periodic checks are performed according to the execution cycle of the target task, i.e., step S503 is executed again.

[0144] S503: The common memory module sends the retrieval results corresponding to the target task to the planning agent. Correspondingly, the planning agent receives the retrieval results corresponding to the target task from the common memory module.

[0145] The common memory module can send the retrieval result through the ABI interface. The ABI interface determines the destination address of the retrieval result (i.e., the planning agent) and sends the retrieval result to the planning agent.

[0146] S504A: The planning agent can adjust the planning subtasks based on the search results.

[0147] S504B: The planning agent sends the subtask to the assembly agent. Correspondingly, the assembly agent receives the subtask from the planning agent.

[0148] S505A: The assembling agent updates the task topology based on the adjusted subtasks.

[0149] S505B: The assembling agent sends the task topology to the executing agent. Correspondingly, the executing agent receives the task topology from the assembling agent.

[0150] S506A: The executing agent sends the updated subtask configuration to the connected agent.

[0151] S506B: Connecting agents update network connection configurations for UE and RAN.

[0152] S507A: The executing agent is based on a new subtask-generating network instance.

[0153] S507B: Executes the agent to deploy the corresponding instance to the GN.

[0154] After the S508 GN deployment instance is completed, it sends the execution result information to the common memory module. Correspondingly, the common memory module receives the execution result information sent by the GN.

[0155] The information retrieval method shown in Figure 4 will be illustrated below with reference to an embodiment.

[0156] Figure 6 is a flowchart illustrating an information retrieval method provided in an embodiment of this application. As shown in Figure 6, the method includes an agent, a common memory module, and a network GPT. The agent can be used to execute a target task. The common memory module can be used to perform retrieval based on the description text of the target task. The network GPT can be used to analyze the feedback information from the agent and send the execution status of the target task to the common memory module.

[0157] Referring to Figure 6, this embodiment may include the following steps:

[0158] S601: The agent sends a description text of the target task to the common memory module. Correspondingly, the common memory module receives the description text of the target task from the agent. The description text of the target task is "Analyze the internet browsing behavior of users in the current community".

[0159] S602: The common memory module retrieves the first search result related to the target task from the knowledge vector base based on the description text of the target task. Alternatively, the common memory module can retrieve the first search result related to the target task from relevant expert knowledge and document information.

[0160] The first search result corresponding to the target task can be semi-structured data. This first search result may include results from historical tasks similar to the target task, links to relevant documents or web databases, etc. Specifically, results from historical tasks similar to the target task are in natural language, while links to relevant documents or web databases are in structured data.

[0161] For example, the first search result could be: "The historical network usage analysis of all users in the current community is as follows: 50% of users use the network for more than 8 hours a day, 30% of users use the network for 2-8 hours a day, and another 20% of users use the network for less than 2 hours a day. The main usage scenarios include social networking platforms, video platforms, etc.; Cell ID: 934; Information source: https: / / dataservice / v1." Here, "The historical network usage analysis of all users in the current community is as follows: 50% of users use the network for more than 8 hours a day, 30% of users use the network for 2-8 hours a day, and another 20% of users use the network for less than 2 hours a day. The main usage scenarios include social networking platforms, video platforms, etc." is unstructured data, while "Cell ID: 934" and "Information source: https: / / dataservice / v1" are structured data.

[0162] S603: The common memory module sends the first search result to the agent. Correspondingly, the agent receives the first search result from the common memory module.

[0163] S604: The agent can perform the target task based on the first search result and generate corresponding feedback information.

[0164] Upon receiving the first search result, the agent can check whether the data format of the first search result is correct and whether the webpage link is accessible. When the first search result meets the agent's requirements, the agent can determine the prompt information based on the first search result and the description text of the target task, and process the prompt using its internal AgentGPT to execute the corresponding orchestration and reasoning task.

[0165] Feedback information may include binary information indicating the execution status of the target task (e.g., 1 for normal, 0 for abnormal), and error descriptions of target task execution errors. Error descriptions may include: "The content of the first search result cannot be correctly parsed due to incorrect data format or missing necessary fields," "Insufficient agent permissions prevent access to resource links in the first search result," "Low relevance of retrieved information due to keyword mismatch or inconsistent data content," "Task failure due to exceeding memory or computing resource limits," "The first search result does not meet the input character restrictions," "Incorrect arrangement logic, incorrect information retrieval order or condition judgment," "Unable to generate correct query conditions or the target data source is unavailable," etc.

[0166] S605: The agent sends feedback information to the common memory module. Correspondingly, the common memory module receives the feedback information from the agent.

[0167] S606: When feedback indicates an error in the target task execution, the common memory module can send an error description to the network GPT. Correspondingly, the network GPT receives the error description from the common memory module.

[0168] S607, the network GPT can analyze this error description to obtain the reason for the abnormal execution of the target task.

[0169] S608: The reason for the abnormal execution of the target task sent by the network GPT to the common memory module.

[0170] S609: The common memory module can generate a retrieval strategy based on the cause of the target task execution exception, and use the retrieval strategy to perform a retrieval to obtain the second retrieval result corresponding to the target task.

[0171] The retrieval strategies include updating search keywords, search scope, and search database permissions. Furthermore, the shared memory module validates the search results to ensure they meet the agent's attribute requirements. For example, it can verify the correctness of the data format, the relevance of the search results to the target task, and the availability of the data source. Additionally, the shared memory module can compress the search results to meet the agent's input character limitations or reduce communication overhead.

[0172] S610: The common memory module sends the second search result to the agent. Correspondingly, the agent receives the second search result from the common memory module.

[0173] The following example, with reference to an embodiment, illustrates how to summarize multiple search data into a single search result. Referring to Figure 7, the agent sends a description text of the target task to the common memory module. The common memory module performs a search based on the description text, obtaining multiple search data. The common memory module can then summarize the multiple search data to obtain a search result and send the search result back to the agent.

[0174] The common memory module can extract keywords and key time information from the description text of the target task. For example, the description text could be "please help me find the statistics of abnormal behavior in Shanghai Century Park in 2024." Keywords could be "the statistics of abnormal behavior" or "Shanghai Century Park," and the key time information could be "2024."

[0175] After obtaining the corresponding search results based on the description text of the target task, the common memory module can also process the search results. For example, it can determine whether there are problems such as irrelevance, duplication, and contradiction in the search results, and summarize and extract keywords to solve the problems in the search results.

[0176] For example, the common memory module processes the search results by: deleting duplicate and redundant search data from the search results; identifying the correct data among contradictory search data based on data quality parameters such as valid time and number of similar results; determining the similarity between each search data and the descriptive text, and sorting each search data based on semantic similarity; determining the weight of each search data, and summarizing each search data based on its weight to generate the final search results.

[0177] For example, the public memory module obtained several search data points in its search results, such as: "Shanghai Century Park has a large number of visitors in summer and a relatively small number of visitors in winter," "A record of a dispute between a visitor and a security guard at the park's Gate 1 on May 2, 2024; a drunk man throwing trash into the lake in the park on August 5, 2024," and "Shanghai Century Park is very safe; no abnormal behavior was detected in 2024." The public memory module processed these search data points to obtain the following search result: "Shanghai Century Park operated normally overall last year, but there were still several sudden and serious incidents, including the dispute between a visitor and a security guard on May 2 and the drunk visitor throwing trash into the lake on August 5."

[0178] It is understood that, in order to achieve the functions in the above embodiments, the communication device includes hardware structures and / or software modules corresponding to each function. Those skilled in the art should readily recognize that, based on the units and method steps described in conjunction with the embodiments disclosed in this application, this application can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed by hardware or by computer software driving hardware depends on the specific application scenario and design constraints of the technical solution.

[0179] Figures 8 and 9 are schematic diagrams illustrating the possible structures of information retrieval devices provided in embodiments of this application. These information retrieval devices can be used to implement the functions of the first network element (or common memory module) or the second network element (or intelligent agent) in the above method embodiments, and thus can also achieve the beneficial effects of the above method embodiments. For example, the information retrieval device can be used to implement the functions of the first network element (or common memory module) or the second network element (or intelligent agent) in the processes shown in Figures 4 to 7.

[0180] The information retrieval device 800 shown in Figure 8 includes a processing unit 810 and a transceiver unit (or communication unit) 820. The information retrieval device 800 is used to implement the receiving and / or transmitting functions of the first network element or the second network element in the above method embodiments. The transceiver unit may include a transmitting unit and a receiving unit, used for transmitting and receiving, respectively.

[0181] Taking the process shown in Figure 4 as an example, when the information retrieval device 800 is used to implement the function of the first network element in the method embodiment shown in Figure 4, specifically, the transceiver unit 820 can be used to receive the description information and feedback information of the target task, and send the first search result and the second search result. The processing unit 810 can be used to perform a search based on the description information of the target task to obtain the first search result or the second search result, so as to improve the stability of the retrieval device in obtaining accurate search results.

[0182] In one possible implementation, the processing unit 810 can update the knowledge base based on acquired external data. The processing unit 810 performs periodic searches based on the description text and corresponding search strategies according to the execution cycle to obtain search results corresponding to the target task. When the third search result obtained in the current execution cycle is inconsistent with the search result obtained in the previous execution cycle, the transceiver unit 820 sends the third search result to the second network element.

[0183] In one possible implementation, the processing unit 810 may verify the second search result based on the search result verification information and pass the verification.

[0184] In one possible implementation, the processing unit 810 can process multiple search data and merge the multiple search data into a single search result; wherein, the processing of multiple search data includes deleting irrelevant search data, deduplication, identifying contradictory search data, and merging multiple search data.

[0185] For a more detailed description of the processing unit 810 and the transceiver unit 820, please refer directly to the description of the process steps and their related features in the above method embodiments, which will not be repeated here.

[0186] The information retrieval device 900 shown in Figure 9 includes a processor 910 and an interface circuit 920. The processor 910 and the interface circuit 920 are coupled to each other. It is understood that the interface circuit 920 can be a transceiver or an input / output interface. Optionally, the information retrieval device 900 may also include a memory 930 for storing instructions executed by the processor 910, or storing input data required by the processor 910 to execute instructions, or storing data generated after the processor 910 executes instructions.

[0187] When the information retrieval device 900 is used to implement the above method embodiment, the processor 910 is used to implement the function of the processing unit 810, and the interface circuit 920 is used to implement the function of the transceiver unit 820.

[0188] It is understood that the processor in the embodiments of this application can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), microprocessors without interlocked piped stages architecture (MIPS), advanced instruction set computers (RISC) machines (ARM), network processors (NPs), field-programmable gate arrays (FPGAs), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. A general-purpose processor can be a microprocessor or any conventional processor.

[0189] The method steps in the embodiments of this application can be implemented in hardware or by a processor executing software instructions. The software instructions can consist of corresponding software modules, which can be stored in random access memory, flash memory, read-only memory, programmable read-only memory, erasable programmable read-only memory, electrically erasable programmable read-only memory, registers, hard disks, portable hard disks, compact disc read-only memory (CD-ROM), or any other form of storage medium well known in the art. An exemplary storage medium is coupled to a processor, enabling the processor to read information from and write information to the storage medium. Of course, the storage medium can also be a component of the processor. The processor and storage medium can reside in an ASIC. Furthermore, the ASIC can reside in a first network device, a second network device, a third network device, or a fourth network device. Alternatively, the processor and storage medium can exist as discrete components in the first network device, the second network device, the third network device, or the fourth network device.

[0190] In the above embodiments, implementation can be achieved entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially in the form of a computer program product. The computer program product includes one or more computer programs or instructions. A computer program is a set of instructions that directs each step of an action of an electronic computer or other device with message processing capabilities. It is typically written in a programming language and runs on a target architecture. When the computer program or instructions are loaded and executed on a computer, the processes or functions described in the embodiments of this application are performed, in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer program or instructions can be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another. For example, the computer program or instructions can be transferred from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium, such as a floppy disk, hard disk, or magnetic tape; it can also be an optical medium, such as a digital video optical disc; or it can be a semiconductor medium, such as a solid-state drive. The computer-readable storage medium can be volatile or non-volatile, or it can include both types of storage media.

[0191] Based on the same technical concept, embodiments of this application also provide a computer-readable storage medium, including a program or instructions, which, when run on a computer, cause the methods in the above method embodiments to be executed.

[0192] Based on the same technical concept, embodiments of this application also provide a computer program product, including instructions that, when run on a computer, cause the methods in the above method embodiments to be executed.

[0193] Based on the same technical concept, this application also provides an information retrieval system to implement the information retrieval method shown in Figures 4 to 7.

[0194] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0195] In the various embodiments of this application, unless otherwise specified or in case of logical conflict, the terminology and / or descriptions of different embodiments are consistent and can be referenced by each other. The technical features of different embodiments can be combined to form new embodiments according to their inherent logical relationship.

[0196] This application can be presented in terms of aspects, embodiments, or features surrounding a system that may include multiple devices, components, modules, etc. It should be understood and appreciated that various systems may include additional devices, components, modules, etc., and / or may not include all devices, components, modules, etc. discussed in conjunction with the accompanying drawings. Furthermore, combinations of these solutions may also be used. Additionally, in the embodiments of this application, words such as "exemplarily," "for example," etc., are used to indicate examples, illustrations, or descriptions. Any embodiment or design described as an "example" in this application should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the term "example" is intended to present concepts in a concrete manner. In the embodiments of this application, "of," "corresponding / relevant," and "corresponding" may sometimes be used interchangeably, and it should be noted that their intended meanings are consistent unless their distinction is emphasized.

[0197] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0198] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to this application. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in one or more blocks of the flowchart illustrations and / or one or more blocks of the block diagrams.

[0199] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means that implement the functions specified in one or more flowcharts and / or one or more block diagrams.

[0200] These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, such that the instructions, which execute on the computer or other programmable apparatus, provide steps for implementing the functions specified in one or more flowcharts and / or one or more block diagrams.

[0201] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the scope of protection of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. An information retrieval method, characterized in that, include: The first network element receives a description text of a target task from the second network element; wherein the second network element is used to execute the target task, which is a task in a network communication service. The first network element performs a search in the knowledge base based on the description text and a first search strategy to obtain the first search result corresponding to the target task; The first network element sends the first search result to the second network element and receives feedback information from the second network element. The feedback information is used to instruct the second network element on the execution status of the target task based on the first search result. When the feedback information indicates that the target task is not executing properly, the first network element obtains the reason for the target task not executing properly and determines a second retrieval strategy based on the reason; wherein, the second retrieval strategy is different from the first retrieval strategy; The first network uses the second retrieval strategy to search the knowledge base based on the description text, and obtains the second retrieval result corresponding to the target task; The first network element sends the second search result to the second network element.

2. The method as described in claim 1, characterized in that, The method further includes: updating the knowledge base based on the acquired external data; The description text includes the execution cycle of the target task, and the method further includes: The first network element performs periodic searches in the knowledge base based on the execution cycle and the description text, using the corresponding search strategy, to obtain the search results corresponding to the target task. When the third search result obtained in the current execution cycle is inconsistent with the search result obtained in the previous execution cycle, the first network element sends the third search result to the second network element.

3. The method as described in claim 1 or 2, characterized in that, The feedback information includes search result verification information. Before the first network element sends the second search result to the second network element, the method further includes: The second search result is verified based on the verification information of the search result and the verification is successful.

4. The method as described in claim 3, characterized in that, The verification information for the search results includes at least one of the following: data format requirements, access restrictions, and input character restrictions for the second network element.

5. The method according to any one of claims 1-4, characterized in that, The first retrieval strategy and the second retrieval strategy include at least one of the following: keywords of the descriptive text, knowledge base, and retrieval scope.

6. The method according to any one of claims 1-5, characterized in that, The reasons for the abnormal execution of the target task include at least one of the following: inability to parse the search results, inability to access the links in the search results, the search results not matching the target task, the search results not meeting the input character restrictions, or inability to execute the target task normally.

7. The method according to any one of claims 1-6, characterized in that, The first search result and / or the second search result include multiple search data, and the method further includes: The first network element processes the multiple search data and merges them into a single search result; The methods for processing the multiple search data include deleting irrelevant search data, deduplication, identifying contradictory search data, and merging multiple search data.

8. An information retrieval device, characterized in that, include: The transceiver unit is used to receive a description text of a target task from a second network element; wherein the second network element is used to execute the target task, which is a task in a network communication service. The processing unit is configured to perform a search in the knowledge base based on the description text using a first search strategy to obtain a first search result corresponding to the target task; The transceiver unit is further configured to send the first search result to the second network element and receive feedback information from the second network element, wherein the feedback information is used to instruct the second network element on the execution status of the target task based on the first search result; When the feedback information indicates that the target task is not executing correctly, the processing unit is further configured to obtain the reason for the target task not executing correctly, and determine a second retrieval strategy based on the reason; and use the second retrieval strategy to search the knowledge base according to the description text to obtain a second retrieval result corresponding to the target task; wherein the second retrieval strategy is different from the first retrieval strategy; The transceiver unit is also used to send the second search result to the second network element.

9. The apparatus as claimed in claim 8, characterized in that, The processing unit is also used to update the knowledge base based on the acquired external data; The description text includes the execution cycle of the target task. The processing unit is further configured to perform periodic searches in the knowledge base based on the execution cycle and the description text, using a corresponding search strategy, to obtain search results corresponding to the target task. When the third search result obtained in the current execution cycle is inconsistent with the search result obtained in the previous execution cycle, the transceiver unit is also used to send the third search result to the second network element.

10. The apparatus as claimed in claim 8 or 9, characterized in that, The feedback information includes search result verification information. Before sending the second search result to the second network element, the processing unit is further configured to: The second search result is verified based on the verification information of the search result and the verification is successful.

11. The apparatus as claimed in claim 10, characterized in that, The verification information for the search results includes at least one of the following: data format requirements, access restrictions, and input character restrictions for the second network element.

12. The apparatus as claimed in any one of claims 8-11, characterized in that, The first retrieval strategy and the second retrieval strategy include at least one of the following: keywords of the descriptive text, knowledge base, and retrieval scope.

13. The apparatus as claimed in any one of claims 8-12, characterized in that, The reasons for the abnormal execution of the target task include at least one of the following: inability to parse the search results, inability to access the links in the search results, the search results not matching the target task, the search results not meeting the input character restrictions, or inability to execute the target task normally.

14. The apparatus according to any one of claims 8-13, characterized in that, The first search result and / or the second search result includes multiple search data. The processing unit is further configured to process the multiple search data and merge the multiple search data into one search result. The methods for processing the multiple search data include deleting irrelevant search data, deduplication, identifying contradictory search data, and merging multiple search data.

15. A computer-readable storage medium, characterized in that, The storage medium stores a computer program or instructions, and when the computer program or instructions are executed by the information retrieval device, the method as described in any one of claims 1-7 is implemented.

16. A computer program product, characterized in that, When the computer program product is executed by a computer, the computer executes the method as described in any one of claims 1-7.

17. An electronic device, characterized in that, The electronic device includes: Memory is used to store computer program instructions; A processor for executing the computer program instructions to support the electronic terminal device in implementing the method as described in any one of claims 1-7.

Images