Communication method, apparatus and system

Abstract

The present application relates to the technical field of communications. Disclosed are a communication method, apparatus and system. The method comprises: a first communication apparatus receiving first information from a second communication apparatus, the first information being used for triggering generation of synthetic data associated with a target object; generating, on the basis of the first information, the synthetic data associated with the target object; and sending the synthetic data to a third communication apparatus. In this way, a target object is introduced, such that synthetic data is associated with a target object, thereby facilitating generation of diverse synthetic data, and thus solving the problems such as data scarcity in the field of communications caused by factors such as privacy protection.

Description

A communication method, apparatus and system

[0001] Cross-reference to related applications

[0002] This application claims priority to Chinese Patent Application No. 202411988121.1, filed on December 27, 2024, entitled "A Communication Method, Apparatus and System", the entire contents of which are incorporated herein by reference. Technical Field

[0003] This application relates to the field of communication technology, and in particular to a communication method, apparatus and system. Background Technology

[0004] In today's rapidly evolving technological landscape, model training and fine-tuning in the field of artificial intelligence (AI) have been core drivers of technological progress. Real-world data has long played an irreplaceable and crucial role in this process. From early, simple image recognition models to today's complex natural language processing systems, vast amounts of real-world data (such as massive amounts of text data, rich image samples, and detailed user behavior records) provide a solid foundation for model growth, enabling them to learn patterns and rules from the real world and thus achieve accurate predictions and decisions.

[0005] Current data acquisition methods are shifting from traditional data collection to a combination of collection and data synthesis. This is because: firstly, acquiring real-world data is becoming increasingly difficult. High collection costs, time-consuming and laborious annotation processes, and reduced data availability due to strict privacy regulations all make it difficult to meet the growing demands of model training by relying solely on real-world data. Secondly, as model application scenarios continue to expand and diversify, real-world data often fails to fully cover certain specific, rare, yet extremely critical data scenarios. Synthetic data, on the other hand, can generate data for these special contexts through algorithms, thereby enhancing the model's generalization ability and adaptability to complex situations.

[0006] With the development of communication technology, AI models are gradually being applied to the field of communication. However, acquiring real-world data in the communication field also faces numerous challenges. Therefore, how to generate diverse synthetic data still requires further research. Summary of the Invention

[0007] This application provides a communication method, apparatus, and system for generating diverse synthetic data.

[0008] In a first aspect, embodiments of this application provide a communication method, which can be executed by a first communication device in a communication network. The "first communication device" in this application can refer to a first communication equipment, a component within the first communication equipment (e.g., a processor, chip, or chip system), or a logic module or software capable of implementing all or part of the functions of the first communication equipment. For example, in the method provided in the first aspect, the first communication device receives first information from a second communication device, the first information being used to trigger the generation of synthetic data associated with a target object; based on the first information, the synthetic data associated with the target object is generated; and the synthetic data is sent to a third communication device.

[0009] By using the above method, the synthetic data is associated with the target object by introducing the target object, which facilitates the generation of diverse synthetic data and solves the problem of data shortage caused by privacy protection and other reasons in the field of communications.

[0010] In one possible design, the first information includes description information of the synthesized data and description information of the target object, wherein the target object includes at least one of the following: at least one role associated with the communication network, at least one function of at least one network element in the communication network, and at least one service associated with the communication network.

[0011] Thus, since the target objects include roles / functions / businesses, that is, the target objects are multiple dimensions of the communication network, the generated synthetic data can cover multiple dimensions of the communication network, which makes it easier to ensure the diversity of the synthetic data.

[0012] In one possible design, the method further includes: sending description information of at least one object to the second communication device, the at least one object including at least one of the following: multiple roles in a role library, multiple functions in a function library, and multiple services in a service library; wherein the target object belongs to the at least one object.

[0013] Alternatively, the method may further include: sending to the second communication device at least one of the following: description information of multiple roles in a role library, description information of multiple functions in a function library, and description information of multiple services in a service library; wherein the at least one role belongs to the multiple roles, the at least one function belongs to the multiple functions, and the at least one service belongs to the multiple services.

[0014] In one possible design, the method further includes: generating at least one of the role library, the function library, and the service library based on at least one of the communication network private data, public information and communication technology (ICT) data, communication standard data, and third-party private data.

[0015] In one possible design, generating synthetic data associated with the target object based on the first information includes: inputting the first information into a preset model to obtain the synthetic data.

[0016] In one possible design, the method further includes: receiving second information from the second communication device, the second information indicating that the receiving end of the synthesized data is the third communication device; the step of sending the synthesized data to the third communication device includes: sending the synthesized data to the third communication device according to the second information.

[0017] In one possible design, the method further includes sending the quantity of the synthesized data to the third communication device.

[0018] In one possible design, the method further includes: receiving a quantity threshold of the synthesized data from the second communication device; the step of generating synthesized data based on the first information includes: generating the synthesized data based on the first information and the quantity threshold of the synthesized data, wherein the quantity of the synthesized data is greater than or equal to the quantity threshold of the synthesized data.

[0019] In one possible design, the method further includes: receiving a deduplication threshold for the synthesized data from the second communication device; generating the synthesized data based on the first information includes: generating the synthesized data based on the first information and the deduplication threshold; wherein the synthesized data is synthesized data obtained after deduplication processing based on the deduplication threshold.

[0020] Thus, by deduplicating the synthesized data, it is easier to ensure the quality of the synthesized data.

[0021] In one possible design, the method further includes: receiving indication information from the second communication device, the indication information being used to indicate that the synthetic data is used for model training or model fine-tuning; and sending the indication information to the third communication device.

[0022] Secondly, embodiments of this application provide a communication method, which can be executed by a second communication device in a communication network. The "second communication device" in this application can refer to a second communication equipment, a component within the second communication equipment (e.g., a processor, chip, or chip system), or a logic module or software capable of implementing all or part of the functions of the second communication equipment. For example, in the method provided in this second aspect, the second communication device obtains first information and sends the first information to a first communication device, the first information being used to trigger the generation of synthetic data associated with a target object.

[0023] In one possible design, the first information includes prompt information and description information of the target object, the target object including at least one of the following: at least one role associated with the communication network, at least one function of at least one network element in the communication network, and at least one service associated with the communication network.

[0024] In one possible design, obtaining the first information includes: generating the prompt information based on at least one of the following: the functional requirements of the network elements of the communication network; the service requirements of the communication network; the process requirements of the communication network; the strategy of the communication network; the computing requirements of the communication network; the perception requirements of the communication network; or the alarm information of the communication network.

[0025] In one possible design, the method further includes: receiving description information of at least one object, the at least one object including at least one of the following: multiple roles in a role library, multiple functions in a function library, and multiple services in a business library; obtaining the first information includes: determining the target object from the at least one object based on the similarity between the description information of the synthesized data and the description information of the at least one object.

[0026] Alternatively, the method may further include: receiving at least one of the following: description information of multiple roles, description information of multiple functions, and description information of multiple services; obtaining the first information includes at least one of the following: determining at least one role from the multiple roles based on the similarity between the description information of the synthesized data and the description information of the multiple roles; determining at least one function from the multiple functions based on the similarity between the description information of the synthesized data and the description information of the multiple functions; and determining at least one service from the multiple services based on the similarity between the description information of the synthesized data and the description information of the multiple services.

[0027] Thus, by determining at least one role / function / business based on the similarity between the descriptions of multiple roles / functions / businesses in the role database, the determined role / function / business is more reasonable, which helps to ensure the quantity of synthesized data.

[0028] In one possible design, the method further includes sending a second message to the first communication device, the second message indicating that the receiving end of the synthesized data is a third communication device.

[0029] In one possible design, the method further includes: sending a threshold number of the synthesized data to the first communication device.

[0030] In one possible design, the method further includes sending a deduplication threshold to the first communication device, the deduplication threshold being used for deduplication processing of the synthesized data.

[0031] In one possible design, the method further includes: sending indication information to the first communication device, the indication information being used to indicate that the synthetic data is used for model training or model fine-tuning.

[0032] Thirdly, embodiments of this application provide a communication method, which can be executed by a fourth communication device in a communication network. The "fourth communication device" in this application can refer to a fourth communication equipment, a component within the fourth communication equipment (e.g., a processor, chip, or chip system), or a logic module or software capable of implementing all or part of the functions of the fourth communication equipment. For example, in the method provided in the third aspect, the fourth communication device obtains network data and generates at least one of a role library, a function library, and a service library based on the network data; wherein the network data includes at least one of the following: private data of the communication network, public information and communication technology (ICT) data, communication standard data, and private data of a third party.

[0033] In one possible design, the method further includes sending at least one of the role library, function library, and service library to the first communication device.

[0034] In one possible design, the fourth communication device is the same as the first communication device.

[0035] Fourthly, this application provides a communication device that has the functions involved in any of the first to third aspects described above. For example, the communication device includes modules, units, or means corresponding to the operations involved in any of the first to third aspects described above. The functions, units, or means can be implemented by software, or by hardware, or by hardware executing corresponding software.

[0036] In one possible design, the communication device includes a processing unit and a communication unit, wherein the communication unit can be used to transmit and receive signals to enable communication between the communication device and other devices; the processing unit can be used to perform some internal operations of the communication device. The functions performed by the processing unit and the communication unit can correspond to the operations involved in any of the first to third aspects described above.

[0037] In one possible design, the communication device includes a processor that can be coupled to a memory. The memory can store necessary computer programs or instructions for implementing the functions involved in any of the first to third aspects described above. The processor can execute the computer programs or instructions stored in the memory, causing the communication device to implement the methods in any possible design or implementation of the first to third aspects described above when the computer programs or instructions are executed.

[0038] In one possible design, the communication device includes a processor and a memory, the memory of which may store necessary computer programs or instructions for implementing the functions involved in any of the first to third aspects described above. The processor may execute the computer programs or instructions stored in the memory, and when the computer programs or instructions are executed, cause the communication device to implement the methods in any possible design or implementation of the first to third aspects described above.

[0039] In one possible design, the communication device includes a processor and an interface circuit, wherein the processor is configured to communicate with other devices via the interface circuit and execute the methods in any possible design or implementation of the first to third aspects described above.

[0040] Understandably, in the fourth aspect mentioned above, the processor can be implemented in hardware or software. When implemented in hardware, the processor can be a logic circuit, integrated circuit, etc.; when implemented in software, the processor can be a general-purpose processor that reads software code stored in memory. Furthermore, there can be one or more processors, and one or more memories. The memory can be integrated with the processor, or the memory and processor can be separate. In specific implementations, the memory can be integrated with the processor on the same chip, or it can be set on different chips. This application does not limit the type of memory or the arrangement of the memory and processor.

[0041] Fifthly, this application provides a communication system that may include a first communication device and a second communication device; wherein the first communication device is used to perform the method described in the first aspect, and the second communication device is used to perform the method described in the second aspect. Optionally, the communication system may further include a third communication device for receiving synthesized data.

[0042] In a sixth aspect, this application provides a computer-readable storage medium storing a computer program (or computer-readable instructions) in which, when a computer reads and executes some or all of the computer-readable instructions, the method in any of the possible designs of the first to third aspects described above is executed.

[0043] For example, a computer-readable storage medium can be any available medium that a computer can access. This includes, but is not limited to, non-transient computer-readable media, random-access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), CD-ROM or other optical disc storage, magnetic disk storage media, or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer.

[0044] In a seventh aspect, this application provides a computer program product that, when read and executed by a computer, causes the method in any of the possible designs of the first to third aspects to be performed.

[0045] Eighthly, this application provides a chip (or chip system) including a processor coupled to a memory storing a computer program; the processor is configured to invoke part or all of the computer program in the memory, such that the method in any of the possible designs of the first to third aspects described above is executed. Attached Figure Description

[0046] Figure 1 is a schematic diagram of a communication network architecture;

[0047] Figure 2 is a schematic diagram of a data plane architecture for a future communication network;

[0048] Figure 3A is a schematic diagram of a network architecture applicable to an embodiment of this application;

[0049] Figure 3B is a schematic diagram of a federated learning architecture provided in an embodiment of this application;

[0050] Figure 4 is a flowchart illustrating the communication method provided in the embodiments of this application;

[0051] Figure 5 is a schematic diagram of synthetic data provided in the embodiments of this application for model training or fine-tuning;

[0052] Figure 6 is an exemplary block diagram of the apparatus involved in the embodiments of this application;

[0053] Figure 7 is a schematic diagram of the structure of a communication device provided in an embodiment of this application. Detailed Implementation

[0054] The technical solutions in the embodiments of this application will now be described with reference to the accompanying drawings. This application will focus on various aspects, embodiments, or features of a system that may include multiple devices, components, modules, etc. It should be understood and appreciated that each system may include additional devices, components, modules, etc., and / or may not include all the devices, components, modules, etc. discussed in conjunction with the accompanying drawings. Furthermore, combinations of these solutions may also be used.

[0055] First, the relevant technical terms involved in the embodiments of this application will be explained. Unless otherwise specified, these explanations are for the purpose of making the embodiments of this application easier to understand, and should not be regarded as a strict limitation on the scope of protection claimed in this application.

[0056] I. Architecture of Communication Networks

[0057] Figure 1 is a schematic diagram of a communication network architecture, which can be a 5th generation (5G) network architecture. This network architecture comprises four components: terminal equipment, access network (AN), core network (CN), and data network (DN). The access network can be a radio access network (RAN). Terminal equipment, access network, and core network are the main components of the above network architecture. Logically, they can be divided into user plane and control plane. The control plane is responsible for the management of the mobile network, while the user plane is responsible for the transmission of service data.

[0058] (1) Terminal equipment

[0059] A terminal device is a device that provides voice and / or data connectivity to a user. Terminal devices may also be referred to as user equipment (UE), terminal, access terminal, terminal unit, terminal station, mobile station (MS), remote station, remote terminal, mobile terminal (MT), wireless communication equipment, terminal agent, or terminal equipment, etc.

[0060] For example, the terminal device can be a handheld device with wireless connectivity, or a vehicle with communication capabilities, such as in-vehicle equipment (e.g., in-vehicle communication device, in-vehicle communication chip). Examples of current terminal devices include: mobile phones, cordless phones, session initiation protocol (SIP) phones, wireless local loop (WLL) stations, personal digital assistant (PDA) devices, handheld devices with wireless communication capabilities, computing devices or other processing devices connected to a wireless modem, tablet computers, computers with wireless transceiver capabilities, laptops, handheld computers, mobile internet devices (MIDs), wearable devices, virtual reality (VR) devices, augmented reality (AR) devices, wireless terminals in industrial control, wireless terminals in self-driving vehicles, wireless terminals in remote medical surgery, wireless terminals in smart grids, wireless terminals in transportation safety, wireless terminals in smart cities, and wireless terminals in smart homes.

[0061] Terminal devices can be deployed on land, including indoors or outdoors, handheld, wearable, or vehicle-mounted; they can also be deployed on water (such as ships); and they can also be deployed in the air (such as airplanes, balloons, and satellites). This application does not limit the specific technologies, device forms, application scenarios, or names used in the terminal devices.

[0062] (2) Access Network

[0063] The access network is deployed close to the terminal equipment, providing network access functionality for authorized users in a specific area. It can determine different quality transmission tunnels to transmit user data based on user level, service requirements, and other factors. The access network manages and utilizes its own resources efficiently, providing access services to terminal equipment on demand, and is responsible for forwarding control signals and service data between the terminal equipment and the core network.

[0064] The access network can be the access network in the 3rd generation partnership project (3GPP). The RAN can also be an open RAN (O-RAN or ORAN), a cloud radio access network (CRAN), or a communication network of two or more of the above.

[0065] Access network equipment is deployed in the access network to connect terminal devices to the wireless network. Access network equipment is typically connected to the core network via a wired link (e.g., fiber optic cable). Access network equipment can also be called access network devices or RAN equipment / nodes. Access network equipment can be a base station, an evolved NodeB (eNodeB), a transmission reception point (TRP), a next-generation NodeB (gNB) in 5G communication networks, or a base station in future communication networks.

[0066] Access network equipment 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), a distributed unit (DU), or a radio unit (RU). The CU performs the functions of the radio resource control (RRC) protocol and PDCP of the base station, and can also perform the functions of the service data adaptation protocol (SDAP). The CU can be further divided into a CU control plane (CP) (i.e., CU-CP) and a CU user plane (UP) (i.e., CU-UP). The DU performs the functions of the RLC and MA layers of the base station, and can also perform some or all of the physical layer functions. For specific descriptions of the above protocol layers, please refer to the relevant 3GPP technical specifications. The CU and DU can be set up separately, or they can be included in the same network element, such as in the baseband unit (BBU). The RU can be included in radio frequency equipment or radio frequency units, such as in a remote radio unit (RRU), an active antenna unit (AAU), or a remote radio head (RRH). In different systems, CU, DU, or RU may have different names, but those skilled in the art will understand their meaning. For example, in an ORAN system, CU can also be called O-CU (open CU), DU can also be called O-DU, and RU can also be called O-RU. Any of the units among CU (or CU-CP, CU-UP), DU, and RU in this application can be implemented through software modules, hardware modules, or a combination of software and hardware modules. RA equipment can be a macro base station, a micro base station, an indoor station, a relay node, or a donor node, etc. The embodiments of this application do not limit the specific technology or equipment form used in the access network equipment.

[0067] (3) Core Network

[0068] The core network is responsible for maintaining the subscription data of the mobile network, managing the network elements of the mobile network, and providing terminal devices with functions such as session management, mobility management, policy management, and security authentication.

[0069] The core network user plane includes user plane function (UPF) network elements. The core network control plane includes, but is not limited to: access and mobility management function (AMF) network elements, session management function (SMF) network elements, authentication server function (AUSF) network elements, network exposure function (NEF) network elements, network function repository function (NRF) network elements, policy control function (PCF) network elements, unified data management (UDM) network elements, and application function (AF) network elements.

[0070] UPF network elements are primarily responsible for connecting to external networks and executing user data packet forwarding according to the routing rules of SMF network elements. For example, uplink data is sent to the data network or other UPF network elements, and downlink data is sent to other UPF network elements or access network devices.

[0071] AMF network elements are mainly responsible for the access management and mobility management of terminal devices, such as the status maintenance of terminal devices, the reachability management of terminal devices, the forwarding of mobility management non-access stratum (MM NAS) messages, and the forwarding of session management (SM) N2 messages.

[0072] SMF (Service Provider Function) network elements are primarily responsible for session management in mobile networks, including establishing sessions for terminal devices, allocating and releasing resources for sessions, such as session quality of service (QoS), session paths, and forwarding rules. For example, they may allocate Internet Protocol (IP) addresses to terminal devices and select UPF (User Provider Function) network elements that provide packet forwarding functions.

[0073] The AUSF network element is primarily responsible for performing security authentication of terminal devices.

[0074] NEF network elements are used to connect other internal network elements of the core network with external devices (such as application servers) of the core network to provide network capability information to external devices, or to provide information from external devices to core network elements.

[0075] The NRF network element is primarily responsible for providing other network elements with the functions of storing and selecting network function entity information.

[0076] The PCF network element is mainly responsible for user policy management, including policy authorization, quality of service and generation of billing rules, and distributing the corresponding rules to the UPF network element through the SMF network element to complete the installation of the corresponding policies and rules.

[0077] UDM network elements are primarily responsible for data management and control. For example, UDM network elements can manage user subscription information, including acquiring subscription information and providing it to other network elements (such as AMF network elements); generating 3GPP authentication credentials for terminal devices; and registering and maintaining the network elements currently serving the terminal devices.

[0078] The AF (Area Function) network element is mainly responsible for providing various application service data to the control plane network elements of the operator's communication network, or obtaining network data and control information from the control plane network elements of the communication network.

[0079] Although not shown, the core network may include other possible network elements, such as network data analysis function (NWDAF) network elements, without any specific limitation.

[0080] (4) Data Network

[0081] A data network, also known as a packet data network (PDN), is a network located outside of the operator's network. An operator's network can connect to multiple data networks. Data networks can be private networks, such as local area networks (LANs), external networks not controlled by the operator, such as the Internet, or dedicated networks jointly deployed by operators; the specific type is not limited.

[0082] It is understood that the following description will use the names of network elements in a 5G communication network as an example, and the embodiments of this application do not limit the names of network elements. Figure 1 illustrates the core network control plane using a service-oriented architecture as an example. In the service-oriented architecture, each control plane network element is connected to a service bus, and the interaction between control plane network elements adopts the service call method, that is, the control plane network element will open services to other control plane network elements for other control plane network elements to call. In other possible implementations, the core network control plane can also adopt a point-to-point communication method. In point-to-point communication, there will be a specific set of messages for the communication interface between control plane network elements. Of course, in future communication networks, the names of these interfaces can remain unchanged, or they can be replaced with other names, and this application does not limit this. In future communication networks, the above-mentioned network elements or devices can still use their names in the 5G communication network, or have other names; the functions of the above-mentioned network elements or devices can be completed by an independent network element, or can be completed by several network elements together, and the embodiments of this application do not limit this.

[0083] The network elements / functional entities in the various possible network architectures described above 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). Optionally, the aforementioned network elements or functional entities can be implemented by a single device, multiple devices working together, or different functional modules within a single device; this application does not specifically limit these possibilities.

[0084] II. Data plane of future communication networks

[0085] Based on the 5G network architecture shown in Figure 1, the current 5G user plane is used to carry session data and cannot meet the requirements of "in-the-path computing" and "arbitrary topology." Therefore, the 5G user plane cannot carry data for future communication networks. To systematically address the challenges of data services and solve the problem that the existing 5G user plane cannot carry data for future communication networks, an independent data plane has been introduced for future communication networks. Figure 2 is a schematic diagram of the data plane architecture of future communication networks. As shown in Figure 2, this data plane architecture includes: a data orchestrator (DO) (and a data controller (DC)), a data agent (DA), and a data storage function (DSF). Optionally, it also includes a data communication proxy (DCP) and / or a data processing function (DPF).

[0086] (1) DO: Supports data service request translation, transforming data service requests into the construction of data bearers, and orchestrating programmable data pipelines to provide the data service. For example, the DO obtains global information about the DA logical network based on the data service capabilities reported by the DA and the logical connection status between DAs; then, the DO selects a suitable DA based on the received data service request, orchestrates the data pipeline, and calculates and constructs the data forwarding path to form the data bearer. The DO sends data forwarding information to the DA through the data forwarding control protocol (DFCP) and updates and deletes data forwarding information as needed.

[0087] DO can be a newly added network element in the core network. For example, the newly added DO can communicate with other network elements in the core network through a service interface, or it can communicate with other network elements in the core network through point-to-point communication. Alternatively, DO can be built into an existing network element (such as an AMF network element or an SMF network element). That is, DO can be an SMF network element or an AMF network element with added data orchestration function.

[0088] In other examples, based on the real-time requirements and cross-domain nature of the data service tasks, the data orchestrator can be subdivided into DO and DC. DO is responsible for coarse-grained, non-real-time data orchestration, while DC is responsible for fine-grained, real-time data orchestration. In this application, "DO" will be used as the example for description.

[0089] (2) DA: Performs data services such as data acquisition, data processing, data storage, data analysis, and data sharing as assigned by the orchestration. For example, a DA can implement a variety of data processing functions, which are reported to the DO as capabilities of the DA during the DA registration period, and capability updates can be reported in a timely manner.

[0090] DA can be built into existing devices (such as terminal devices, access network devices) or network elements (such as AMF network elements, SMF network elements), or it can be deployed independently, without any specific limitations.

[0091] (3) DSF: A storage extension component for DA when large-scale data storage or long-term storage is required. DSF can be built into existing devices or network elements, or it can be deployed independently, without any specific limitation.

[0092] In addition to DSF, Figure 2 may also include other data storage functions, such as the analysis data repository function (ADRF) and / or the sensing data repository function (SDRF), without any specific limitations.

[0093] (4) DCP: Provides an efficient data transmission mechanism that supports multiple transmission protocols, such as Transmission Control Protocol (TCP), User Datagram Protocol (UDP), Quick UDP Internet Connection (QUIC), or other transmission protocols. Data producers can send data to the DCP, and data consumers can subscribe to and pull data from the DCP. The DCP can be built into existing devices or network elements, or it can be deployed independently; there are no specific limitations.

[0094] (5) DPF: A special type of DA that performs data analysis and processing functions. DPF can be built into existing devices or network elements, or it can be deployed independently, with no specific limitation.

[0095] III. Large Language Model

[0096] Large language models (LLMs) are artificial intelligence models trained on large datasets that can perform a wide range of tasks. When the parameters of an LLM exceed a certain scale, it will produce unexpected capabilities, such as language understanding, intent understanding, and multi-turn dialogue memory.

[0097] LLM has a wide range of applications. For example, it can be used for natural language generation, such as generating various types of text, including essay writing, story creation, and poetry writing. For instance, it can create a beautiful prose piece based on a given theme like "the beauty of autumn," depicting the scenery and atmosphere of autumn. Another example is its use in intelligent question-answering systems. As the core of such systems, it can answer various user questions, from common-sense questions (such as "What is the highest mountain in the world?") to specialized questions (such as "How to understand the Schrödinger equation in quantum mechanics?"), providing answers based on its learned knowledge. Furthermore, LLM can be used in machine translation, assisting in translation between different languages. It accurately translates text from one language into another, taking into account context, culture, and other factors during the translation process to make the translation more natural and fluent.

[0098] IV. Composite Data

[0099] Synthetic data is data generated using algorithms, rather than data directly obtained from actual observations, measurements, or records. It can simulate the characteristics and distribution of real data and can be used for various purposes, such as supplementing datasets, testing systems, or protecting privacy. For example, in the image domain, images can be synthesized using generative adversarial networks (GANs). A GAN consists of a generator and a discriminator. The generator attempts to produce realistic images, while the discriminator determines whether an image is real or generated. Through continuous training, the generator can produce synthetic images with similar features to real images, such as synthesized face images or landscape images.

[0100] There are several methods for generating synthetic data. For example, method 1 uses statistical models: this method generates data using statistical distributions. For instance, if the temperature data for a certain region is known to follow a normal distribution, synthetic temperature data conforming to that distribution can be generated by estimating the parameters of the normal distribution (mean and standard deviation). Method 2 uses machine learning models: besides generative adversarial networks (GANs) used for image synthesis, autoregressive language models (such as LLMs) can also be used to synthesize text data in natural language processing. By learning from large amounts of text, these models can generate synthetic data based on given prompts.

[0101] Synthetic data has various applications. For example, it can be used for data augmentation; specifically, synthetic datasets can serve as pre-training datasets, i.e., large-scale datasets used to pre-train models in machine learning, especially deep learning. This pre-training process is typically performed before the model is fine-tuned for a specific task (such as text classification or image recognition), aiming to allow the model to learn general feature representations, thus enabling faster convergence and improved performance on the specific task during subsequent fine-tuning stages. Alternatively, synthetic datasets can also serve as instruction datasets, containing a series of steps and parameters used in machine learning and deep learning to guide the fine-tuning of pre-trained models on specific tasks or datasets. Fine-tuning is a crucial process for adapting general models to specific application domains; through these instruction sets, the model can better adapt and optimize its performance on specific tasks.

[0102] Based on the above-mentioned terminology, this application will study the relevant implementation of generating synthetic data.

[0103] The Economist published an article titled "AI firms will soon exhaust most of the internet's data," citing predictions from Epoch AI research firm that all high-quality text data on the internet will be used up by 2028, while machine learning datasets may exhaust all "high-quality language data" by 2026. The future demand for more high-quality text will be exponential, far exceeding the rate of data generation (approximately 1% to 7% per year), leading to a near depletion of high-quality public data. Therefore, current data acquisition methods are shifting from traditional data collection to a combination of collection and data synthesis.

[0104] One approach involves generating synthetic data (such as instruction data) using a self-instruction method. This instruction dataset is used for supervised fine-tuning (SFT) and reinforcement learning from human feedback (RLHF). The 1.3B model (with approximately 1.3 billion parameters) after SFT and RLHF outperforms the un-fine-tuned 175B model (with approximately 175 billion parameters), demonstrating the effectiveness of using the synthetic instruction dataset for fine-tuning.

[0105] Self-Instruction is a method for generating synthetic data, particularly important in Natural Language Processing (NLP). Its core idea is to allow the model to generate instructions and corresponding responses, thereby expanding the training data. The specific process of Self-Instruction is as follows: First, some instructions are randomly selected from 175 seed instructions (manually written instructions / problems accumulated from the business side). Then, the model (such as an LLM) refers to these instructions to generate a series of similar instructions. The model then determines whether the instruction is a "classification" problem or a "generation" problem, and adopts different generation methods accordingly. This distinction is based on experimental results: for generalized classification tasks, the model tends to generate input content with the same label as the output, resulting in poor data diversity and balance. For classification tasks, the model first generates possible outputs, and then combines the instructions and outputs to generate the input; for non-classification tasks, the input and output are generated sequentially. If an instruction does not require input, the input is empty. After the above steps, a batch of synthetic data is initially obtained. This data needs further filtering to improve quality. Then, deduplication is performed based on algorithms, keyword filtering, and text length filtering. The filtered synthetic data can then be added to the "seed instruction," and this cycle continues, generating more instructions in a continuous stream.

[0106] Using the above method, the model can generate the required synthetic data simply by specifying data synthesis prompts. While the amount of synthetic data can be easily increased, diversity is difficult to ensure: typically, a single data synthesis prompt can only generate one instance. To create large-scale, diverse synthetic data, such as 1 billion instances, a large number of diverse prompts are needed. Therefore, self-instructions and similar methods cannot achieve scalable synthetic data creation. Furthermore, continuously generating a series of similar instructions from seed instructions can lead to the homogenization of synthetic data.

[0107] With the development of communication technology, AI models are gradually being applied to the communication field. However, acquiring real-world data in the communication field also faces numerous challenges. Data is the foundation for training communication network models. Communication networks involve complex scenarios and require diverse data, which may originate from different devices (such as terminal devices, access network devices, or network management / cloud management systems). Acquiring real-world communication network data is particularly difficult. For example, base station-level data is proprietary to vendors and lacks standardization. In the operations and maintenance field, although the network generates massive amounts of data daily (such as call history reports (CHR) data), this data is highly homogenized, consisting of results from identical parameter templates (with differences less than 1‰). Using the Self-Instruction method to generate synthetic data in the communication field would compromise data diversity.

[0108] Based on this, embodiments of this application provide a communication method, apparatus, and system for generating diverse synthetic data.

[0109] First, the network architecture to which the embodiments of this application are adapted will be introduced. The technical solutions in the embodiments of this application can be applied to various communication networks, such as wireless local area networks (WLAN), wireless fidelity (Wi-Fi) systems, 4th generation (4G) networks (such as long term evolution (LTE) networks), 5G networks (such as new radio (NR) systems), future communication networks, or other similar communication networks, without limitation.

[0110] Figure 3A is a schematic diagram of a network architecture applicable to an embodiment of this application. As shown in Figure 3A, the network architecture may include a first communication device and a second communication device, optionally including a third communication device, and optionally including a fourth communication device (not shown). The second communication device sends first information to the first communication device, and then the first communication device generates composite data based on the first information and sends the composite data to the third communication device. The first information can be found below. The second and third communication devices may be the same device or different devices. The fourth communication device is used to generate at least one of a role library, a function library, and a business library. The descriptions of the role library, function library, and business library can be found below. The fourth communication device and the first communication device may be the same device or different devices.

[0111] For example, the first communication device, the second communication device, the third communication device and the fourth communication device are all communication equipment or components of communication equipment in a communication network. The specific communication equipment / network elements included in the communication network can be referred to the description in Figure 1 or Figure 2.

[0112] For example, the first communication device is a first communication equipment or a component of the first communication equipment, such as a chip or chip system disposed in the first communication equipment; the second communication device is a second communication equipment or a component of the second communication equipment, such as a chip or chip system disposed in the second communication equipment; the third communication device is a third communication equipment or a component of the third communication equipment, such as a chip or chip system disposed in the third communication equipment; and the fourth communication device is a fourth communication equipment or a component of the fourth communication equipment, such as a chip or chip system disposed in the fourth communication equipment.

[0113] For example, the second communication device is a DO / DC, the first communication device is a DSF network element, the third communication device is a terminal device, and the fourth communication device is the same as the first communication device. For instance, the third communication device might be a client device in a federated learning (FL) architecture. Federated learning, while fully protecting user data privacy and security, efficiently completes the model learning task by enabling collaboration between various edge devices and the central server. The FL architecture shown in Figure 3B is currently the most widely used training architecture in the FL field. The FedAvg algorithm is the foundational algorithm of FL, and its general algorithm flow is as follows:

[0114] (1) The central server initializes the model to be trained. It then broadcasts this information to all client devices. The central server can be a base station or a mobile edge computing (MEC) network element.

[0115] (2) In the t∈[1,T] round, the client device k∈[1,K] is based on the local dataset. For the received global model Perform E epochs of training to obtain the local training results. This is then reported to the central server. In the example shown in Figure 3B, the local training results sent by client devices n, k, and m are denoted as G, respectively. n G k G m .

[0116] (3) The central server aggregates and collects the local training results from all (or some) of the client devices. Assume the set of clients uploading local models in round t is... The central server will use the number of samples from the corresponding client devices as weights to calculate the new global model. The specific update rule is as follows: The central server then sends the latest version of the global model. The broadcast is sent to all client devices for a new round of training.

[0117] (4) Repeat steps (2) and (3) until the model finally converges or the number of training rounds reaches the upper limit.

[0118] As can be seen in the FL framework, distributed nodes (such as client device k) perform local training based on the received synthetic data and report the local training results to the central server. The central server itself does not have a dataset; it is only responsible for fusing the training results from the distributed nodes to obtain a global model, which is then distributed to the distributed nodes.

[0119] It is understood that the network architecture shown in Figure 3A may also include other possible devices / network elements, and no specific limitations are imposed. The network architecture shown in Figure 3A is only one possible example. The network architecture and business scenarios described in the embodiments of this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided in the embodiments of this application. As those skilled in the art will know, with the evolution of communication network architecture and the emergence of new business scenarios, the technical solutions provided in the embodiments of this application are also applicable to similar technical problems.

[0120] It is understood that in the embodiments of this application, "send" and "receive" indicate the direction of signal transmission. For example, "send information to XX" can be understood as the destination of the information being XX, and "send information" can include direct sending or indirect sending through other communication devices, communication apparatuses, units, or modules. "Receive information from YY" can be understood as the source of the information being YY, and "receive information" can include receiving directly from YY or receiving indirectly from YY through other communication devices, communication apparatuses, units, or modules. In addition, "send" can also be understood as the "output" of the chip interface, and "receive" can also be understood as the "input" of the chip interface. In other words, "send" or "receive" can be performed between devices, such as between access network devices and terminal devices through an air interface, or "send" or "receive" can be performed within a device, such as between components, modules, chips, software modules, or hardware modules within a device through a bus, wiring, or interface.

[0121] Based on the network architecture shown in Figure 3A, the communication method provided by this application is described below with reference to specific embodiments. The communication method provided by this application involves interaction between multiple communication devices, such as a first communication device, a second communication device, and a third communication device. The communication method provided by this application involves a role library, a function library, and a business library; the relevant content of the role library, function library, and business library will be introduced here first.

[0122] (1) An introduction to the implementation of generating role libraries, function libraries and business libraries.

[0123] This application does not limit the executing entity for generating the role library, function library, and service library. For example, the executing entity can be a fourth communication device, such as an NWDAF network element or a DSF network element. Exemplarily, the role library, function library, and service library can be generated based on network data. For example, network data includes at least one of the following: private data of the communication network, public information and communication technology (ICT) data, communication standard data, and private data of a third party.

[0124] Among these, communication network-private data can also be understood as operator-private data. For example, communication network-private data may include at least one of the following: minimized drive test (MDT) data, measurement report (MR) data, call history (CHR) data, traffic statistics, billing information, user subscription data, functional data of one or more network elements in the communication network, and may also include other possible information, which will not be listed one by one. For example, MDT data, MR data, CHR data, traffic statistics, and billing information can be collectively referred to as dynamic data, while user subscription data can be referred to as static data.

[0125] Public ICT data can include data from ICT-related papers, white papers, standards, blogs, and other sources.

[0126] Communication standard data can include standard protocol data in the field of communications, such as 3GPP standard protocol data.

[0127] In the telecommunications field, a third party can be any entity other than a telecommunications service provider (such as an operator) or a network equipment manufacturer (such as a base station or core network equipment). For example, a third party can be a terminal manufacturer or an application (APP) manufacturer; the third party's private data may include documentation from the terminal manufacturer, such as specifications, designs, tests, and applications, as well as documentation from the application (APP) manufacturer.

[0128] For example, roles, functions, and business operations can be generated based on MDT data. Specifically, a portion of data can be selected from the MDT data (it could be an entire article, a paragraph, a sentence, etc.), and the selected data and prompt information 1 can be input into a preset model (such as LLM). For example, prompt information 1 could be "This text is most likely to be used for doing / reading / writing / testing...what?" Correspondingly, LLM will output the corresponding functions, such as identifying network coverage blind spots and optimizing switching parameters.

[0129] Alternatively, the selected data and prompt information 2 can be input into a preset model (such as LLM). For example, prompt information 2 could be "Who is most concerned about / involved in / related to this text...". Accordingly, LLM will output the corresponding roles, such as telecommunications engineer, telecommunications university teacher, etc.

[0130] Alternatively, the selected data and prompt information 3 can be input into a preset model (such as LLM). For example, prompt information 3 could be "This text is most likely to appear / be related to / involve / ...which business"; accordingly, LLM will output the corresponding business, such as switching business, live streaming upload business, etc.

[0131] Because dynamic data (such as MDT data) is generated dynamically and is dynamically related to location, air interface channel information, network status, users, etc., it is time-varying and can therefore guarantee the diversity of generated roles / functions / services.

[0132] For another example, roles, functions, and services can be generated based on user subscription data. Operators have numerous subscribed users, potentially in the millions or even hundreds of millions. The data on user subscriptions is stored in the operator's UDR / UDM, including identity information, service subscription information, and QoS information. However, due to privacy protection requirements, this data needs to be de-identified; that is, information such as identity information is anonymized to generate de-identified information. Specifically, a portion of the user subscription data can be selected (it could be an entire text, a paragraph, a sentence, etc.). This selected data, along with prompt message 1, is input into a preset model (such as LLM). For example, prompt message 1 could be "This text is most likely used for / reading / writing / testing…what?" Correspondingly, the LLM will output the corresponding functions, such as network access control and QoS management.

[0133] Alternatively, the selected data and prompt information 2 can be input into a preset model (such as LLM). For example, prompt information 2 could be "Who is most likely to comply with / regulate / use this text...". Accordingly, LLM will output the corresponding roles, such as user roles after privacy is removed, IoT / vehicle terminals, telecommunications regulatory departments, operators, etc.

[0134] Alternatively, the selected data and prompt information 3 can be input into a preset model (such as LLM). For example, prompt information 3 could be "This text is most likely to appear / be related to / involve / …which business"; accordingly, LLM will output the corresponding business.

[0135] To give another example, roles, functions, and services can be generated based on the functional data of one or more network elements in a communication network. A communication network contains numerous network elements, such as base stations, UPF network elements, AMF network elements, and SMF network elements. A portion of the data (which could be an entire document, a paragraph, or a sentence) can be selected from the functional data of these network elements. This selected data, along with prompts, is then input into a preset model (such as an LLM). Correspondingly, the LLM will output the corresponding role, function, or service.

[0136] It is understandable that after generating multiple roles, more roles can be derived based on the relationships between them (i.e., Persona-to-Persona). Similarly, after generating multiple functions, more functions can be derived based on the relationships between them (i.e., Function-to-Function). For example, registration management is related to authentication and authorization; corresponding authentication and authorization functions can be derived from registration management. Furthermore, after generating multiple services, more services can be derived based on the relationships between them (i.e., Service-to-Service). Optionally, taking roles as an example (functions or services can refer to this), after generating multiple roles, deduplication can be performed on them. There are various specific implementations for deduplication, and this application embodiment does not limit this.

[0137] (2) An introduction to the storage of the role library, function library and business library.

[0138] This application does not limit the storage entity for the role library, function library, and service library. For example, the storage entity and the execution entity can be the same device / network element or different devices / network elements. When the storage entity and the execution entity are different devices / network elements, the execution entity can send the generated role library, function library, and service library to the storage entity; there is no specific limitation. For example, the role library, function library, and service library can be stored in a DSF network element, an ADRF network element, or other possible network elements; there is no specific limitation.

[0139] The roles in the role library are those associated with the communication network, such as communication engineers, telecommunications regulatory departments, privacy-free user roles, and IoT / vehicle terminals. The functions in the function library are those of network elements in the communication network, such as identifying network coverage blind spots and optimizing switching parameters. The services in the service library are those associated with the communication network, such as live streaming services, multimedia services, voice services, and video services.

[0140] Roles in the role library can be stored by category, such as roles related to network infrastructure construction and maintenance (e.g., telecommunications engineers, network operations engineers), terminal roles (e.g., mobile phones, IoT terminals, vehicle terminals, live streaming users, enterprise users, etc.), and network operation and service roles (e.g., operators, business support system engineers). Further, optionally, taking "telecommunications engineer" as an example, "telecommunications engineer" can also be a role category, specifically including roles such as "telecommunications engineer a" and "telecommunications engineer b".

[0141] Functions in the function library can be stored according to network elements or categories. Taking storage by network element as an example, functions of AMF network elements (such as mobility management) and functions of SMF network elements (such as session management) can be stored.

[0142] The services in the service database can be stored by category, such as information transmission and interaction services (e.g., live streaming and voice services), terminal management services (e.g., call charge services and mobile payment services), and network operation and maintenance services (e.g., network equipment monitoring services, troubleshooting services, and performance optimization services).

[0143] It is understood that the above-described category-based storage is only one possible example. In specific implementations, it can be a nested category form. For example, taking roles as an example, category a includes category a1 and category a2, category a1 includes category a11 and category a12, and category a11 specifically includes roles a111 and role a112. This application does not limit the specific storage form.

[0144] Figure 4 is a flowchart illustrating a communication method provided in an embodiment of this application. As shown in Figure 4, the process may include:

[0145] S401, the second communication device receives the first information.

[0146] For example, the second communication device is a DO / DC or a component of a DO / DC, and will be described below as "the second communication device is a DO / DC".

[0147] For example, the DO / DC obtaining the first information can be replaced by the DO / DC determining the first information or the DO / DC generating the first information. There are various scenarios that trigger the DO / DC to obtain the first information, such as the DO / DC obtaining the first information based on a request from a third communication device (e.g., a request for synthesized data).

[0148] (1) Introduce the first piece of information.

[0149] The first piece of information is used to trigger the generation of composite data associated with the target object. For example, the first piece of information includes descriptive information about the composite data and descriptive information about the target object.

[0150] The first piece of information can also be called a prompt message or prompt word, or other possible descriptions. In this embodiment, the example of "generating synthetic data through a model" is used. The model can be a statistical model, a machine learning model such as LLM, or other possible models, without specific limitations. For ease of description, the model that generates the synthetic data is referred to as Model A in this embodiment.

[0151] The first piece of information can be understood as the prompt for model A, providing a guiding input to instruct model A to generate the corresponding results (i.e., synthetic data). The descriptive information of the synthetic data describes what kind of synthetic data is desired; for example, it describes at least one of the semantic, logical, and format requirements of the synthetic data. The descriptive information of the target object describes the target object associated with the synthetic data.

[0152] There are several types of Prompt modes, such as zero-shot hint mode and few-shot hint mode. Zero-shot hint mode directly tells the model what kind of synthetic data is needed (i.e., the descriptive information of the input synthetic data), such as "Analyze the scenarios or parameters that cause QoS degradation" or "Analyze the propagation characteristics of millimeter-wave bands in 5G networks, including path loss and penetration loss." Few-shot hint mode provides the model with a small number of examples (usually a very limited number, such as a few) as descriptive information for the synthetic data. These examples guide the model to generate subsequent data in a similar manner. For example, when asking the model to generate mathematical problem-solving steps, first give it two or three correct solutions to simple math problems as examples, allowing the model to understand the problem-solving approach and format, and then let it solve new math problems.

[0153] Since the first information in this embodiment also includes the description information of the target object, that is, the description information of the target object is further added on the basis of the zero-sample prompt mode and the small-sample prompt mode, the Prompt mode corresponding to the first information in this embodiment can be understood as the zero-sample prompt mode or the small-sample prompt mode based on the target object.

[0154] It is understandable that in other examples, the Prompt mode may also include other prompt modes, without any specific limitations.

[0155] (2) The specific implementation of the description information of the DO / DC to obtain the synthetic data is introduced.

[0156] For example, the DO / DC can generate descriptive information for the synthetic data based on at least one of the following: ① functional requirements of the network elements of the communication network; ② service requirements of the communication network; ③ process requirements of the communication network; ④ policies of the communication network, such as QoS policies, traffic policies, billing policies, security policies, etc.; ⑤ computational requirements of the communication network; ⑥ perception requirements of the communication network; ⑦ alarm information of the communication network. At least one of ① to ⑦ can be sent to the DO / DC by other devices or network elements. For example, the policies of the communication network are usually stored in the PCF network element, which can actively send the policies of the communication network to the DO / DC; or, the PCF network element can send the policies of the communication network to the DO / DC based on a request or query from the DO / DC.

[0157] For example, alarm information in communication networks includes: bandwidth-related alarms, such as "Link bandwidth utilization consistently exceeds 90%." Excessive link bandwidth utilization leads to data transmission congestion, increasing transmission latency and causing QoS degradation; packet loss alarms, such as "Network interface has accumulated 1000 packet losses in the past 10 minutes." Significant packet loss severely impacts data integrity and accuracy, thus degrading QoS; and latency alarms, such as "Average transmission latency for a specific service flow exceeds 500 milliseconds." High latency leads to QoS degradation. Therefore, based on these alarms, the DO / DC can generate descriptive information for the synthetic data, which can be described as "analyzing the scenarios or parameters causing QoS degradation."

[0158] To give another example, the functional requirements of a network element in a communication network are: In a communication network, a large number of mobile terminals need to conduct wireless communication in various complex environments to achieve signal transmission and reception. Therefore, based on this functional requirement, the description information of the synthetic data generated by DO / DC can be: "Construct a base station network element that can stably transmit and receive wireless signals with a large number of mobile terminals in complex environments."

[0159] For another example, the computational requirements of network elements in a communication network include encryption and decryption requirements. Therefore, based on this functional requirement, the description information of the synthetic data generated by DO / DC can be "Please design a computational task involving the encryption and decryption process of data transmission in a communication network".

[0160] (3) The specific implementation of DO / DC to obtain the target object is introduced.

[0161] The target object includes at least one of the following: at least one role associated with the communication network, at least one function of at least one network element in the communication network, and at least one service associated with the communication network. Here, "associated" can be replaced with "related" or "involved". It should be understood that, taking "at least one role" as an example (at least one function or at least one service can be understood similarly), the role in "at least one role" can refer to a type of role in the role library (e.g., a communication engineer), or it can refer to a specific role in the role library (e.g., communication engineer a, communication engineer b, communication engineer c), without limitation.

[0162] For example, the DO / DC can obtain the target object based on the descriptive information of the synthetic data. For instance, a DSF network element can send descriptive information of at least one object to the DO / DC, and correspondingly, the DO / DC receives the descriptive information of at least one object, whereby the at least one object includes at least one of the following: multiple roles in a role library, multiple functions in a function library, and multiple services in a service library. Furthermore, the DO / DC determines the target object from the at least one object based on the similarity between the descriptive information of the synthetic data and the descriptive information of the at least one object.

[0163] Specifically, regarding "obtaining at least one role": DSF network elements can send descriptions of multiple roles in the role library to DO / DC. Each of these multiple roles can refer to a type of role (such as a communications engineer or a network maintenance engineer). In other words, DSF network elements can send descriptions of role categories in the role library to DO / DC without sending descriptions of specific roles (such as communications engineer a, communications engineer b, and communications engineer c) to save transmission overhead. It should be understood that each of the multiple roles can also refer to a specific role, meaning that DSF network elements can also send specific roles from the role library to DO / DC.

[0164] Accordingly, the DO / DC determines (or selects) at least one role from multiple roles based on the similarity between the description information of the synthetic data and the description information of the multiple roles. For example, if the multiple roles include roles 1 to 10, the DO / DC can select at least one role with a similarity greater than or equal to a threshold based on the similarity between the description information of the synthetic data and the description information of the 10 roles; or, if the number of at least one role is 5, the DO / DC can select the 5 roles with the highest similarity based on the similarity between the description information of the synthetic data and the description information of the 10 roles.

[0165] Here, similarity can refer to cosine similarity, without any specific limitation. Thus, by determining at least one character based on the similarity between the descriptive information of multiple characters in the character database, the determined at least one character becomes more reasonable, ensuring a sufficient quantity of synthesized data.

[0166] Regarding "obtaining at least one function": DSF network elements can send descriptions of multiple functions from the function library to DO / DC. Accordingly, DO / DC determines (or selects) at least one function from the multiple functions based on the similarity between the descriptions of the synthesized data and the descriptions of the multiple functions. For a detailed description, please refer to the above section on "obtaining at least one role".

[0167] Regarding "obtaining at least one service": DSF network elements can send description information of multiple services from the service database to DO / DC. Accordingly, DO / DC determines (or selects) at least one service from the multiple services based on the similarity between the description information of the composite data and the description information of the multiple services. For a detailed description, please refer to the above content on "obtaining at least one role".

[0168] S402, the second communication device (such as DO / DC) sends the first information to the first communication device; correspondingly, the first communication device receives the first information.

[0169] For example, the first communication device is a DSF network element or a component of a DSF network element. The following description will use "the first communication device is a DSF network element" as an example.

[0170] For example, the DO / DC can send the first information to the DSF network element through the interface between the DO / DC and the DSF network element; or, the DO / DC can also send the first information to the DSF network element through other network elements, without any specific limitation.

[0171] S403, the first communication device (such as a DSF network element) generates composite data based on the first information.

[0172] (1) The implementation of DSF network element to generate synthetic data is described.

[0173] For example, the DSF network element inputs the first information into a preset model (i.e., model A) to obtain synthetic data; wherein, model A may be pre-configured in the DSF network element.

[0174] If the first information mentioned above includes description information of the synthesized data and description information of the target object, for example, the description information of the synthesized data is "analyze the scenarios or parameters that cause QoS degradation", and the target object includes at least one role (such as a live broadcast user), at least one function (such as an access management function), and at least one service (such as a switching service). After the DSF network element inputs the first information into model A, model A can generate at least one piece of synthesized data based on "analyze the scenarios or parameters that cause QoS degradation from the perspective of the live broadcast user".

[0175] Understandably, assuming that a live user is a type of role in a role library, and the specific roles for live users in the role library include live user a, live user b, and live user c, then the DSF network element can directly input the description information of the live user contained in the first information into model A, so that model A generates at least one composite data. Alternatively, the DSF network element can also obtain the specific roles in the "live user" category, such as live user a, live user b, and live user c, from the role library based on the description information of the live user contained in the first information, and input the description information of live user a, live user b, and live user c into model A. Then, model A can generate at least one composite data from the perspective of live user a, at least one composite data from the perspective of live user b, and at least one composite data from the perspective of live user c. Specifically, when the DSF network element is the main storage entity of the role database, the DSF network element can directly obtain the specific role in the "live broadcast user" role category from the role database; when other devices / network elements are the main storage entities of the role database, the DSF network element can obtain the specific role in the "live broadcast user" role category through interaction with other devices / network elements, without any specific limitation.

[0176] Furthermore, Model A can generate at least one synthetic data point based on "analyzing the scenarios or parameters that cause QoS degradation from the perspective of access management functions." For details, please refer to the description of roles (i.e., live streaming users) above.

[0177] Furthermore, Model A can generate at least one synthetic data point based on "analyzing the scenarios or parameters that cause QoS degradation from the perspective of switching services." For details, please refer to the description of roles (i.e., live streaming users) above.

[0178] Furthermore, Model A can generate at least one synthetic data based on "analyzing the scenarios or parameters that cause QoS degradation from the perspective of combining live streaming users and access management functions".

[0179] Additionally, Model A can generate at least one synthetic data based on "analyzing the scenarios or parameters that cause QoS degradation from the perspective of combining live streaming users, access management functions, and switching services"; these will not be listed one by one.

[0180] (2) Describe the other information sent by DO / DC to DSF network elements.

[0181] Optionally, in addition to the first information mentioned above, the DO / DC may also send other possible information to the DSF network element.

[0182] For example, the DO / DC can send a threshold value for the amount of synthetic data to the DSF network element; correspondingly, the DSF network element generates synthetic data based on the first information and the threshold value, wherein the amount of synthetic data generated is greater than or equal to the threshold value. Optionally, if the DO / DC does not send the threshold value for the amount of synthetic data to the DSF network element, the DSF network element can generate synthetic data according to the default value.

[0183] For example, the DO / DC can send a deduplication threshold for the synthesized data to the DSF network element; correspondingly, the DSF network element generates synthesized data based on the first information and the deduplication threshold. The generated synthesized data is obtained after deduplication processing according to the deduplication threshold. There are various specific implementations for deduplication processing, such as using MinHash or embedding-based deduplication, i.e., using a text embedding model like OpenAI's text-embedding-3-small to calculate an embedding for each synthesized data, and then filtering out synthesized data with a cosine semantic similarity greater than or equal to the deduplication threshold.

[0184] For example, the DO / DC can send at least one of the following to the DSF network element: at least one role's limiting conditions, at least one function's limiting conditions, and at least one service's limiting conditions. Taking a role as an example (functions or services can be referenced here), to generate more granular synthetic data, after obtaining at least one role, the DO / DC can further obtain at least one role's limiting conditions (such as limiting the role's location, behavior, habits, etc.) and send them to the DSF network element; correspondingly, the DSF network element can generate synthetic data based on the prompt information, the limiting conditions of the prompt information, and at least one of the limiting conditions of at least one role, at least one function, and at least one service. It is understood that, taking roles as an example, the limiting conditions for different roles can be the same or different, and there is no specific limitation.

[0185] Taking the constraint of at least one role as an example (the constraint of at least one function or at least one business can be referred to), following the example above, assuming that the target object includes at least one role as a live user, and that a live user is a type of role in the role library, then the DSF network element can obtain the specific role that meets the constraint from the role library of the "live user" type of role based on the description information and constraint of the live user. The subsequent implementation can refer to the description above.

[0186] For example, DO / DC can send a second message to the DSF network element, which is used to indicate that the receiver of the synthesized data is a third communication device.

[0187] It is understood that some or all of the above information (such as the first information, the threshold for the number of synthesized data, the deduplication threshold, the second information, the limiting conditions for at least one role, the limiting conditions for at least one function, the limiting conditions for at least one business, etc.) can be carried in the same message, or they can be carried in different messages, without any specific limitation.

[0188] S404, the first communication device (such as a DSF network element) sends composite data to the third communication device.

[0189] For example, the third communication device is a terminal device or a component of a terminal device. In the following description, "the third communication device is a terminal device" will be used as an example.

[0190] For example, the DSF network element sends composite data to the terminal device based on the received second information.

[0191] Optionally, the DSF network element can also send other possible information to the terminal device.

[0192] For example, the DSF network element can also send the amount of composite data to the terminal device.

[0193] For example, the DSF network element can also send instruction information to the terminal device, which indicates whether the synthetic data is used for model training or model fine-tuning. This instruction information can be sent from the DO / DC to the DSF network element, meaning the DO / DC manages the specific purpose of the synthetic data.

[0194] For example, DSF network elements can also send model training or fine-tuning parameters (such as learning rate, batch size, loss function, etc.) to terminal devices. These parameters can be sent to the DSF network element by DO / DC or other network elements.

[0195] This application does not limit the application scenarios of the synthetic data. For example, the terminal device mentioned above can be a client device for federated fine-tuning. The terminal device can perform local fine-tuning of the model (such as model B) based on the synthetic data.

[0196] Using the above method, a role library, a function library, and a service library are generated based on network data. This ensures that when synthetic data needs to be generated, the target object determined by the DO / DC can include at least one of at least one role, at least one function, and at least one service. In other words, the target object covers multiple dimensions of the communication network. Furthermore, the synthetic data generated by the DSF network element based on the description information of the synthetic data and the description information of the target object covers multiple dimensions of the communication network. Compared with the Self-Instruction method for generating synthetic data, this method can improve the diversity of synthetic data.

[0197] As mentioned above, synthetic data can be used for model training or model fine-tuning. Therefore, a large-scale communication industry model can be generated based on the synthetic data in the embodiments of this application. As shown in Figure 5, assuming that a large-scale communication industry model needs to be generated, synthetic data 1 (for model training) and synthetic data 2 (for model fine-tuning) can be generated through model A. Synthetic data 1 (along with some real data, such as real data from the communication industry) is used to train model B (model B can be the initial large-scale communication industry model). After training is completed, synthetic data 2 is used to fine-tune the trained model B, and finally, a larger-scale communication industry model with better performance is obtained.

[0198] Regarding the above embodiments, it is understood that:

[0199] (1) The above focuses on describing the differences between different processes. In the various processes of this application, unless otherwise specified or logically conflicting, the terminology and / or descriptions of different processes are consistent and can be referenced from each other. In addition, different implementations or different examples can also be referenced from each other.

[0200] (2) The various numerical designations used in this application are merely for descriptive convenience and are not intended to limit the scope of this application. The step numbers in the above flowcharts are only examples of the execution process and do not constitute a restriction on the order of execution of the steps. That is, the size of each step number does not imply the order of execution; the execution order of each step should be determined by its function and internal logic. Furthermore, not all steps shown in the flowcharts are mandatory steps; some steps may be added or deleted based on actual needs.

[0201] The above primarily describes the solutions provided in the embodiments of this application from the perspective of device / network element interaction. It is understood that, to achieve the above functions, the device / network element may include corresponding hardware structures and / or software modules for executing each function. Those skilled in the art should readily recognize that, in conjunction with the units and algorithm steps of the various examples described in the embodiments disclosed herein, the embodiments of this application can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed in hardware or by computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0202] This application embodiment can divide the device / network element into functional units according to the above method example. For example, each function can be divided into a separate functional unit, or two or more functions can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0203] In the case of using integrated units, FIG6 shows a possible exemplary block diagram of the device involved in the embodiments of this application. As shown in FIG6, the device 600 may include a processing unit 602 and a communication unit 603. The processing unit 602 is used to control and manage the operation of the device 600. The communication unit 603 is used to support communication between the device 600 and other devices. Optionally, the communication unit 603 is also called a transceiver unit, and may include a receiving unit and / or a sending unit, respectively used to perform receiving and sending operations. The device 600 may also include a storage unit 601 for storing the program code and / or data of the device 600.

[0204] (1) The device 600 can be the first communication device in the above embodiments. The processing unit 602 can support the device 600 in performing the actions of the first communication device in the above method embodiments. Alternatively, the processing unit 602 mainly performs the internal actions of the first communication device in the method embodiments, and the communication unit 603 can support communication between the device 600 and other devices.

[0205] For example, in one embodiment, the communication unit 603 is configured to: receive first information from a second communication device, the first information being used to trigger the generation of synthetic data associated with a target object; the processing unit 602 is configured to: generate the synthetic data associated with the target object based on the first information; and the communication unit 603 is further configured to: send the synthetic data to a third communication device.

[0206] In one possible design, the first information includes description information of the synthesized data and description information of the target object, wherein the target object includes at least one of the following: at least one role associated with the communication network, at least one function of at least one network element in the communication network, and at least one service associated with the communication network.

[0207] In one possible design, the processing unit 602 is further configured to: send description information of at least one object to the second communication device, the at least one object including at least one of the following: multiple roles in a role library, multiple functions in a function library, and multiple services in a service library; wherein the target object belongs to the at least one object.

[0208] In one possible design, the processing unit 602 is further configured to: generate at least one of the role library, the function library, and the service library based on at least one of the communication network private data, public information and communication technology (ICT) data, communication standard data, and third-party private data.

[0209] In one possible design, the processing unit 602 is specifically used to: input the first information into a preset model to obtain the synthesized data.

[0210] In one possible design, the communication unit 603 is further configured to: receive second information from the second communication device, the second information indicating that the receiving end of the synthesized data is the third communication device; the step of sending the synthesized data to the third communication device includes: sending the synthesized data to the third communication device according to the second information.

[0211] In one possible design, the communication unit 603 is also used to: send the quantity of the synthesized data to the third communication device.

[0212] In one possible design, the communication unit 603 is further configured to: receive a quantity threshold of the synthesized data from the second communication device; the processing unit 602 is specifically configured to: generate the synthesized data according to the first information and the quantity threshold of the synthesized data, wherein the quantity of the synthesized data is greater than or equal to the quantity threshold of the synthesized data.

[0213] In one possible design, the communication unit 603 is further configured to: receive a deduplication threshold for the synthesized data from the second communication device; the processing unit 602 is specifically configured to: generate the synthesized data based on the first information and the deduplication threshold; wherein the synthesized data is synthesized data obtained after deduplication processing based on the deduplication threshold.

[0214] In one possible design, the communication unit 603 is further configured to: receive instruction information from the second communication device, the instruction information being used to instruct the synthetic data to be used for model training or model fine-tuning; and send the instruction information to the third communication device.

[0215] (2) The device 600 can be the second communication device in the above embodiments. The processing unit 602 can support the device 600 in performing the actions of the second communication device in the above method embodiments. Alternatively, the processing unit 602 mainly performs the internal actions of the second communication device in the method embodiments, and the communication unit 603 can support communication between the device 600 and other devices.

[0216] For example, in one embodiment, the processing unit 602 is used to: obtain first information; the communication unit 603 is used to: send the first information to a first communication device, wherein the first information is used to trigger the generation of synthetic data associated with the target object.

[0217] In one possible design, the first information includes prompt information and description information of the target object, the target object including at least one of the following: at least one role associated with the communication network, at least one function of at least one network element in the communication network, and at least one service associated with the communication network.

[0218] In one possible design, the processing unit 602 is specifically configured to: generate the prompt information according to at least one of the following: the functional requirements of the network elements of the communication network; the service requirements of the communication network; the process requirements of the communication network; the strategy of the communication network; the computing requirements of the communication network; the perception requirements of the communication network; or the alarm information of the communication network.

[0219] In one possible design, the communication unit 603 is further configured to: receive description information of at least one object, the at least one object including at least one of the following: multiple roles in a role library, multiple functions in a function library, and multiple services in a service library; the processing unit 602 is specifically configured to: determine the target object from the at least one object based on the similarity between the description information of the synthesized data and the description information of the at least one object.

[0220] In one possible design, the communication unit 603 is further configured to: send second information to the first communication device, the second information being used to indicate that the receiving end of the synthesized data is a third communication device.

[0221] In one possible design, the communication unit 603 is also used to: send a threshold number of the synthesized data to the first communication device.

[0222] In one possible design, the communication unit 603 is further configured to: send a deduplication threshold to the first communication device, the deduplication threshold being used for deduplication processing of the synthesized data.

[0223] In one possible design, the communication unit 603 is further configured to: send indication information to the first communication device, the indication information being used to indicate that the synthetic data is used for model training or model fine-tuning.

[0224] It should be understood that the division of units in the above device is merely a logical functional division. In actual implementation, they can be fully or partially integrated into a single physical entity, or they can be physically separated. Furthermore, all units in the device can be implemented entirely through software calls from processing elements; all units can be implemented entirely in hardware; or some units can be implemented through software calls from processing elements, while others are implemented in hardware. For example, each unit can be a separate processing element, or it can be integrated into a chip within the device. Alternatively, it can be stored as a program in memory, called and executed by a processing element of the device. Moreover, these units can be fully or partially integrated together, or implemented independently. The processing element here can also be called a processor, which can be an integrated circuit with signal processing capabilities. In the implementation process, the operations or units described above can be implemented through integrated logic circuits in the processor element or through software calls from processing elements.

[0225] In one example, a unit in any of the above devices can be one or more integrated circuits configured to implement the above methods, such as one or more application-specific integrated circuits (ASICs), or one or more digital signal processors (DSPs), or one or more field-programmable gate arrays (FPGAs), or a combination of at least two of these integrated circuit forms. As another example, when a unit in the device can be implemented in the form of a processing element scheduler, the processing element can be a processor, such as a central processing unit (CPU), or other processor capable of calling programs. Furthermore, these units can be integrated together and implemented as a System-on-a-Chip (SoC).

[0226] The receiving unit described above is an interface circuit of the device, used to receive signals from other devices. For example, when the device is implemented as a chip, the receiving unit is an interface circuit for the chip to receive signals from other chips or devices. The transmitting unit described above is an interface circuit of the device, used to transmit signals to other devices. For example, when the device is implemented as a chip, the transmitting unit is an interface circuit for the chip to transmit signals to other chips or devices.

[0227] Based on the above embodiments, this application also provides a communication device. Referring to FIG7, the communication device 700 may include a processor 701. Optionally, the communication device 700 may further include a memory 702, which may be disposed inside or outside the communication device 700. It is understood that FIG7 only shows the main components of the communication device, and the communication device may further include a transceiver (not shown in the figure).

[0228] Specifically, processor 701 can be a CPU, a network processor (NP), or a combination of a CPU and an NP. Processor 701 may further include a hardware chip. The aforementioned hardware chip can be an ASIC, a programmable logic device (PLD), or a combination thereof. The aforementioned PLD can be a complex programmable logic device (CPLD), an FPGA, generic array logic (GAL), or any combination thereof.

[0229] The processor 701 and memory 702 are interconnected. Optionally, the processor 701 and memory 702 are interconnected via bus 703; bus 703 can be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of illustration, only one thick line is used in Figure 7, but this does not indicate that there is only one bus or one type of bus.

[0230] In one alternative implementation, memory 702 is used to store programs, etc. Specifically, the program may include program code, which includes computer operation instructions. Memory 702 may include RAM, and may also include non-volatile memory, such as one or more disk storage devices. Processor 701 executes the application program stored in memory 702 to implement the above-mentioned functions, thereby realizing the functions of communication device 700.

[0231] For example, the communication device 700 may be the first communication device, the second communication device, or the third communication device in the above embodiments.

[0232] In one embodiment, when the communication device 700 implements the functions of the first communication device in the above method embodiment, the transceiver can perform the transmit and receive operations executed by the first communication device in the above method embodiment; the processor 701 can perform other operations besides the transmit and receive operations executed by the first communication device in the above method embodiment. Specific details can be found in the relevant descriptions in the above embodiments, and will not be elaborated upon here.

[0233] In one embodiment, when the communication device 700 implements the functions of the second communication device in the above method embodiment, the transceiver can perform the transmit and receive operations executed by the second communication device in the above method embodiment; the processor 701 can perform other operations besides the transmit and receive operations executed by the second communication device in the above method embodiment. Specific details can be found in the relevant descriptions in the above embodiments, and will not be elaborated upon here.

[0234] The terms "system" and "network" in this application embodiment are used interchangeably. "At least one" refers to one or more, and "multiple" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B 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, and C" includes A, B, C, AB, AC, BC, or ABC. And, unless otherwise specified, the ordinal numbers such as "first" and "second" mentioned in this application embodiment are used to distinguish multiple objects and are not used to limit the order, sequence, priority, or importance of multiple objects.

[0235] In the embodiments of this application, words such as "exemplarily" and "for example" are used to indicate examples, illustrations, or descriptions. Any embodiment or design scheme described as an "example" in this application should not be construed as being more preferred or advantageous than other embodiments or design schemes. 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.

[0236] 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.

[0237] 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.

[0238] 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.

[0239] 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.

Claims

1. A communication method, characterized in that, The method is applied to a first communication device in a communication network, and the method includes: Receive first information from the second communication device, the first information being used to trigger the generation of synthetic data associated with the target object; Based on the first information, generate synthetic data associated with the target object; The synthesized data is sent to a third communication device.

2. The method according to claim 1, characterized in that, The first information includes description information of the synthesized data and description information of the target object. The target object includes at least one of the following: at least one role associated with the communication network, at least one function of at least one network element in the communication network, and at least one service associated with the communication network.

3. The method according to claim 2, characterized in that, The method further includes: Send description information of at least one object to the second communication device, wherein the at least one object includes at least one of the following: multiple roles in a role library, multiple functions in a function library, and multiple services in a service library; The target object belongs to the at least one object.

4. The method according to claim 3, characterized in that, The method further includes: At least one of the role library, the function library, and the business library is generated based on at least one of the following: private data from the communication network, public information and communication technology (ICT) data, communication standard data, and private data from a third party.

5. The method according to any one of claims 1 to 4, characterized in that, The step of generating synthetic data associated with the target object based on the first information includes: The first information is input into a preset model to obtain the synthesized data.

6. The method according to any one of claims 1 to 5, characterized in that, The method further includes: Receive second information from the second communication device, the second information being used to indicate that the receiving end of the synthesized data is the third communication device; Sending the synthesized data to the third communication device includes: Based on the second information, the synthesized data is sent to the third communication device.

7. The method according to any one of claims 1 to 6, characterized in that, The method further includes: The amount of synthesized data sent to the third communication device.

8. The method according to any one of claims 1 to 7, characterized in that, The method further includes: A threshold for the number of synthesized data received from the second communication device; The step of generating synthetic data based on the first information includes: The synthetic data is generated based on the first information and the quantity threshold of the synthetic data, wherein the quantity of the synthetic data is greater than or equal to the quantity threshold of the synthetic data.

9. The method according to any one of claims 1 to 8, characterized in that, The method further includes: The deduplication threshold for receiving the synthesized data from the second communication device; The step of generating synthetic data based on the first information includes: The synthesized data is generated based on the first information and the deduplication threshold; The synthesized data is the synthesized data obtained after deduplication processing according to the deduplication threshold.

10. The method according to any one of claims 1 to 9, characterized in that, The method further includes: Receive instruction information from the second communication device, the instruction information being used to instruct the synthetic data to be used for model training or model fine-tuning; The instruction information is sent to the third communication device.

11. A communication method, characterized in that, The method is applied to a second communication device in a communication network, and the method includes: Obtain first information; The first information is sent to the first communication device, and the first information is used to trigger the generation of synthetic data associated with the target object.

12. The method according to claim 11, characterized in that, The first information includes prompt information and description information of the target object, wherein the target object includes at least one of the following: at least one role associated with the communication network, at least one function of at least one network element in the communication network, and at least one service associated with the communication network.

13. The method according to claim 12, characterized in that, The acquisition of the first information includes: The prompt message is generated based on at least one of the following: Functional requirements of the network elements of the communication network; The service requirements of the communication network; The process requirements of the communication network; The strategy of the communication network; The computational requirements of the communication network; The sensing requirements of the communication network; or Alarm information from the communication network.

14. The method according to claim 12 or 13, characterized in that, The method further includes: Receive description information of at least one object, wherein the at least one object includes at least one of the following: multiple roles in a role library, multiple functions in a function library, and multiple services in a business library; The acquisition of the first information includes: The target object is determined from the at least one object based on the similarity between the description information of the synthesized data and the description information of the at least one object.

15. The method according to any one of claims 11 to 14, characterized in that, The method further includes: Send a second message to the first communication device, the second message indicating that the receiving end of the synthesized data is a third communication device.

16. The method according to any one of claims 11 to 15, characterized in that, The method further includes: A threshold number of synthesized data to be sent to the first communication device.

17. The method according to any one of claims 11 to 16, characterized in that, The method further includes: A deduplication threshold is sent to the first communication device, and the deduplication threshold is used for deduplication processing of the synthesized data.

18. The method according to any one of claims 11 to 17, characterized in that, The method further includes: Send an instruction message to the first communication device, the instruction message being used to instruct the synthetic data to be used for model training or model fine-tuning.

19. A communication device, characterized in that, Includes units for performing the method as described in any one of claims 1 to 18.

20. A communication device, characterized in that, The device includes a processor coupled to a memory in which a computer program is stored; the processor is configured to invoke part or all of the computer program in the memory such that the method as described in any one of claims 1 to 18 is executed.

21. A communication system, characterized in that, The communication system includes a first communication device and a second communication device, wherein the first communication device is used to perform the method as described in any one of claims 1 to 10, and the second communication device is used to perform the method as described in any one of claims 11 to 18.

22. A computer-readable storage medium, characterized in that, The storage medium stores a computer program that, when some or all of the computer program is executed by a computer, causes the method described in any one of claims 1 to 18 to be performed.

23. A computer program product, characterized in that, When the computer reads and executes the computer program product, the method described in any one of claims 1 to 18 is performed.

Images