Knowledge base construction method, network element and storage medium

Abstract

The invention provides a knowledge base construction method, a network element and a storage medium. The method comprises the following steps: converting multi-modal data into a vector based on an embedded model under the condition that the multi-modal data is not a vector; and storing the vector corresponding to the multi-modal data to a knowledge base of a second network element.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of wireless communication, for example, to a knowledge base construction method, a network element and a storage medium. BACKGROUND

[0002] Retrieval-Augmented Generation (RAG) is a new technology for large model applications. In order to improve the inference quality of the core network multi-modal large model, RAG can be considered to be applied to the core network, but the current core network Artificial Intelligence (AI) architecture does not support RAG technology, and the current network element does not have the ability to construct a vector database. In addition, collecting and transmitting multi-modal data in the core network may have a security risk of data leakage. How to construct a safe and reliable knowledge base for multi-modal data is a problem to be solved. SUMMARY

[0003] The present application provides a knowledge base construction method, a network element and a storage medium.

[0004] The present application provides a knowledge base construction method, a network element and a storage medium.

[0005] In the case of non-vector multi-modal data, the multi-modal data is converted into a vector based on an embedding model;

[0006] The vector corresponding to the multi-modal data is stored in the knowledge base of the second network element.

[0007] The present application provides a knowledge base construction method, a network element and a storage medium.

[0008] Receiving the vector converted by the first network element from the multi-modal data;

[0009] Storing the vector in the knowledge base.

[0010] The present application provides a knowledge base construction method, a network element and a storage medium.

[0011] Sending a knowledge base update or deletion instruction to the second network element, the knowledge base update or deletion instruction being used to instruct the second network element to update or delete a specified target vector;

[0012] Receiving an update or deletion result instruction of the second network element.

[0013] The present application provides a knowledge base construction method, a network element and a storage medium.

[0014] Receiving a knowledge base update or deletion instruction sent by the first network element;

[0015] updating or deleting the specified target vector according to the knowledge base;

[0016] sending an updating or deleting result indication to the first network element.

[0017] The embodiment of the present application further provides a network element, which comprises a memory, a processor and a computer program stored in the memory and executable on the processor, and the processor implements the knowledge base construction method described above when executing the program.

[0018] The embodiment of the present application further provides a computer readable storage medium, which stores a computer program, and the program is executed by a processor to implement the knowledge base construction method described above. BRIEF DESCRIPTION OF DRAWINGS

[0019] Figure 1 a flow chart of a knowledge base construction method provided by an embodiment;

[0020] Figure 2 a flow chart of another knowledge base construction method provided by an embodiment;

[0021] Figure 3 a flow chart of still another knowledge base construction method provided by an embodiment;

[0022] Figure 4 a flow chart of yet another knowledge base construction method provided by an embodiment;

[0023] Figure 5 a schematic diagram of a capability information registration process of a second network element provided by an embodiment

[0024] Figure 6 a schematic diagram of a knowledge base construction process provided by an embodiment

[0025] Figure 7 a schematic diagram of a process in which one network element requests another network element to construct a knowledge base provided by an embodiment

[0026] Figure 8 a schematic diagram of a knowledge base construction process provided by an embodiment

[0027] Figure 9 a schematic diagram of a knowledge base updating or deleting process provided by an embodiment

[0028] Figure 10 a structural schematic diagram of a knowledge base construction device provided by an embodiment;

[0029] Figure 11 a structural schematic diagram of another knowledge base construction device provided by an embodiment;

[0030] Figure 12 FIG. 6 is a schematic diagram of a structure of a knowledge base construction apparatus according to an embodiment;

[0031] Figure 13 FIG. 7 is a schematic diagram of a structure of a knowledge base construction apparatus according to an embodiment;

[0032] Figure 14 FIG. 8 is a schematic diagram of a structure of a network element according to an embodiment. DETAILED DESCRIPTION

[0033] The present application will be described in detail below with reference to the drawings and embodiments. It can be understood that the specific embodiments described herein are merely intended for explaining the present application, but not limiting the present application. It should be noted that the embodiments in the present application and the features in the embodiments can be combined with each other in any manner without conflict. In addition, it should be noted that only the parts related to the present application are shown in the drawings for the convenience of description, but not all the structures.

[0034] The Network Data Analytics Function (NWDAF) is a 5G Core (5GC) Network Function (NF) located in the control plane that can perform statistical data and Machine Learning related tasks. The NWDAF can interact with different entities for different purposes, e.g. collect data from event subscriptions provided by the Access and Mobility Management Function (AMF), Session Management function (SMF), The User plane function (UPF), Policy Control function (PCF), The Unified Data Management (UDM), Network Slice Admission Control Function (NSACF), Application Function (AF), directly or through the Network Exposure Function (NEF), and Operation Administration and Maintenance (OAM); perform analytics and data collection using the Data Collection Coordination and Coordination and Delivery Function (DCCF); retrieve information from data repositories, e.g. UDR for user related information through the UDM or PFD information through the NEF (PFDF); collect location information data from LCS system; store and retrieve information from the Analytic Data Repository Function (ADRF); analyze and collect data from the Messaging Framework Adaptor Function (MFAF); retrieve information about NFs, e.g. NF related information from the NRF; provide analytics on demand to consumers, e.g. provide batch data related to an analytics ID; provide accuracy information for analytics IDs; provide Machine Learning (ML) model accuracy information or ML model accuracy degradation indication about ML models.A single instance or multiple instances of NWDAF can be deployed in a Public Land Mobile Network (PLMN). A NWDAF can contain the following logical functions:

[0035] Analytics Logical Function (AnLF): used to perform inference, derive analytics information (i.e. derive statistics and / or predictions from analytics consumer requests) and expose analytics services (i.e. Nnwdaf_AnalyticsSubscription or Nnwdaf_AnalyticsInfo);

[0036] Model Training Logical Function (MTLF): a logical function in NWDAF used to train ML models and expose new training services (e.g. provide trained ML models).

[0037] A NWDAF can contain one MTLF or one AnLF, or both.

[0038] DCCF is responsible for coordinating data collection and distribution requested by NF consumers, preventing data sources from handling multiple subscriptions to the same data, and preventing multiple notifications containing the same information from being sent due to uncoordinated requests from data consumers.

[0039] DCCF is applicable to: NWDAF requesting data from data sources (e.g. for computing analytics); NF consumer requesting analytics from NWDAF data sources; NF consumer requesting data from ADRF data sources; ADRF receiving data from NF data sources.

[0040] The 5G system architecture supports ADRF to store collected data and analytics. ADRF exposes Nadrf service for information storage and retrieval of data by other 5GC network functions (e.g. NWDAF).

[0041] Based on a request from a network function or a configuration on DCCF, DCCF can determine ADRF and interact with it directly or indirectly to request or store data. The way of interaction can be:

[0042] Direct: DCCF requests data storage in ADRF through Nadrf service or Ndccf_DataManagement_Notify (e.g. when ADRF requests DCCF to notify data collection). In addition, DCCF retrieves data from ADRF through Nadrf service.

[0043] Indirect: The DCCF requires the messaging framework to store data in the ADRF via the Nadrf service or Nmfaf_3daDataManagement_Configure. The messaging framework can contain one or more adapters that convert between protocols defined by 3GPP.

[0044] The consumer network function can specify in the request sent to the DCCF that data provided by the data source needs to be stored in the ADRF.

[0045] The ADRF stores data received in the Nadrf_DataManagement_StorageRequest sent directly from the network function or in the Ndccf_DataManagement_Notify / Nmfaf_3caDataManagement_Notify or Nnwdaf_DataManagement_Notify sent from the DCCF, MFAF or NWDAF.

[0046] RAG (Retrieval-Augmented Generation) is an emerging approach in natural language processing that combines the strengths of both retrieval and generation techniques. Specifically, RAG models rely on both the internal knowledge of pre-trained language models and the retrieval of relevant documents from external databases to enhance the accuracy and diversity of generated results when generating text.

[0047] The workflow of RAG mainly includes:

[0048] Retrieval phase: The model retrieves the most relevant document fragments from the external database based on the given input query. This step usually uses information retrieval techniques such as BM25 or vector retrieval.

[0049] Generation phase: Then, the input and the retrieved document fragments are input into the generation model. The generation model combines the input and the retrieved document fragments to generate more rich and accurate output.

[0050] When using large language models for inference, the parameter configurations used can significantly affect the quality and characteristics of the output content. The following are some common inference parameters and their roles: temperature, maximum length, prompt, short burst length, repetition penalty, Top-k sampling, Top-p (or Nucleus) sampling, batch size, and softness. These parameters can be used individually or in combination to improve the quality of the output analysis / prediction content. Depending on the specific application scenario, adjusting these parameters can significantly improve the output effect.

[0051] The embodiments of the present application provide a core network architecture supporting RAN, wherein the network element has a vector conversion or storage function, so that the multi-modal data is stored in the knowledge base in the form of vectors, and the knowledge base is constructed. The vectors in the knowledge base can be used for retrieval enhancement and multi-modal analysis or prediction.

[0052] Figure 1 A flowchart of a knowledge base construction method is provided for an embodiment. The method can be applied to a first network element, which is a network element that converts multi-modal data into vectors, such as NWDAF or NF. The converted vectors are stored in the knowledge base of a second network element, which has a vector storage function, such as ADRF or a new network function. As shown in Figure 1 The method provided by the embodiment includes steps 110 and 120.

[0053] In step 110, the multi-modal data is converted into a vector based on an embedding model when the multi-modal data is not in a vector form.

[0054] In step 120, the vector corresponding to the multi-modal data is stored in the knowledge base of the second network element.

[0055] In the embodiment, the multi-modal data can include pictures, audio, video, and / or text, etc. The multi-modal data obtained by the first network element can be in a vector form. If not, the first network element can convert the multi-modal data into a vector based on a local embedding model. If the first network element does not support vector conversion or lacks computing power, it can also request other network elements that support vector conversion or have sufficient computing power to perform the conversion, and then obtain the converted vector and store it.

[0056] In the embodiments of the present application, the knowledge base can also be referred to as a vector library, a vector database, a vector knowledge base, a knowledge database, or a vector knowledge database, etc.

[0057] In the embodiments of the present application, the function of the knowledge base can be implemented in ADRF or a new 5GC network element, such as a vector knowledge base network element.

[0058] In an embodiment of the present application, the embedding model can also be referred to as a vector conversion model, mainly referring to a conversion model used to convert multi-modal data into vectors.

[0059] In an embodiment, the method further comprises: sending a storage indication message to the second network element; and the storage indication message comprises at least one of the following: a link of the vector file; a size of the vector file; a type of the knowledge base; configuration information of the knowledge base; a knowledge base manufacturer; embedding model information; a knowledge classification identifier; and a knowledge base identifier.

[0060] In an embodiment, different knowledge corresponds to different knowledge classification identifiers; and one knowledge classification identifier corresponds to at least one knowledge base identifier.

[0061] In an embodiment, the apparatus further comprises: receiving a storage result indication of the second network element, and the storage result indication comprises at least one of the following:

[0062] a storage success indication, wherein the storage success indication information comprises a knowledge base identifier and / or a storage processing identifier used for storing the vector;

[0063] at least one of a storage failure indication and a storage failure reason.

[0064] In an embodiment, the method further comprises: receiving a vector conversion request before converting the multi-modal data into vectors based on the embedding model.

[0065] In an embodiment, the method further comprises: registering or updating multi-modal capability information to a third network element; and the multi-modal capability information comprises at least one of the following: a data type supporting multi-modal analysis or prediction; and a multi-modal data vector conversion capability; and the data type comprises at least one of the following: text, audio, video, and picture. The third network element can be an NRF.

[0066] In an embodiment, converting the multi-modal data into vectors based on the embedding model comprises:

[0067] In a case where the vector conversion is not supported locally or in a case where there is a shortage of computing power, sending a vector conversion request to a target network element supporting vector conversion, the vector conversion request being used to request the target network element to convert the multi-modal data into vectors based on the embedding model;

[0068] receiving the converted vectors.

[0069] In an embodiment, the method further comprises:

[0070] sending a discovery request of a network element supporting vector conversion to a third network element;

[0071] receive the list of network elements supporting vector conversion returned by the third network element and determine the target network element.

[0072] In an embodiment, the embedding model information is included in the discovery request.

[0073] In an embodiment, the embedding model information includes at least one of: an embedding model identifier, or an embedding model type, or an embedding model manufacturer.

[0074] In an embodiment, in the case where the target network element is not instructed to store the converted vector to the second network element, receive the vector conversion information returned by the target network element, the vector conversion information including a vector file of the vector, a link containing the vector file, a size of the vector file, and / or the embedding model information.

[0075] In the case where the target network element is instructed to store the converted vector to the second network element, store the converted vector to the second network element at the corresponding address by the target network element, or store the converted vector to the second network element having the capability of building a knowledge base, having sufficient memory margin, and meeting the vector conversion request by the target network element.

[0076] In an embodiment, converting the multi-modal data into vectors based on the embedding model includes:

[0077] In the case where the vector conversion request of the network element not supporting vector conversion or having insufficient computing power is received, converting the multi-modal data into vectors based on the embedding model.

[0078] In an embodiment, the method further includes:

[0079] In the case where the instruction of storing the converted vector to the second network element is not received, send the vector conversion information, the vector conversion information including a vector file of the vector, a link containing the vector file, a size of the vector file, and / or the embedding model information.

[0080] In the case where the instruction of storing the converted vector to the second network element is received, store the converted vector to the second network element at the corresponding address, or to the second network element having the capability of building a knowledge base, having sufficient memory margin, and meeting the vector conversion request.

[0081] In an embodiment, the method further includes:

[0082] send the vector storage information to the network element having the storage function;

[0083] The vector storage information includes at least one of:

[0084] a vector file of the vector; a link containing the vector file; a size of the vector file; embedding model information.

[0085] In an embodiment, the vector conversion request comprises at least one of the following: a file of the multi-modal data; a link of the file containing the multi-modal data; a type of the multi-modal data; embedding model information; an indication of storing to the second network element after conversion is completed; address information of the second network element; a knowledge classification identifier; a knowledge base identifier.

[0086] In an embodiment, the first network element is a NWDAF or an NF; and the second network element is an ADRF.

[0087] In an embodiment, the multi-modal data comprises at least one of the following: text, picture, video, and audio.

[0088] Figure 2 A flowchart of another method for constructing a knowledge base is provided for an embodiment, which can be applied to a second network element, which can be a network element for storing vectors or constructing a knowledge base. As shown in FIG. 10, the method provided by the embodiment comprises steps 1010 and 1020. Figure 2

[0089] In step 1010, a vector converted from multi-modal data sent by a first network element is received.

[0090] In step 1020, the vector is stored in a knowledge base.

[0091] In an embodiment, the method further comprises:

[0092] sending registration or update information to a third network element, the registration or update information comprising at least one of the following:

[0093] whether the second network element has the capability of constructing a knowledge base; a type of the knowledge base supported by the second network element; manufacturer information of the knowledge base supported by the second network element; memory remaining information.

[0094] In an embodiment, the memory remaining information comprises at least one of the following:

[0095] a memory remaining amount available for data storage;

[0096] a memory remaining amount available for model storage;

[0097] a memory remaining amount available for the knowledge base;

[0098] a percentage of a memory remaining amount available for data storage;

[0099] a percentage of a memory remaining amount available for model storage;

[0100] a percentage of a memory remaining amount available for the knowledge base;

[0101] a memory remaining shortage indication.

[0102] In an embodiment, the third network element is a NRF. ​

[0103] Figure 3 A flowchart of another knowledge base construction method provided by an embodiment is applicable to a first network element which can send an indication of updating or deleting a vector to a second network element storing the vector. The first network element is, for example, a NWDAF or an NF, and the second network element is, for example, an ADRF. As shown in Figure 3 the method provided by the embodiment includes steps 310 and 320.

[0104] In step 310, a knowledge base update or deletion indication is sent to the second network element, the knowledge base update or deletion indication being used to instruct the second network element to update or delete a specified target vector.

[0105] In step 320, an update or deletion result indication of the second network element is received.

[0106] In an embodiment, the knowledge base update or deletion indication includes at least one of the following: a knowledge base identifier; a storage processing identifier used to indicate a vector in the knowledge base; a vector file to be updated or deleted; a link containing the vector file; and a size of the vector file.

[0107] In an embodiment, at least one knowledge base is established in a second network element, each knowledge base corresponding to a unique knowledge base identifier.

[0108] In an embodiment, the update or deletion result indication includes at least one of the following:

[0109] a knowledge base identifier of successful update or deletion;

[0110] a storage processing identifier of successful update or deletion;

[0111] a knowledge base identifier of failed update or deletion;

[0112] a storage processing identifier of failed update or deletion;

[0113] a failure reason of update or deletion.

[0114] Figure 4 A flowchart of another knowledge base construction method provided by an embodiment is applicable to a second network element storing vectors and capable of updating or deleting the vectors. As shown in Figure 4 the method provided by the embodiment includes:

[0115] Step 410, receiving a knowledge base update or deletion indication sent by a first network element.

[0116] Step 420, updating or deleting a specified target vector according to the knowledge base update or deletion indication.

[0117] Step 430, sending an update or deletion result indication to the first network element.

[0118] In an embodiment, the knowledge base update or deletion indication includes at least one of the following: knowledge base identifier; storage processing identifier for indicating a vector in the knowledge base; vector file to be updated or deleted; link containing the vector file; and vector file size.

[0119] In an embodiment, the update or deletion result indication includes at least one of the following:

[0120] Knowledge base identifier of successful update or deletion;

[0121] Storage processing identifier of successful update or deletion;

[0122] Knowledge base identifier of failed update or deletion;

[0123] Storage processing identifier of failed update or deletion;

[0124] Failure reason of update or deletion.

[0125] The knowledge base construction method of the embodiments of the present application, the first network element (such as NWDAF) can use an embedded model to realize conversion of multi-modal data into vectors, and store the converted vector data into the knowledge base of the second network element (such as ADRF); if a local NWDAF does not support vector conversion or is insufficient in computing power, another NWDAF can be found and requested to perform vector conversion, or another 5GC network element can be requested to perform vector conversion, and instructed to store the converted vectors into the knowledge base; the first network element can also request to update or delete vectors in the knowledge base.

[0126] In addition, if a network element (such as NWDAF1) does not support vector conversion or is insufficient in computing power, another target network element (such as NWDAF2 with an embedded model) can be requested to perform vector conversion, and the converted vectors are returned to the network element or stored into the second network element according to the requirements of the network element.

[0127] The third network element (such as NRF) can receive the construction knowledge base capability, knowledge base information, and / or memory capacity information registered or updated by the second network element, and can also receive the multi-modal analysis or prediction capability, whether to support vector conversion, and / or the data types of multi-modal data supported for processing registered or updated by the first network element; the third network element supports discovering the first network element for vector conversion, and supports discovering the second network element for vector storage and knowledge base construction.

[0128] The second network element can register or update the construction knowledge base capability, knowledge base information, and / or memory capacity information to the third network element, can also receive vectors corresponding to multi-modal data, store the vectors to construct the knowledge base, and support updating and deleting vectors in the knowledge base.

[0129] The following describes the knowledge base construction method of the present application through some examples.

[0130] Example 1

[0131] In this embodiment, the registration or update process of the capability information of the second network element is mainly described. Figure 5 FIG. 1 is a schematic diagram of a second network element capability information registration process provided by an embodiment. Figure 5 As shown, the second network element is a 5GC network element (for example, ADRF, or other network elements), which can register or update the AI ​​model lifecycle capability to the third network element (such as NRF), and the registration or update information includes at least one of the following:

[0132] Whether there is the ability to build a knowledge base, where the knowledge base can be composed of vectors and / or matrices and can be used in RAG technology;

[0133] Types of supported knowledge bases;

[0134] Manufacturer information supporting the knowledge base;

[0135] Memory remaining information.

[0136] The memory remaining information may include at least one of the following:

[0137] Internal margin available for data storage;

[0138] The amount of memory available for model storage;

[0139] The amount of memory available for the knowledge base;

[0140] The percentage of memory remaining corresponding to the data model and / or knowledge base;

[0141] Memory shortage indication, for example, if the memory surplus of the ADRF local for data, model and / or knowledge base is less than a preset percentage of the image, a memory shortage indication may be sent to the NRF).

[0142] The above information may be included in the NF profile of the ADRF or other network elements.

[0143] The third network element may store the relevant information and return an indication of the success or failure of the registration / update.

[0144] In the embodiment of the present application, the knowledge base may also be referred to as a vector library, a vector database, a vector knowledge base, a knowledge database, or a vector knowledge database, etc. The function of the knowledge base may be implemented in the ADRF or may be a new 5GC network element, such as a vector knowledge base network element.

[0145] Embodiment 2

[0146] In this embodiment, the knowledge base construction process is mainly described. Figure 6 A schematic diagram of a knowledge base construction process is provided for an embodiment. As shown, the knowledge base construction process mainly includes: Figure 6

[0147] 1a. The first network element (such as NWDAF) sends a discovery request for a second network element (such as ADRF) supporting the construction of a knowledge base to a third network element (such as NRF), which can include the type of knowledge base, the manufacturer of the knowledge base, and / or memory requirements, etc.

[0148] 1b. The NRF matches the appropriate ADRF according to the discovery request information in 1a and the stored NF profile of the ADRF (see Embodiment 1), and returns the ADRF list to the NWDAF.

[0149] 2. The NWDAF collects multi-modal data such as text (such as the context of the UE), pictures, videos, and / or audio, etc. from other network elements (such as AF, RAN, and / or UE, etc.). It can also collect multi-modal data that has been converted into vectors.

[0150] 4. After the vector conversion is completed, the NWDAF stores the converted vectors into the knowledge base. The message can carry the following information: the converted vector file, the link with the vector file, the size of the vector file, the type of the knowledge base, the configuration information of the knowledge base, the manufacturer of the knowledge base, the embedding model information used for conversion (such as the embedding model identifier, the manufacturer of the embedding model, and / or the type of the embedding model, etc.), the knowledge classification ID, and / or the knowledge base ID, etc.

[0151] 5. The ADRF performs vector storage and may construct a new knowledge base for vector storage.

[0152] 6. The ADRF returns a storage success indication, which can include the assigned knowledge base ID (which can be included in the case that the NWDAF does not carry the knowledge base ID in step 4) and can also include the transaction ID; or the ADRF returns a storage failure indication and can also return the storage failure reason, such as the number of available knowledge bases or the knowledge base capacity reaching the upper limit, or insufficient memory, etc.

[0153] ​In the embodiments of the present application, the knowledge classification ID can be used to identify the classification of some knowledge, for example, knowledge classification ID 1 corresponds to UE session management related information, and knowledge classification ID 2 corresponds to UE mobility management related information. One knowledge classification ID can correspond to multiple knowledge base IDs. On this basis, for a knowledge base constructed by one NWDAF1, another NWDAF can analyze the knowledge classification according to the request sent by the consumer network element, and then send the knowledge classification ID to the ADRF and retrieve the related data. In the embodiments of the present application, one ADRF can establish one or more knowledge bases, and the knowledge base and the knowledge base ID are in a one-to-one correspondence. The transaction ID can be used to indicate part of the vector in a knowledge base.

[0154] In the embodiments of the present application, step 1 can occur before steps 2 and 3, or after steps 2 and 3.

[0155] In the embodiments of the present application, the embedding model can also be referred to as a vector conversion model, mainly referring to a conversion model used to convert multi-modal data into vectors.

[0156] Embodiment 3

[0157] In the embodiments, the process in which one network element requests another network element to construct a knowledge base is mainly described. Figure 7 A schematic diagram of a process in which one network element requests another network element to construct a knowledge base is provided in an embodiment. As shown in the figure, the NWDAF1 can request the NWDAF2 to perform vector conversion, and the NWDAF1 and the NWDAF2 can both be understood as a first network element. The knowledge base construction process mainly includes: Figure 7

[0158] 1. The NWDAF1 decides to perform vector conversion of multi-modal data (such as text, pictures, video, audio, etc.), but does not have an embedding model locally, does not have vector conversion capability, or cannot perform vector conversion due to insufficient computing power, and can send a NWDAF discovery request supporting vector conversion to a third network element (such as an NRF), which can include an embedding model ID, an embedding model type, and / or an embedding model manufacturer, etc.

[0159] 2. The NRF matches according to the pre-stored local information (i.e., the capability information registered or updated by the NWDAF), and returns a list of NWDAFs meeting the requirements.

[0160] ​3. NWDAF 1 selects NWDAF 2 to perform vector conversion. NWDAF 1 sends a vector conversion request to NWDAF 2, which can contain: the file of multi-modal data to be converted, the link of the file containing multi-modal data, the type of the file of multi-modal data, the embedding model information used for vector conversion (such as embedding model ID, embedding model type, embedding model vendor and / or configuration information of embedding model, etc.).

[0161] The vector conversion request can also contain an indication of storing the converted vector in the second network element (such as ADRF) after conversion; if the indication is contained, NWDAF 1 can also provide the following information: ADRF address information (such as ADRF ID), knowledge classification ID and / or knowledge base ID.

[0162] 4. NWDAF 2 uses the embedding model to perform vector conversion on the file of multi-modal data.

[0163] 5a. If the indication of storing the converted vector in the ADRF after conversion is not sent by NWDAF 1 in step 3, NWDAF 2 sends the converted vector file, the link of the vector file, the size of the vector file, the embedding model information used (such as model ID and / or embedding model type, etc.) to NWDAF 1.

[0164] 5b. If the indication of storing the converted vector in the ADRF after conversion is sent by NWDAF 1 in step 3 and the address information of ADRF is carried, NWDAF 2 can store the converted vector in the corresponding ADRF according to the ADRF address information; if the indication of storing the converted vector in the ADRF after conversion is sent by NWDAF 1 but the ADRF address information is not carried, NWDAF 2 can discover an ADRF that supports vector database capability and has enough memory through NRF, and supports the knowledge base type and vendor required by NWDAF 1.

[0165] NWDAF 2 stores the vector file in ADRF, and this message contains: the converted vector file, the link of the vector file, the vector database type and vendor information, the embedding model information used (embedding model ID, embedding model type and / or embedding model vendor), knowledge classification ID and / or knowledge base ID.

[0166] 5c. ADRF performs vector storage, which can construct a new knowledge base for vector storage (see step 5 in embodiment 2).

[0167] 5d. The ADRF returns a storage success indication, which can include the assigned knowledge base ID (which can be included in the case that the knowledge base ID is not carried by the NWDAF in step 3) and can also include the transaction ID; or, the ADRF returns a storage failure indication and can also return the storage failure cause, such as the number of available knowledge bases or the knowledge base capacity reaching the upper limit, or insufficient memory, etc.

[0168] 5e. The NWDAF 2 sends the information received in step 5d to the NWDAF 1.

[0169] Embodiment 4

[0170] In this embodiment, the process of constructing a knowledge base for one network element is described. Figure 8 A schematic diagram of a knowledge base construction process provided by an embodiment is shown in FIG. 4. As shown, the first network element can be an NF. The knowledge base construction process mainly includes: Figure 8

[0171] 1. The NWDAF sends a knowledge base construction indication to the NF and indicates that the vector conversion is performed using the embedded model in it, which can include: the embedded model information used, the knowledge classification ID, the knowledge base ID for storing after the vector conversion is completed, the knowledge base configuration information, the type, the knowledge base manufacturer information, the address information (ADRF ID) of the second network element (ADRF), and / or the data name / type to be converted, etc. It should be noted that step 1. is optional, and the NF can also trigger the vector storage autonomously.

[0172] 2. The target network element (such as a 5GC network element or an NF) performs vector conversion using the local embedded model according to the requirements of the NWDAF.

[0173] 3. The target network element stores the converted vector into the knowledge base, and this message includes: the converted vector file, or a link with the vector file, the vector database type, the manufacturer information, the embedded model information used (embedded model ID and / or embedded model type, embedded model manufacturer), the knowledge classification ID, and / or the knowledge base ID.

[0174] 4. The ADRF stores the vector and the embedded model information used according to the indication.

[0175] 5. The ADRF returns a storage success indication to the NF, which can include the assigned knowledge base ID (which can be included in the case that the knowledge base ID is not carried by the NWDAF in step 3) and can also include the transaction ID; or, the ADRF returns a storage failure indication and can also return the storage failure cause, such as the number of available knowledge bases or the knowledge base capacity reaching the upper limit, or insufficient memory, etc.

[0176] ​6. The NF sends the information received in step 5d to the NWDAF.

[0177] Embodiment 5

[0178] In this embodiment, the process of updating or deleting vectors in the knowledge base is mainly described. Figure 9 A schematic diagram of a knowledge base updating or deleting process provided by an embodiment is shown in FIG. 5. As shown in FIG. 5, the first network element can be the NWDAF or other 5GC network element, and the second network element can be the ADRF. The knowledge base updating or deleting process mainly includes: Figure 9

[0179] 1. The NWDAF or other 5GC network element sends the knowledge base content (vector) update / deletion indication to the ADRF, which can carry one or more sets of knowledge base ID + transaction ID and one or more vector files to be updated or a link containing the vector file and / or the size of the vector file, etc.

[0180] 2. The ADRF locates the vector to be updated or deleted according to the received knowledge base ID and transaction ID.

[0181] 3. The ADRF returns the knowledge base content (vector) update / deletion success or failure indication to the NWDAF, including one or more sets of knowledge base ID + transaction ID successfully updated or deleted and the failed knowledge base ID + transaction ID.

[0182] If the update or deletion fails, the ADRF can return the failure reason: for example, the knowledge base is found but the target vector is not found in the knowledge base, or the knowledge base is not found, etc.

[0183] If the update fails, the ADRF can return the failure reason: for example, insufficient memory, etc.

[0184] The present application also provides a knowledge base construction device. As shown in FIG. 6, the knowledge base construction device includes: Figure 10 The conversion module 510 is configured to convert the multi-modal data into a vector based on an embedding model in the case that the multi-modal data is not a vector;

[0185] The storage module 520 is configured to store the vector corresponding to the multi-modal data into the knowledge base of the second network element.

[0186] In an embodiment, the device further includes:

[0187]

[0188] ​​The storage indication module is configured to send a storage indication message to the second network element, and the storage indication message includes at least one of the following: a link of the vector file; a size of the vector file; a type of the knowledge base; configuration information of the knowledge base; a knowledge base manufacturer; embedded model information; a knowledge classification identifier; and a knowledge base identifier.

[0189] In an embodiment, different knowledge corresponds to different knowledge classification identifiers, and one knowledge classification identifier corresponds to at least one knowledge base identifier.

[0190] In an embodiment, the apparatus further includes:

[0191] The result receiving module is configured to receive a storage result indication of the second network element, and the storage result indication includes at least one of the following: a storage success indication, the storage success indication information including a knowledge base identifier and / or a storage processing identifier used for storing the vector; and at least one of a storage failure indication and a storage failure reason.

[0192] In an embodiment, before converting the multi-modal data into the vector based on the embedded model, the apparatus further includes:

[0193] The request receiving module is configured to receive a vector conversion request.

[0194] In an embodiment, the apparatus further includes:

[0195] The registration or updating module is configured to register or update multi-modal capability information to a third network element, and the multi-modal capability information includes at least one of the following: a data type supporting multi-modal analysis or prediction; and a multi-modal data vector conversion capability.

[0196] The data type includes at least one of the following: text, audio, video, and picture.

[0197] In an embodiment, the conversion module 510 is specifically configured to:

[0198] In a case where vector conversion is not supported locally or in a case where computing power is insufficient, a vector conversion request is sent to a target network element supporting vector conversion, the vector conversion request is used to request the target network element to convert multi-modal data into a vector based on an embedded model, and the converted vector is received.

[0199] In an embodiment, the apparatus further includes a discovery module configured to send a discovery request of a network element supporting vector conversion to a third network element, and receive a list of network elements supporting vector conversion returned by the third network element and determine a target network element.

[0200] In an embodiment, the discovery request includes embedded model information.

[0201] In an embodiment, the embedding model information comprises at least one of: an embedding model identifier, an embedding model type, and an embedding model manufacturer.

[0202] In an embodiment, the discovery module is further configured to, in a case where it is not indicated that the target network element is to store the converted vector to the second network element, receive vector conversion information returned by the target network element, the vector conversion information comprising a vector file of the vector, a link containing the vector file, a size of the vector file, and / or embedding model information.

[0203] In a case where it is indicated that the target network element is to store the converted vector to the second network element, the converted vector is stored to the second network element at a corresponding address by the target network element, or the converted vector is stored to a second network element that has the capability of building a knowledge base, has sufficient memory margin, and meets the vector conversion request by the target network element.

[0204] In an embodiment, the converting the multi-modal data into a vector based on the embedding model comprises: in a case where a vector conversion request of a network element that does not support local vector conversion or has insufficient computing power is received, converting the multi-modal data into a vector based on the embedding model.

[0205] In an embodiment, the apparatus further comprises a conversion information sending module configured to, in a case where an indication of storing the converted vector to the second network element is not received, send vector conversion information, the vector conversion information comprising a vector file of the vector, a link containing the vector file, a size of the vector file, and / or embedding model information.

[0206] The storage module 520 is configured to, in a case where an indication of storing the converted vector to the second network element is received, store the converted vector to the second network element at a corresponding address, or to a second network element that has the capability of building a knowledge base, has sufficient memory margin, and meets the vector conversion request.

[0207] In an embodiment, the apparatus further comprises a storage information sending module configured to send vector storage information to a network element having a storage function, the vector storage information comprising at least one of: a vector file of the vector; a link containing the vector file; a size of the vector file; and embedding model information.

[0208] In an embodiment, the vector conversion request comprises at least one of: a file of the multi-modal data; a link containing the file of the multi-modal data; a type of the multi-modal data; embedding model information; an indication of storing to a second network element after conversion is completed; address information of the second network element; a knowledge classification identifier; and a knowledge base identifier.

[0209] In an embodiment, the first network element is a NWDAF or an NF, and the second network element is an ADRF.

[0210] In an embodiment, the multi-modal data comprises at least one of: text, picture, video, audio.

[0211] The knowledge base construction device proposed in the embodiment belongs to the same inventive concept as the knowledge base construction method proposed in the above embodiments, and the technical details not described in detail in the embodiment can be referred to the above any embodiment, and the embodiment has the same beneficial effects as performing the knowledge base construction method.

[0212] The present application also provides another knowledge base construction device. As shown in the Figure 11 The knowledge base construction device comprises:

[0213] The vector receiving module 610 is configured to receive the vector converted from the multi-modal data sent by the first network element;

[0214] The vector storage module 620 is configured to store the vector into the knowledge base.

[0215] In an embodiment, the device further comprises:

[0216] The registration or update module is configured to send registration or update information to the third network element, and the registration or update information comprises at least one of:

[0217] Whether it has the ability to construct a knowledge base; the type of supported knowledge base; the manufacturer information of the supported knowledge base; the memory remaining information.

[0218] In an embodiment, the memory remaining information comprises at least one of:

[0219] The memory remaining available for data storage;

[0220] The memory remaining available for model storage;

[0221] The memory remaining available for the knowledge base;

[0222] The percentage of memory remaining available for data storage;

[0223] The percentage of memory remaining available for model storage;

[0224] The percentage of memory remaining available for the knowledge base;

[0225] The memory remaining shortage indication.

[0226] In an embodiment, the third network element is NRF.

[0227] The knowledge base construction device of the embodiment belongs to the same inventive concept as the knowledge base construction method of the above embodiments, and the technical details not described in detail in the embodiment can be seen from any of the above embodiments, and the embodiment has the same beneficial effects as the knowledge base construction method.

[0228] The embodiment of the present application further provides another knowledge base construction device. As shown in Figure 12 The knowledge base construction device comprises:

[0229] The update or deletion instruction module 710 is configured to send a knowledge base update or deletion instruction to a second network element, the knowledge base update or deletion instruction being used to instruct the second network element to update or delete a specified target vector;

[0230] The result receiving module 720 is configured to receive an update or deletion result instruction of the second network element.

[0231] In an embodiment, the knowledge base update or deletion instruction comprises at least one of the following:

[0232] A knowledge base identifier; a storage processing identifier of a vector in the knowledge base; a vector file to be updated or deleted; a link containing the vector file; and a size of the vector file.

[0233] In an embodiment, at least one knowledge base is established in a second network element, and each knowledge base corresponds to a unique knowledge base identifier.

[0234] In an embodiment, the update or deletion result instruction comprises at least one of the following:

[0235] A knowledge base identifier of successful update or deletion;

[0236] A storage processing identifier of successful update or deletion;

[0237] A knowledge base identifier of failed update or deletion;

[0238] A storage processing identifier of failed update or deletion;

[0239] A failure reason of update or deletion.

[0240] The knowledge base construction device of the embodiment belongs to the same inventive concept as the knowledge base construction method of the above embodiments, and the technical details not described in detail in the embodiment can be seen from any of the above embodiments, and the embodiment has the same beneficial effects as the knowledge base construction method.

[0241] The embodiment of the present application further provides another knowledge base construction device. As shown in Figure 13 The knowledge base construction device comprises:

[0242] The indication receiving module 810 is configured to receive the knowledge base update or deletion indication sent by the first network element;

[0243] The update or deletion module 820 is configured to update or delete the specified target vector according to the knowledge base update or deletion indication;

[0244] The update or deletion result indication is sent to the first network element.

[0245] In an embodiment, the knowledge base update or deletion indication comprises at least one of the following: knowledge base identifier; storage processing identifier for indicating the vector in the knowledge base; vector file to be updated or to be deleted; link containing the vector file; and vector file size.

[0246] In an embodiment, the update or deletion result indication comprises at least one of the following:

[0247] Knowledge base identifier of successful update or deletion;

[0248] Storage processing identifier of successful update or deletion;

[0249] Knowledge base identifier of failed update or deletion;

[0250] Storage processing identifier of failed update or deletion;

[0251] Reason for failed update or deletion.

[0252] The knowledge base construction apparatus proposed in the embodiment belongs to the same inventive concept as the knowledge base construction method proposed in the above embodiments, and the technical details not described in detail in the embodiment can be referred to the above embodiments, and the embodiment has the same beneficial effects as the knowledge base construction method.

[0253] The present application also provides a network element, Figure 14 A structural schematic diagram of a network element provided in an embodiment is shown in the figure, and the network element provided in the present application comprises a processor 910 and a memory 920. Figure 14 The processor 910 in the network element can be one or more processors. Figure 14 The memory 920 is configured to store one or more programs; and the one or more programs are executed by the one or more processors 910, so that the one or more processors 910 implement the knowledge base construction method as described in the embodiments of the present application.

[0254] The network element further comprises a communication device 930, an input device 940 and an output device 950.

[0255] The processor 910, the memory 920, the communication device 930, the input device 940 and the output device 950 in the network element can be connected through a bus or other means, Figure 14 The bus connection is taken as an example.

[0256] The input device 940 can be used to receive input digital or character information, and generate key signal input related to user settings and function control of the network element. The output device 950 can include a display device such as a display screen.

[0257] The communication device 930 can include a receiver and a transmitter. The communication device 930 is configured to perform information receiving and transmitting communication under the control of the processor 910.

[0258] The memory 920, as a computer readable storage medium, can be configured to store software programs, computer executable programs and modules, such as program instructions / modules corresponding to the knowledge base construction method described in the embodiments of the present application (for example, modules in the knowledge base construction device). The memory 920 can include a program storage area and a data storage area, wherein the program storage area can store an operating system and application programs required by at least one function; the data storage area can store data created according to the use of the network element, etc. In addition, the memory 920 can include a high-speed random access memory, and can also include a non-volatile memory, such as at least one magnetic disk storage device, a flash memory device, or other non-volatile solid-state storage device. In some examples, the memory 920 can further include a memory remotely arranged with respect to the processor 910, and these remote memories can be connected to the network element through a network. Examples of the above network include but are not limited to the Internet, an intranet, a local area network, a mobile communication network and a combination thereof.

[0259] The embodiments of the present application also provide a storage medium, which stores a computer program, and the computer program is executed by a processor to implement the knowledge base construction method described in any of the embodiments of the present application. The method is applied to a first network element, and includes: in the case of non-vector of multi-modal data, converting the multi-modal data into a vector based on an embedding model; storing the vector corresponding to the multi-modal data into a knowledge base of a second network element. Alternatively, the method is applied to a second network element, and includes: receiving a vector converted from multi-modal data sent by a first network element; storing the vector into a knowledge base. Alternatively, the method is applied to a first network element, and includes: sending a knowledge base update or deletion instruction to a second network element, the knowledge base update or deletion instruction being used to instruct the second network element to update or delete a specified target vector; receiving an update or deletion result instruction of the second network element. Alternatively, the method is applied to a second network element, and includes: receiving a knowledge base update or deletion instruction sent by a first network element; updating or deleting a specified target vector according to the knowledge base update or deletion instruction; sending an update or deletion result instruction to the first network element.

[0260] The embodiments of the present application further provide a computer storage medium, including computer programs / instructions, which are executed by a processor to implement the knowledge base construction method described in any of the embodiments of the present application. The method is applied to a first network element and includes: converting multi-modal data into a vector based on an embedding model in a case where the multi-modal data is not a vector; and storing the vector corresponding to the multi-modal data in a knowledge base of a second network element. Alternatively, the method is applied to the second network element and includes: receiving a vector converted from multi-modal data sent by the first network element; and storing the vector in the knowledge base. Alternatively, the method is applied to the first network element and includes: sending a knowledge base update or deletion instruction to the second network element, where the knowledge base update or deletion instruction is used to instruct the second network element to update or delete a specified target vector; and receiving an update or deletion result instruction of the second network element. Alternatively, the method is applied to the second network element and includes: receiving a knowledge base update or deletion instruction sent by the first network element; updating or deleting a specified target vector according to the knowledge base update or deletion instruction; and sending an update or deletion result instruction to the first network element.

[0261] The computer storage medium of the embodiments of the present application can adopt any combination of one or more computer readable media. The computer readable medium can be a computer readable signal medium or a computer readable storage medium. The computer readable storage medium may, for example, be, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or apparatus, or any combination of the above. More specific examples (a non-exhaustive list) of the computer readable storage medium include an electrical connection having one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read only memory (ROM), an erasable programmable read only memory (EPROM), a flash memory, an optical fiber, a portable CD-ROM, an optical storage device, a magnetic storage device, or any suitable combination of the above. The computer readable storage medium can be any tangible medium containing or storing a program that can be used by or in connection with an instruction execution system, apparatus or device.

[0262] A computer readable signal medium can include a propagated data signal with computer executable code embodied therein, for example, in baseband or as part of a carrier wave. Such a propagated signal can take any of a variety of forms, including, but not limited to, electro-magnetic, optical, or any suitable combination thereof. A computer readable signal medium can be any computer readable medium that can be involved in

[0263] The code can be transmitted in any form, including, but not limited to, radio frequency (RF) signals, digital signals, light, sound, or other media. The media can take a variety of forms, including, but not limited to, wire, cable, fiber optics, and / or wireless media. The computer program code can be implemented in any of a variety of programming languages, including, but not limited to, C, C++, Java, and / or Visual Basic, among others.

[0264] Computer program code for carrying out operations of the present application can be written in any combination of one or more programming languages, including an object oriented programming language such as Java, Smalltalk, C++ or the like, and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code can execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection can be made to an external computer (for example, through the Internet using an Internet Service Provider).

[0265] The embodiments of the present application also provide a computer program product, including computer program / instructions, which, when executed by a processor, implement the video encoding method according to any of the above embodiments.

[0266] The above description is merely illustrative of the application, and not in limitation of the application's scope since modifications can be made by those skilled in the art, now or in the future, without departing from the scope of the application.

[0267] Those skilled in the art will appreciate that the term user terminal encompasses any suitable type of wireless user device, such as a mobile phone, a portable data processing portable network browser or a vehicle mounted mobile station.

[0268] In general, the various embodiments of the application can be implemented in hardware or special purpose circuits, software, logic or any combination thereof. For example, some aspects can be implemented in hardware, while other aspects can be implemented in software that can be executed by a controller, microprocessor or other computing device, although the application is not limited thereto. The software implementing some aspects of the application

[0269] Embodiments of the application can be implemented by computer program instructions on a data processor of a mobile device, for example in a processor entity, or by hardware, or by a combination of software and hardware. Computer program instructions can be in the form of assembly instructions, Instruction Set Architecture (ISA) instructions, machine instructions, machine dependent instructions, microcode, firmware instructions, state-setting data, or in any combination of one or more programming languages, written in any combination of one or more of a plurality of programming languages. As such, embodiments of the application can be viewed as being composed of various interrelated computer processes.

[0270] Any block diagrams of logical flows of the application can represent program steps, or logical states of interconnected logic circuits, blocks and functions, or a combination of program steps and logic circuits, blocks and functions. The computer program can be stored on a memory. The memory can be of any suitable type and can be implemented using any suitable data storage technology, such as, but not limited to, a random access memory (RAM), a read-only memory (ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. The computer readable media can include non-transitory storage media. The data processor can be of any suitable type and can be implemented using any suitable combination of hardware and software. The data processor can be, but is not limited to, a general purpose computer, a microprocessor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA) or any combination thereof.

[0271] The foregoing detailed description of the exemplary embodiments of the application has been presented for the purposes of illustration and description. Many modifications and variations of the above described embodiments of the application are possible in light of the above teachings. It is therefore, intended to cover within the ambit of the application, all modifications and variations as long as they fall within the scope of the application. Accordingly, the proper scope of the application is to be determined solely from the appended claims.

Claims

1. A knowledge base construction method, applied to a first network element, characterized in that: include: In the case where the multimodal data is non-vector, converting the multimodal data into a vector based on the embedding model; The vector corresponding to the multimodal data is stored in a knowledge base of the second network element.

2. The method according to claim 1, characterized in that Also includes: Sending a storage indication message to the second network element; the storage indication message includes at least one of the following: Link to vector file; size of vector file; type of knowledge base; configuration information of knowledge base; knowledge base manufacturer; embedded model information; knowledge classification identifier; knowledge base identifier.

3. The method according to claim 2, characterized in that Different knowledge corresponds to different knowledge classification identifiers; A knowledge classification identifier corresponds to at least one knowledge base identifier.

4. The method according to claim 1, wherein Also includes: Receive a storage result indication from the second network element, where the storage result indication includes at least one of the following: A storage success indication, wherein the storage success indication information includes a knowledge base identifier and / or a storage processing identifier for storing the vector; At least one of a storage failure indication and a storage failure reason.

5. The method according to claim 1, wherein Before converting the multimodal data into a vector based on the embedding model, the method further includes: Receives a vector conversion request.

6. The method according to claim 1, characterized in that Also includes: Registering or updating multimodal capability information with a third network element; the multimodal capability information includes at least one of the following: Supports data types for multimodal analysis or prediction; multimodal data vector conversion capabilities; The data type includes at least one of the following: text, audio, video, and picture.

7. The method according to claim 1, characterized in that Converting the multimodal data into vectors based on an embedding model includes: When vector conversion is not supported locally or computing power is insufficient, a vector conversion request is sent to a target network element that supports vector conversion, wherein the vector conversion request is used to request the target network element to convert the multimodal data into a vector based on the embedding model; Receives the transformed vector.

8. The method according to claim 7, characterized in that Also includes: Sending a discovery request of a network element supporting vector conversion to a third network element; Receive the list of network elements supporting vector conversion returned by the third network element and determine the target network element.

9. The method according to claim 8, characterized in that The discovery request includes embedded model information.

10. The method according to claim 2 or 9, characterized in that The embedded model information includes at least one of the following: embedded model identification, embedded model type, embedded model manufacturer.

11. The method according to claim 7, characterized in that Also includes: In a case where the target network element is not instructed to store the converted vector in the second network element, receiving vector conversion information returned by the target network element, the vector conversion information including a vector file of the vector, a link containing the vector file, a size of the vector file, and / or embedded model information; When the target network element is instructed to store the converted vector in the second network element, the converted vector is stored in the second network element of the corresponding address through the target network element, or the converted vector is stored in the second network element that has the ability to build a knowledge base, has sufficient memory margin and meets the vector conversion request through the target network element.

12. The method according to claim 1, characterized in that Converting the multimodal data into vectors based on an embedding model includes: When a vector conversion request is received from a network element that does not locally support vector conversion or has insufficient computing power, the multimodal data is converted into vectors based on the embedding model.

13. The method according to claim 12, characterized in that Also includes: In a case where no instruction to store the converted vector in the second network element is received, sending vector conversion information, the vector conversion information including a vector file of the vector, a link containing the vector file, a size of the vector file, and / or embedded model information; When an instruction to store the converted vector in a second network element is received, the converted vector is stored in the second network element at the corresponding address, or in a second network element that has the ability to build a knowledge base, has sufficient memory margin and meets the vector conversion request.

14. The method according to claim 1, wherein Also includes: Sending vector storage information to a network element with storage function; The vector storage information includes at least one of the following: A vector file of the vector; a link containing the vector file; a size of the vector file; and embedded model information.

15. The method according to claim 1, wherein The vector conversion request includes at least one of the following: The file of the multimodal data; a link to a file containing the multimodal data; the type of the multimodal data; Embedded model information; an instruction to store in the second network element after conversion is completed; address information of the second network element; knowledge classification identifier; knowledge base identifier.

16. The method according to claim 1, characterized in that The first network element is an NWDAF or a NF; The second network element is ADRF.

17. The method according to claim 1, wherein The multimodal data includes at least one of the following: text, picture, video, and audio.

18. A knowledge base construction method, applied to a second network element, characterized in that: include: receiving a vector obtained by converting multimodal data sent by the first network element; The vector is stored in a knowledge base.

19. The method according to claim 18, characterized in that Also includes: Sending registration or update information to the third network element, where the registration or update information includes at least one of the following: Whether it has the ability to build a knowledge base; the types of knowledge bases supported; the manufacturer information of the supported knowledge bases; and the memory remaining information.

20. The method according to claim 19, characterized in that The memory remaining information includes at least one of the following: The amount of memory available for data storage; The amount of memory available for model storage; The amount of memory available for the knowledge base; The percentage of memory available for data storage; The percentage of memory available for model storage; The percentage of memory available for the knowledge base; Low memory indication.

21. The method according to claim 19, wherein The third network element is NRF.

22. A knowledge base construction method, applied to a first network element, characterized in that: include: Sending a knowledge base update or deletion instruction to the second network element, where the knowledge base update or deletion instruction is used to instruct the second network element to update or delete a specified target vector; Receive an update or deletion result indication of the second network element.

23. The method according to claim 22, characterized in that The knowledge base update or deletion instruction includes at least one of the following: Knowledge base identifier; a storage processing identifier for indicating a vector in the knowledge base; a vector file to be updated or deleted; a link containing the vector file; and the size of the vector file.

24. The method according to claim 22, characterized in that At least one knowledge base is established in a second network element, and each knowledge base corresponds to a unique knowledge base identifier.

25. The method according to claim 22, wherein The update or deletion result indication includes at least one of the following: The ID of the knowledge base that was successfully updated or deleted; The storage transaction identifier of a successful update or deletion; Update or delete failed knowledge base identification; Update or delete the failed storage processing identifier; The reason why the update or delete failed.

26. A knowledge base construction method, applied to a second network element, characterized in that: include: receiving a knowledge base update or deletion instruction sent by the first network element; updating or deleting a specified target vector according to the knowledge base update or deletion instruction; Send an update or deletion result indication to the first network element.

27. The method according to claim 26, characterized in that The knowledge base update or deletion instruction includes at least one of the following: Knowledge base identifier; a storage processing identifier for indicating a vector in the knowledge base; a vector file to be updated or deleted; a link containing the vector file; and the size of the vector file.

28. The method according to claim 26, characterized in that The update or deletion result indication includes at least one of the following: The ID of the knowledge base that was successfully updated or deleted; The storage transaction identifier of a successful update or deletion; Update or delete failed knowledge base identification; Update or delete the failed storage processing identifier; The reason why the update or delete failed.

29. A network element, characterized in that: include: memory, and one or more processors; The memory is configured to store one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the knowledge base construction method according to any one of claims 1 to 28.

30. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the knowledge base construction method as described in any one of claims 1 to 28 is implemented.

Images