Knowledge data acquisition method, knowledge base construction method, and electronic device

Abstract

Embodiments of the present application provide a knowledge data acquisition method, a knowledge base construction method, and an electronic device. The knowledge data acquisition method comprises: receiving a target request; acquiring, from a target knowledge plug-in package, knowledge data corresponding to the target request; and feeding back the knowledge data, wherein the target knowledge plug-in package comprises an atomic granularity knowledge plug-in package for knowledge data acquisition, and the atomic granularity knowledge plug-in package comprises a plug-in identifier, knowledge directory information, and knowledge content.

Description

Knowledge data acquisition method, knowledge base construction method and electronic device

[0001] Cross-reference to related applications

[0002] The present application claims priority to the Chinese patent application No. 202410811897.X, filed on June 21, 2024, and entitled "Knowledge data acquisition method, knowledge base construction method and electronic device", the content of which is incorporated herein by reference in its entirety. TECHNICAL FIELD

[0003] The present application belongs to the field of computer, and particularly relates to a knowledge data acquisition method, a knowledge base construction method and an electronic device. BACKGROUND

[0004] With the continuous development of technology, large language models (LLM) such as LLM for intelligent question answering have been more and more widely applied.

[0005] Due to the possible lack of knowledge reserves of LLM itself, knowledge expansion is often achieved by using knowledge plugins in the related art. Specifically, LLM often calls external knowledge data by means of knowledge plugins to achieve the effect of providing LLM with expanded knowledge. In this way, there is a problem of insufficient flexibility in acquiring knowledge data for LLM. SUMMARY

[0006] The embodiments of the present application provide a knowledge data acquisition method, a knowledge base construction method and an electronic device to solve the problem of insufficient flexibility in acquiring knowledge data in the related art.

[0007] In a first aspect, the embodiments of the present application provide a knowledge data acquisition method, comprising: receiving a target request; acquiring knowledge data corresponding to the target request from a target knowledge plugin package; and feeding back the knowledge data; wherein the target knowledge plugin package comprises an atomic granularity knowledge plugin package for knowledge data acquisition, and the atomic granularity knowledge plugin package comprises a plugin identifier, knowledge directory information and knowledge content.

[0008] In a second aspect, the embodiments of the present application provide a knowledge base construction method, comprising: acquiring knowledge corpus for a target application scenario; processing the knowledge corpus to obtain P atomic granularity knowledge plugin packages, each of the P atomic granularity knowledge plugin packages comprising a plugin identifier, knowledge directory information and knowledge content itself; obtaining a knowledge base based on the P atomic granularity knowledge plugin packages; and P is a positive integer.

[0009] In a third aspect, an electronic device is provided, which includes a processor and a memory. The memory stores programs or instructions running on the processor. When the programs or instructions are executed by the processor, the steps of the method according to the first aspect and / or the second aspect are implemented.

[0010] In a fourth aspect, a computer readable storage medium is provided, which stores programs or instructions. When the programs or instructions are executed, the steps of the method according to the first aspect and / or the second aspect are implemented.

[0011] In a fifth aspect, a computer program product is provided, which includes a computer program. When the computer program is executed by a processor, the steps of the method according to the first aspect and / or the second aspect are implemented. BRIEF DESCRIPTION OF DRAWINGS

[0012] FIG. 1 is a flowchart of a knowledge data acquisition method according to an embodiment of the present application;

[0013] FIG. 2 is a flowchart of another knowledge data acquisition method according to an embodiment of the present application;

[0014] FIG. 3 is a flowchart of another knowledge data acquisition method according to an embodiment of the present application;

[0015] FIG. 4 is a flowchart of another knowledge data acquisition method according to an embodiment of the present application;

[0016] FIG. 5 is a flowchart of a knowledge base construction method according to an embodiment of the present application;

[0017] FIG. 6-1 is a schematic diagram of a knowledge plug-in construction process according to an embodiment of the present application;

[0018] FIG. 6-2 is a schematic diagram of a knowledge plug-in configuration management method for a large model application according to an embodiment of the present application;

[0019] FIG. 7 is a schematic diagram of a corpus pipeline according to an embodiment of the present application;

[0020] FIG. 8 is a schematic diagram of a target knowledge base according to an embodiment of the present application;

[0021] FIG. 9 is a structural block diagram of a data acquisition apparatus according to an embodiment of the present application;

[0022] FIG. 10 is a structural block diagram of a knowledge base construction apparatus according to an embodiment of the present application;

[0023] FIG. 11 is a schematic diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0024] With reference to the drawings and the embodiments of the present application, the technical solutions in the embodiments of the present application will be clearly described. Obviously, the described embodiments are only some of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art are within the scope of protection of the present application.

[0025] The terms "first", "second", etc. in the specification and claims of the present application are used to distinguish similar objects, and are not used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments of the present application can be implemented in an order other than those illustrated or described herein, and the objects distinguished by "first", "second", etc. are generally of a kind and do not limit the number of objects, for example, the first object can be one or more. In addition, "and / or" in the specification and claims indicates at least one of the connected objects, and the character " / ", generally indicates that the objects before and after are in a "or" relationship.

[0026] As described in the background section, LLMs often call external knowledge data with the help of knowledge plugins to provide extended knowledge for LLMs. In this way, there is a problem of insufficient flexibility in obtaining knowledge data for LLMs. The embodiments of the present application can flexibly obtain knowledge data from the atomic granularity knowledge plugin package in the target knowledge plugin package, which includes plugin identification, knowledge directory information and knowledge content, to solve the problem of insufficient flexibility in obtaining knowledge data in the related art.

[0027] The knowledge data obtaining method provided by the embodiments of the present application can be applied to a scenario where an external knowledge base exists in a target model (such as a large language model), and the target knowledge plugin package can be a plugin package in the external knowledge base of the target model. The external knowledge base plays an increasingly important role in improving the application performance of a large language model (LLM). The external knowledge base can provide the LLM with more abundant, accurate and timely knowledge reserves, thereby significantly enhancing the comprehensiveness and accuracy of the LLM in answering questions, and ultimately optimizing the overall experience of users.

[0028] In actual application process, different application demand parties have diversified demands for the external knowledge base. Therefore, different external knowledge bases need to be configured according to different application demand parties. In this way, there is a problem of poor flexibility of knowledge base content adjustment in the related art. The atomic granularity knowledge plug-in package in the target knowledge plug-in package in the embodiment of the present application can be a reusable plug-in package. The combined knowledge plug-in package meeting the application scenario requirement can be obtained by combining the reusable atomic granularity knowledge plug-in package. The combined knowledge plug-in package can exist in the target knowledge plug-in package. In this way, since the combined knowledge plug-in package is obtained by combining a plurality of atomic granularity knowledge plug-in packages, each of the plurality of atomic granularity knowledge plug-in packages is a reusable knowledge plug-in package. Various combined knowledge plug-in packages can be obtained by using these reusable atomic granularity knowledge plug-in packages. In turn, flexible adjustment of the knowledge base content can be realized, and the problem of poor flexibility of knowledge base content adjustment in the related art is solved.

[0029] In addition, the knowledge base configured in the related art is tightly coupled with the large language model software. The update of the knowledge in the knowledge base is limited by the software release cycle of the large language model. The knowledge base can only be updated at the software update time of the large language model. The target knowledge plug-in package in the embodiment of the present application can be dynamically updated during the running period of the knowledge base, without the need for updating again during the software release cycle. In turn, the knowledge base update and the large language model software are decoupled. For example, in the embodiment of the present application, after the knowledge data is fed back, an update operation can be performed on the target knowledge plug-in package to obtain an updated target knowledge plug-in package. The updated target knowledge plug-in package is used for subsequent acquisition of knowledge data. Compared with the update of the external knowledge base in the process of updating the large model in the related art, the embodiment of the present application can update the target knowledge plug-in, decouple the update of the external knowledge base and the update of the target model, and realize flexible update of the external knowledge base.

[0030] Meanwhile, the target knowledge plug-in package can be a plug-in package from a finished product library. The content of the plug-in package in the finished product library is not allowed to be modified during the running period of the external knowledge base. In this way, the stability and security of the target knowledge plug-in package can be ensured, and it is also convenient to update the target knowledge plug-in package by acquiring the plug-in package from the finished product library. The finished product library can contain atomic granularity knowledge plug-in packages and combined knowledge plug-in packages. In this way, the required knowledge plug-in package can be conveniently obtained from the finished product library.

[0031] In the embodiment of the present application, in order to further improve the security, the atomic granularity knowledge plug-in package can contain identity verification information, and each plug-in package in the target knowledge plug-in package is a plug-in package whose identity verification information is verified.

[0032] In addition, the embodiment of the present application further provides a knowledge base construction method, comprising: acquiring knowledge corpus for a target application scenario; processing the knowledge corpus to obtain P atomic granularity knowledge plug-in packages, each of the P atomic granularity knowledge plug-in packages comprising plug-in identification, knowledge directory information and knowledge content itself; obtaining a knowledge base based on the P atomic granularity knowledge plug-in packages; P is a positive integer. Since the knowledge base is constructed based on the atomic granularity knowledge plug-in packages, the atomic granularity knowledge plug-in packages comprise plug-in identification, knowledge directory information and knowledge content, knowledge data can be flexibly acquired from the atomic granularity knowledge plug-in packages, and the problem of insufficient flexibility in acquiring knowledge data in the related art is solved.

[0033] It should be understood that the method for acquiring knowledge data provided by the embodiment of the present application can be executed by a target device. The target device can be one electronic device or multiple electronic devices cooperating with each other. The electronic device can be a server, such as a physical server, a server cluster composed of multiple servers, and a cloud server capable of cloud computing. For example, the target device can be all devices in a distributed system or part of the devices in the distributed system.

[0034] The method provided by the embodiment of the present application will be described in detail below in combination with the accompanying drawings, embodiments and application scenarios.

[0035] FIG. 1 is a flowchart of a method for acquiring knowledge data provided by the embodiment of the present application. As shown in FIG. 1, the method for acquiring knowledge data provided by the embodiment of the present application comprises the following steps.

[0036] Step 110, receiving a target request.

[0037] In the embodiment of the present application, the target request can be a question and answer request in any field, for example, the target request can be a request in the communication field involving a tutorial for configuring a certain service in network management software, or a request in the education field involving a tutorial for solving a certain problem.

[0038] In the embodiment of the present application, the target device can directly acquire the target request through an external device (such as a keyboard or a mouse), or can receive the target request transmitted by another device. The embodiment of the present application does not limit the source of the target request.

[0039] Step 120, acquiring knowledge data corresponding to the target request from a target knowledge plug-in package; wherein the target knowledge plug-in package comprises an atomic granularity knowledge plug-in package for acquiring knowledge data, and the atomic granularity knowledge plug-in package comprises plug-in identification, knowledge directory information and knowledge content.

[0040] In the embodiment of the present application, the plug-in identifier of the atomic granularity knowledge plug-in package can be unique identifier information (for example, plug-in ID) of the atomic granularity knowledge plug-in package, used to distinguish different atomic granularity knowledge plug-in packages. The knowledge directory information can be directory information of the knowledge content in the atomic granularity knowledge plug-in package, which can be determined by the knowledge content in the atomic granularity knowledge plug-in package. For example, when the knowledge content in the atomic granularity knowledge plug-in package is the content of one document, the knowledge directory information can include the content summary of each different paragraph of the document; when the knowledge content in the atomic granularity knowledge plug-in package is the content of 10 documents, the knowledge directory information can include the content summary of each of the 10 documents. In the embodiment of the present application, the knowledge directory information can also include position indication information, for example, page number information where the paragraph is located or page number information where the document is located. Taking the knowledge content in the atomic granularity knowledge plug-in package as the content of 10 documents as an example, the knowledge directory information can also include the starting page number information of each of the 10 documents.

[0041] In addition, the atomic granularity knowledge plug-in package can also include at least one of the name, description, citation description name, dependent plug-in list (ID list of other atomic granularity knowledge plug-in packages dependent on the atomic granularity knowledge plug-in package), and knowledge format of the atomic granularity knowledge plug-in. The name, description, and citation description name of the atomic granularity knowledge plug-in can facilitate the user to view the information of the atomic granularity knowledge plug-in package; the dependent plug-in list of the atomic granularity knowledge plug-in package can represent the relevance of the knowledge content in the atomic granularity knowledge plug-in package, and the atomic granularity knowledge plug-in package can be combined according to the relevance of the knowledge content in the atomic granularity knowledge plug-in package, so as to obtain a combined knowledge plug-in package. The knowledge format of the atomic granularity knowledge plug-in package can facilitate the storage of the knowledge content in the atomic granularity knowledge plug-in package in the database of the corresponding format in the target knowledge base.

[0042] After receiving the target request, the target device can obtain the knowledge data corresponding to the target request from the target knowledge plug-in package. Since the target knowledge plug-in package includes the atomic granularity knowledge plug-in package for knowledge data acquisition, the target device can obtain the knowledge data corresponding to the target request from the knowledge content included in the atomic granularity knowledge plug-in package. The knowledge data can be answer type data corresponding to the question type information included in the target request, for example, when the target request is a request related to the tutorial of configuring a certain service in the network management software, the knowledge data can be the tutorial content of configuring the certain service in the network management software. In addition, the format of the knowledge data in the embodiment of the present application can not be limited, for example, at least one of the vector format, the knowledge graph format, the text format, and the picture format. For example, if the target request is a request related to the knowledge graph of the gateway software, the format of the obtained knowledge data corresponding to the target request can be the knowledge graph format.

[0043] In the embodiments of the present application, the target knowledge plug-in package can exist in a target knowledge base (for example, an external knowledge base of a target model). The target knowledge base can be any database including knowledge data corresponding to a target request. Since the knowledge data can be in any format, accordingly, there can be a multi-modal database in the target knowledge base, for example, the target knowledge base can be a database including at least one of a vector database, a knowledge graph database, a text database, and a picture database, so as to provide knowledge data in a vector format, a knowledge graph format, and a text format. If the target knowledge base is a multi-modal database, in the process of obtaining knowledge data corresponding to the target request from the target knowledge base by the target device, data corresponding to the target request can be obtained from each modal database, and the data obtained from the multiple databases are collectively used as knowledge data.

[0044] In the embodiments of the present application, a plurality of atomic granularity knowledge plug-in packages can be combined to obtain a combined knowledge plug-in package. The target knowledge plug-in package can include the combined knowledge plug-in package. The target device can combine a plurality of atomic granularity knowledge plug-in packages to obtain a combined knowledge plug-in package in the form of code. Accordingly, the combination of different atomic granularity knowledge plug-in packages can be realized by modifying the code. For example, when the target request is used to request a tutorial for configuring a plurality of services in a network management software, the knowledge data can be the tutorial content of the tutorial for configuring a plurality of services in the network management software, one atomic granularity knowledge plug-in package can include knowledge data for configuring one service in the network management software, and the combination of a plurality of atomic granularity knowledge plug-in packages can obtain a combined knowledge plug-in package including the knowledge data. It should be noted that the atomic granularity knowledge plug-in package can be a reusable knowledge plug-in package and can be a knowledge plug-in package in the smallest unit. That is, one atomic granularity knowledge plug-in package can be used in the next combination operation after being used in one combination operation, so as to realize the reuse of the same atomic granularity knowledge plug-in package. In this way, different combined knowledge plug-in packages can be obtained by using different atomic granularity knowledge plug-in packages. The atomic granularity knowledge plug-in package can be directly imported into the target knowledge base as a whole.

[0045] It should be understood that since the target knowledge plug-in package can include a combined knowledge plug-in package, and the combined knowledge plug-in package also includes an atomic granularity knowledge plug-in package, the target knowledge plug-in package including the atomic granularity knowledge plug-in package in the embodiments of the present application can cover the following three cases: first, only the atomic granularity knowledge plug-in package exists in the target knowledge plug-in package; second, only the combined knowledge plug-in package exists in the target knowledge plug-in package; and third, both the atomic granularity knowledge plug-in package and the combined knowledge plug-in package exist in the target knowledge plug-in package.

[0046] In the embodiment of the present application, if the target knowledge base includes a multi-modal database, the knowledge data corresponding to the data format can be stored in the database of the corresponding mode according to the data format, so that the target device obtains the knowledge data corresponding to the target request from the target knowledge plug-in package of the target knowledge base. After the target device obtains the knowledge data, step 130 can be performed to feed back the knowledge data.

[0047] Step 130, feeding back the knowledge data.

[0048] In the embodiment of the present application, the target device can provide a feedback window, and the knowledge data obtained from the target knowledge base can be directly and completely fed back to the user through the feedback window. In addition, after obtaining the knowledge data, the target device can also feed back the knowledge data to another device for processing. In the embodiment of the present application, when the target device feeds back the knowledge data, the knowledge data can also be sorted according to the relevance to the target request, and the sorting result can be fed back and displayed. The embodiment of the present application does not limit this.

[0049] In the knowledge data acquisition method provided in the embodiment of the present application, a target request is received, and knowledge data corresponding to the target request is obtained from a target knowledge plug-in package. The knowledge data is fed back. The target knowledge plug-in package includes an atomic granularity knowledge plug-in package for knowledge data acquisition, and the atomic granularity knowledge plug-in package includes a plug-in identifier, knowledge directory information and knowledge content. In this way, since the target knowledge plug-in package includes an atomic granularity knowledge plug-in package for knowledge data acquisition, and the atomic granularity knowledge plug-in package includes a plug-in identifier, knowledge directory information and knowledge content, the knowledge data can be flexibly obtained from the atomic granularity knowledge plug-in package, thereby solving the problem of insufficient flexibility in acquiring knowledge data in the related art.

[0050] It should be noted that in an embodiment of the present application, the target knowledge plug-in package includes a combined knowledge plug-in package, the combined knowledge plug-in package is obtained by combining the atomic granularity knowledge plug-in packages, and the atomic granularity knowledge plug-in packages are reusable knowledge plug-in packages. Since the combined knowledge plug-in package in the target knowledge plug-in package is obtained by combining a plurality of atomic granularity knowledge plug-in packages, each of the plurality of atomic granularity knowledge plug-in packages is a reusable knowledge plug-in package, various combined knowledge plug-in packages can be obtained through these reusable atomic granularity knowledge plug-in packages, and the flexibility of adjusting the content of the knowledge base can be realized, thereby solving the problem of poor flexibility in adjusting the content of the knowledge base in the related art.

[0051] In the embodiments of the present application, the combined knowledge plug-in package can include a plug-in identifier, a list of included atomic granularity knowledge plug-in packages, and a query strategy. The plug-in identifier of the combined knowledge plug-in package can be unique identification information of the combined knowledge plug-in package, used to distinguish the combined knowledge plug-in package; the list of included atomic granularity knowledge plug-in packages can be an ID list of the atomic granularity knowledge plug-in packages included in the combined knowledge plug-in package, and knowledge data information can be obtained according to the ID information of the atomic granularity knowledge plug-in packages in the ID list; the query strategy can be a priority query strategy of the atomic granularity knowledge plug-in packages, and when the target device obtains knowledge data based on a target request, the knowledge data can be obtained according to the priority query strategy of the atomic granularity knowledge plug-in packages in the target knowledge plug-in package. Different query strategies can result in different weights of the same atomic granularity knowledge plug-in package, so that the same atomic granularity knowledge plug-in package has different priorities in different application scenarios. In this way, the association between the knowledge data and the target request can be stronger, and the accuracy of the knowledge data can be improved.

[0052] In addition, the combined knowledge plug-in package can also include the name, description, and citation description name of the combined knowledge plug-in package. In the embodiments of the present application, the knowledge content can be stored and used in the form of a plug-in package. For example, the knowledge content can be encapsulated as an atomic granularity knowledge plug-in package and / or a combined knowledge plug-in package.

[0053] In the embodiments of the present application, the target knowledge plug-in package can be a plug-in package in an external knowledge base of a target model. The target model can be any machine learning model capable of providing feedback according to a request, such as a large language model (LLM). The target device can feed back answer content corresponding to the target request through the target model according to the target request. Accordingly, the external knowledge base can provide data support in the process of the target model answering the target request. For example, after the target model receives the target request, the target device obtains knowledge data corresponding to the target request in the external knowledge base (for example, step 120). After obtaining the knowledge data, the target device feeds back the knowledge data to the target model. The target model can directly feed back the knowledge data as answer content to the user, or the target model can further process the knowledge data, and finally obtain answer content corresponding to the target request to feed back to the user.

[0054] In the embodiments of the present application, in order to make the answer content provided by the target model more accurate, the application scenario of the target model can be limited, for example, the target model answering the communication field is different from the target model answering the education field. Correspondingly, the target knowledge plug-in package can be a plug-in package matching the application scenario of the target model, and the content contained in the target knowledge plug-in package changes with the change of the application scenario of the target model.

[0055] The type of the knowledge plug-in package can be determined by the business range size of the application scenario, that is, whether it is an atomic granularity knowledge plug-in package or a combined knowledge plug-in package. In one application scenario, an atomic granularity knowledge plug-in package can be constructed based on the A knowledge content, and in another application scenario, a combined knowledge plug-in package can be constructed based on the A knowledge content. That is, the same knowledge content can construct an atomic granularity knowledge plug-in package in one application scenario, and can also construct a combined knowledge plug-in package composed of multiple atomic granularity knowledge plug-in packages in another application scenario.

[0056] If the application scenario of the target model is the network management knowledge question and answer scenario, the atomic granularity knowledge plug-in package can be the knowledge plug-in package of network management help and the knowledge plug-in package of field fault handling. Further, the two atomic granularity knowledge plug-in packages are combined to obtain a combined knowledge plug-in package, which can be a network management knowledge question and answer combined plug-in package satisfying the application scenario, that is, a target knowledge plug-in package matching the network management knowledge question and answer scenario. If the application scenario of the target model is the network management help knowledge question and answer scenario, the atomic granularity knowledge plug-in package is the knowledge plug-in package of network management page help and the knowledge plug-in package of network management business help. The two atomic granularity knowledge plug-in packages are combined to obtain a combined knowledge plug-in package, that is, a target knowledge plug-in package matching the network management help knowledge question and answer scenario.

[0057] Since the target knowledge plug-in package is a plug-in package matching the application scenario of the target model, the content contained in the target knowledge plug-in package can change with the change of the application scenario of the target model. The content in the target knowledge plug-in package can be more matched with the application scenario of the target model through flexible combination of atomic granularity knowledge plug-in packages, and the accuracy of the target model is further improved. For example, the target knowledge plug-in package in one scenario includes a combined knowledge plug-in package (composed of atomic granularity knowledge plug-in package a, atomic granularity knowledge plug-in package b and atomic granularity knowledge plug-in package c), and the target knowledge plug-in package in another scenario includes another combined knowledge plug-in package (composed of atomic granularity knowledge plug-in package a, atomic granularity knowledge plug-in package b and atomic granularity knowledge plug-in package d).

[0058] In the embodiments of the present application, in addition to the target model and the target knowledge base, a finished product library for storing atomic granularity knowledge plug-in packages and combined knowledge plug-in packages can also exist on the target device. The content of the plug-in packages in the finished product library is not allowed to be modified in the running period of the external knowledge base. The target knowledge plug-in package can be a plug-in package from the finished product library, and the content in the target knowledge plug-in package is not allowed to be modified in the running period of the external knowledge base, so as to ensure the stability of the target knowledge plug-in package and limit the data source in the external knowledge base, thereby improving the stability of the external knowledge base.

[0059] In the embodiments of the present application, in order to further improve the stability and security of the target knowledge base (for example, the external knowledge base of the LLM) of the target model, the atomic granularity knowledge plug-in package and the combined knowledge plug-in package can also include unique identity verification information, and each plug-in package in the target knowledge plug-in package is a plug-in package whose identity verification information passes verification. Compared with the related art of importing any document into the external knowledge base, which affects the stability of the external knowledge base, the embodiments of the present application can improve the stability and security of the external knowledge base by verifying the identity verification information of the plug-in package.

[0060] FIG. 2 is a flowchart of another knowledge data acquisition method provided by the embodiments of the present application. As shown in FIG. 2, the data acquisition method provided by the embodiments of the present application includes the following steps.

[0061] Step 210, constructing M atomic granularity knowledge plug-in packages.

[0062] In the embodiments of the present application, the knowledge content to be processed can be processed by a corpus pipeline to perform data cleaning, slicing and the like, and then the knowledge content to be processed with high correlation and being the smallest reusable unit is compressed and packaged with the description information (for example, the plug-in identifier and the knowledge directory information described above) of the atomic granularity knowledge plug-in package according to the correlation between the knowledge content to be processed, so as to complete the construction of an atomic granularity knowledge plug-in, and the process is repeated to obtain M atomic granularity knowledge plug-in packages, such as a help manual atomic granularity knowledge plug-in package, a field fault handling atomic granularity knowledge plug-in package and the like. In this way, the knowledge content can be refined to the smallest reusable unit, facilitating subsequent flexible combination and configuration.

[0063] Step 220, constructing N combined knowledge plug-in packages based on the M atomic granularity knowledge plug-in packages.

[0064] After obtaining the M atomic granularity knowledge plug-in packages, a plurality of plug-in packages in the M atomic granularity knowledge plug-in packages can be combined to construct N combined knowledge plug-in packages, and the combined knowledge plug-in packages at least include two atomic granularity knowledge plug-in packages. Finally, based on the N combined knowledge plug-in packages, a target knowledge plug-in package is obtained, and the target knowledge plug-in package is used for subsequent acquisition of knowledge data. In the embodiment of the present application, by processing the knowledge content into atomic granularity knowledge plug-in packages and combined knowledge plug-in packages, the knowledge content can be flexibly combined, and then the flexible adjustment of the knowledge base content can be realized.

[0065] In step 230, based on at least part of the M atomic granularity knowledge plug-in packages and the N combined knowledge plug-in packages, a target knowledge plug-in package is obtained.

[0066] In the embodiment of the present application, at least part of the M atomic granularity knowledge plug-in packages and the N combined knowledge plug-in packages can be atomic granularity knowledge plug-in packages, or can be combined knowledge plug-in packages, or can contain both atomic granularity knowledge plug-in packages and combined knowledge plug-in packages.

[0067] Before the target device receives a target request, steps 210-230 can be performed to obtain a target knowledge plug-in package.

[0068] In step 240, a target request is received.

[0069] In step 250, knowledge data corresponding to the target request is acquired from the target knowledge plug-in package; wherein the target knowledge plug-in package includes an atomic granularity knowledge plug-in package for knowledge data acquisition, and the atomic granularity knowledge plug-in package includes a plug-in identifier, knowledge directory information and knowledge content.

[0070] In the embodiment of the present application, there can also be a combined knowledge plug-in package in the target knowledge plug-in package, and the combined knowledge plug-in package is a plug-in package obtained by combining a plurality of atomic granularity knowledge plug-in packages. The atomic granularity knowledge plug-in package can be the smallest unit of reusable knowledge plug-in package. The target knowledge plug-in package can be located in the external knowledge base of the target model.

[0071] In step 260, the knowledge data is fed back.

[0072] In step 230 provided in the embodiments of the present application, after obtaining the N combined knowledge plug-in packages, the application scenario of the target model can be acquired; K combined knowledge plug-in packages matching the application scenario can be acquired from the N combined knowledge plug-in packages, and the K combined knowledge plug-in packages are determined as the target knowledge plug-in packages; wherein K is a positive integer less than or equal to N, for example, K is 1. The target knowledge plug-in packages are acquired according to the application scenario of the target model, so that the target knowledge plug-in packages are more matched with the application scenario of the target model, and further, the target knowledge base (external knowledge base) provides more accurate knowledge data to the target model.

[0073] In the embodiments of the present application, the M atomic granularity knowledge plug-in packages and the N combined knowledge plug-in packages are stored in the finished product warehouse in the form of finished products. That is, when the atomic granularity knowledge plug-in packages and the combined knowledge plug-in packages are constructed, they can be put into the finished product warehouse in the form of finished products, so as to obtain the target knowledge plug-in packages subsequently. In an example, after the application scenario of the target model is acquired, the target knowledge plug-in packages matching the application scenario can be acquired from the finished product warehouse.

[0074] In order to better provide knowledge data, the data acquisition method provided in the embodiments of the present application can also update the target knowledge base. FIG. 3 is a flowchart of another knowledge data acquisition method provided in the embodiments of the present application. As shown in FIG. 3, the data acquisition method provided in the embodiments of the present application includes the following steps.

[0075] Step 310, receiving a target request.

[0076] Step 320, acquiring knowledge data corresponding to the target request from a target knowledge plug-in package in an external knowledge base of a target model; wherein the target knowledge plug-in package includes an atomic granularity knowledge plug-in package for knowledge data acquisition, and the atomic granularity knowledge plug-in package includes a plug-in identifier, knowledge directory information and knowledge content.

[0077] Step 330, feeding back the knowledge data.

[0078] Step 340, performing an update operation on the target knowledge plug-in package to obtain an updated target knowledge plug-in package; wherein the updated target knowledge plug-in package is used for subsequent acquisition of knowledge data.

[0079] In the embodiments of the present application, after the knowledge data is fed back, step 340 can be performed to perform an update operation on the target knowledge plug-in package to obtain an updated target knowledge plug-in package; wherein the updated target knowledge plug-in package is used for subsequent acquisition of knowledge data. The update operation can be updated during the running period of the external knowledge base to provide more real-time knowledge data.

[0080] In the embodiments of the present application, the target device can provide a graphical user interface (GUI), and receive target information input by a target user in the graphical user interface through the GUI, the target information being used to indicate a plug-in package to be operated, and based on the target information, the plug-in package to be operated is determined. The plug-in package to be operated can be used to update the target knowledge plug-in package. For example, the user confirms that an atomic granularity knowledge plug-in package is deleted through the GUI, and the atomic granularity knowledge plug-in package can be the plug-in package to be operated. Alternatively, the user imports a combined knowledge plug-in package through the GUI, and the combined knowledge plug-in package can be the plug-in package to be operated. The target device can complete the update of the target knowledge base by receiving the target information input by the user, and the update of the external knowledge base and the update of the target model are decoupled, so that the update of the external knowledge base can meet the rapid update of knowledge data.

[0081] In the embodiments of the present application, the target device can also determine the plug-in package to be operated by running code. For example, the target device obtains an update request through an application programming interface (API), and thus obtains the plug-in package to be operated.

[0082] In the embodiments of the present application, the update operation can include deletion, addition and replacement. In an example, the plug-in package to be operated is determined through the above steps, and the plug-in package to be operated is a plug-in package to be deleted in the target knowledge plug-in package, a plug-in package to be replaced in the target knowledge plug-in package, or a plug-in package to be added to the target knowledge plug-in package. The target device can determine the plug-in package to be operated by using unique identification information (ID) of the plug-in package, and can also determine the plug-in package to be operated by using Chinese description information of the plug-in package.

[0083] In the case that the plug-in package to be operated is a plug-in package to be deleted in the target knowledge plug-in package, the plug-in package to be operated in the target knowledge plug-in package is deleted.

[0084] In the case that the plug-in package to be operated is a plug-in package to be replaced in the target knowledge plug-in package, a replacement package of the plug-in package to be operated is obtained, the replacement package being a finished product from a finished product library and not being allowed to be modified during the running period of the external knowledge base, and the plug-in package to be operated in the target knowledge plug-in package is replaced by using the replacement package.

[0085] In the embodiments of the present application, before the replacement package is used to replace the to-be-operated plug-in package in the target knowledge plug-in package, the identity verification information of the replacement package can be verified. If the verification is passed, the replacement package is used to replace the to-be-operated plug-in package in the target knowledge plug-in package.

[0086] In the embodiments of the present application, if the target device acquires the replacement package through the API, the replacement package can come from an update package including update data of the target model.

[0087] Correspondingly, in the case that the to-be-operated plug-in package is a plug-in package to be added to the target knowledge plug-in package, the to-be-operated plug-in package is a finished product from a finished product library and is not allowed to be modified in the running period of the external knowledge library. Before the to-be-operated plug-in package is added to the target knowledge plug-in package, the to-be-operated plug-in package can be verified. If the verification is passed, the to-be-operated plug-in package is added. If the target device acquires the to-be-operated plug-in package through the API, the to-be-operated plug-in package can come from an update package including update data of the target model. Through deletion, addition and replacement of plug-in packages in the target knowledge plug-in package, update of knowledge data in the target knowledge library can be realized, and knowledge data with more real-time and accuracy can be provided.

[0088] In the embodiments of the present application, the combined knowledge plug-in package can be a combined knowledge plug-in package based on application program (APP) granularity. One business in an application program can correspond to one atomic granularity knowledge plug-in package. In the embodiments of the present application, atomic granularity knowledge plug-in packages can be combined into a combined knowledge plug-in package meeting specific application requirements according to actual scenarios of users and product requirements of application programs. For example, the application program is a network management knowledge question and answer software, and then the atomic granularity network management help knowledge plug-in and the field fault handling FAQ knowledge plug-in can be included. In this way, the atomic granularity knowledge plug-in package is combined into an APP granularity for use, which can increase the business range of the APP. Correspondingly, the application scenarios of the target model can also match the business range of the application program, and the application scenarios of the target model are increased.

[0089] In this embodiment, the target knowledge plugin package can be imported along with the installation of the target model, and the target knowledge plugin package can be updated along with the update of the target model. In this embodiment, when installing the target model, it can be installed using the target model's installation package. The installation package can include not only the target model's data but also the target knowledge plugin package from the finished product library, which can then be imported together. Furthermore, during the update of the target model, it can be updated using the update package mentioned above, which includes the updated data of the target model. This update package includes the plugin package to be operated on, and the target knowledge plugin package is updated using the plugin package to be operated on. This allows the update of the target knowledge base provided in this embodiment to not only be decoupled from the update of the target model but also to achieve a holistic update.

[0090] In this embodiment, the target model is a large language model, the target request includes a target question, and the knowledge data includes the answer to the target question. The large language model receives the target question, retrieves the corresponding knowledge data from an external knowledge base, learns from the knowledge data, and then outputs the target answer corresponding to the target question, thus realizing a knowledge-based question-and-answer scenario using the large language model.

[0091] Figure 4 is a flowchart illustrating a knowledge data acquisition method provided in an embodiment of this application. As shown in Figure 4, the data acquisition method provided in this embodiment includes the following steps.

[0092] Step 405: Construct M atomic-level knowledge plugin packages.

[0093] Step 410: Based on the M atomic-level knowledge plugin packages, construct N combined knowledge plugin packages. Each of the M atomic-level knowledge plugin packages is a reusable knowledge plugin package.

[0094] Step 415: Obtain the application scenario of the target model.

[0095] Step 420: Obtain K combined knowledge plugin packages that match the application scenario from the N combined knowledge plugin packages, where K is a positive integer less than or equal to N.

[0096] Step 425: Determine the K combined knowledge plugin packages as the target knowledge plugin packages.

[0097] Step 430: Receive the target request.

[0098] Step 435: Obtain knowledge data corresponding to the target request from the target knowledge plugin package; wherein, the target knowledge plugin package includes an atomic-level knowledge plugin package for knowledge data acquisition, and the atomic-level knowledge plugin package includes a plugin identifier, knowledge directory information, and knowledge content.

[0099] Step 440: Feedback the knowledge data.

[0100] Step 445: If it is necessary to perform an update operation on the target knowledge plugin package, determine the plugin package to be operated on. The plugin package to be operated on is a plugin package to be deleted from the target knowledge plugin package, a plugin package to be replaced from the target knowledge plugin package, or a plugin package to be added to the target knowledge plugin package.

[0101] Step 450: If the plugin package to be operated on is a plugin package to be deleted in the target knowledge plugin package, delete the plugin package to be operated on in the target knowledge plugin package.

[0102] Step 455: If the plugin package to be operated is the plugin package to be replaced in the target knowledge plugin package, obtain the replacement package of the plugin package to be operated, wherein the replacement package is a finished product that cannot be modified during the runtime of the plug-in knowledge base; use the replacement package to replace the plugin package to be operated in the target knowledge plugin package.

[0103] Step 460: If the plugin package to be operated is a plugin package to be added to the target knowledge plugin package, obtain the plugin package to be operated, wherein the plugin package to be operated is a finished product that cannot be modified during the runtime of the external knowledge base; add the plugin package to be operated to the target knowledge plugin package.

[0104] Step 465: Obtain the updated target knowledge plugin package; wherein the updated target knowledge plugin package is used for subsequent acquisition of knowledge data.

[0105] In the knowledge data acquisition method provided in this application embodiment, since the target knowledge plugin package includes atomic-level knowledge plugin packages for knowledge data acquisition, and each atomic-level knowledge plugin package includes a plugin identifier, knowledge directory information, and knowledge content, knowledge data can be flexibly acquired from the atomic-level knowledge plugin packages, solving the problem of insufficient flexibility in acquiring knowledge data in related technologies. Furthermore, since the combined knowledge plugin packages within the target knowledge base are plugin packages obtained by combining multiple atomic-level knowledge plugin packages, each of these multiple atomic-level knowledge plugin packages is a reusable knowledge plugin package. Various combined knowledge plugin packages can be obtained through these reusable atomic-level knowledge plugin packages, thereby enabling flexible adjustment of the knowledge base content and solving the problem of poor flexibility in adjusting knowledge base content in related technologies.

[0106] Figure 5 is a flowchart illustrating a knowledge base construction method provided in an embodiment of this application. As shown in Figure 5, the knowledge base construction method provided in this embodiment includes the following steps.

[0107] Step 510: Obtain knowledge corpus for the target application scenario.

[0108] Step 520: Process the knowledge prediction to obtain P atomic-level knowledge plugin packages. Each of the P atomic-level knowledge plugin packages includes a plugin identifier, knowledge directory information, and the knowledge content itself. Furthermore, each of the P atomic-level knowledge plugin packages can be a reusable knowledge plugin package for the target application scenario.

[0109] Step 530: Based on the P atomic-level knowledge plugin packages, obtain the knowledge base; P is a positive integer.

[0110] It should be understood that the knowledge base obtained in step 530 may contain only atomic-level knowledge plugin packages, only combined knowledge plugin packages, or both atomic-level and combined knowledge plugin packages.

[0111] In one embodiment, step 530, which involves obtaining a knowledge base based on the P atomic-level knowledge plugin packages, includes: selecting at least one atomic-level knowledge plugin package from the P atomic-level knowledge plugin packages and adding it to the knowledge base, thereby obtaining the knowledge base. In this embodiment, the knowledge base may contain only atomic-level knowledge plugin packages.

[0112] In another embodiment, step 530, obtaining a knowledge base based on the P atomic-level knowledge plugin packages, includes: combining Q atomic-level knowledge plugin packages from the P atomic-level knowledge plugin packages to obtain a combined knowledge plugin package; where Q is a positive integer and Q is less than or equal to P; and obtaining a knowledge base based on the combined knowledge plugin package. The process of obtaining the knowledge base based on the combined knowledge plugin package may include: placing the combined knowledge plugin package into the knowledge base to obtain the knowledge base; that is, in this case, the knowledge base may only contain the combined knowledge plugin package. In the process of obtaining the knowledge base based on the combined knowledge plugin package, in addition to placing the combined knowledge plugin package into the knowledge base, at least one atomic-level knowledge plugin package from the P atomic-level knowledge plugin packages can also be placed into the knowledge base. In this way, the knowledge base can simultaneously contain both atomic-level knowledge plugin packages and combined knowledge plugin packages.

[0113] As can be seen from the above analysis, the embodiments of this application can obtain a knowledge base based on the P atomic-level knowledge plugin packages and at least some of the plugin packages in the combined knowledge plugin package.

[0114] In this embodiment of the application, the target device for executing the knowledge base construction method and the target device for executing the data acquisition method can be the same target device or different devices. This embodiment of the application does not limit this.

[0115] In this embodiment, the target application scenario can be any scenario in any field, such as the communications field or the education field. The knowledge corpus of the target application scenario can be an audited knowledge corpus to ensure the accuracy of the knowledge data. After obtaining the knowledge corpus of the target application scenario, the knowledge corpus can be processed by data cleaning, segmentation, and formatting. Then, the knowledge corpus data of reusable units within the target application scenario (such as network management help knowledge data in a network management Q&A scenario) can be compressed, packaged, and obtained P atomic-level knowledge plugin packages. Each of the P atomic-level knowledge plugin packages can be a reusable knowledge plugin package for the target application scenario. Based on the P atomic-level knowledge plugin packages, a knowledge base is obtained. Through the atomic-level knowledge plugin packages, knowledge content can be flexibly combined to achieve flexible changes to the knowledge base content. The constructed knowledge base can be used as an external knowledge base for the target model in the data acquisition method provided in the above embodiments, or it can be applied to other application scenarios related to the knowledge base.

[0116] In this embodiment, the knowledge base can be updated during its runtime to obtain an updated knowledge base, making the knowledge data in the knowledge base more real-time and accurate. Updating the knowledge base can be done by updating the atomic-level knowledge plugin packages (e.g., deleting or replacing a specific atomic-level knowledge plugin package), or by updating the combined knowledge plugin packages in the knowledge base (e.g., deleting some of the atomic-level knowledge plugin packages).

[0117] To better understand the knowledge base construction method and knowledge data acquisition method provided in the embodiments of this application, further discussion is provided below. It should be understood that the following discussion is merely illustrative and not limiting.

[0118] This application provides a method for configuring and managing knowledge plugins in a large-scale model application. As shown in Figures 6-1 and 6-2, the method for configuring and managing knowledge plugins in a large-scale model application provided in this application may involve a knowledge plugin modeling module, a knowledge plugin management module, a software installation and deployment module, a data storage module, and a large-scale model application management module. Details of the modules are shown in Table 1.

[0119] Table 1

[0120] Referring to Figures 6-1, 6-2, 7, and 8, the method for knowledge plugin configuration management of large model applications provided in this application embodiment may include the following steps.

[0121] Step 1: Define the knowledge plugin metamodel: This model comprehensively describes all attributes of the knowledge plugin, including plugin description, knowledge file storage format, knowledge dependencies, and knowledge retrieval strategies, providing a basic framework for subsequent knowledge plugin configuration. This step may involve defining atomic-level knowledge plugin packages and combined knowledge plugin packages. Specifically: Atomic-level knowledge plugin package: The smallest unit of a knowledge plugin, such as a knowledge plugin for network management support or a FAQ knowledge plugin for field fault handling. The configuration information for the atomic-level knowledge plugin package is shown in Table 2.

[0122] Table 2

[0123] Combined Knowledge Plugin Package: A plugin package obtained by combining atomic-level knowledge plugin packages according to the external business needs of the product, such as a network management knowledge Q&A combined plugin package. The configuration information of the combined knowledge plugin package is shown in Table 3.

[0124] Table 3

[0125] Step Two: Constructing Atomic-Grained Knowledge Plugin Packages. Referring to Figure 6-1, atomic-grained knowledge plugin packages and / or combined knowledge plugin packages can be constructed using the knowledge plugin construction module. The constructed atomic-grained or combined knowledge plugin packages can be placed in the finished product library. This allows for the subsequent retrieval of required plugin packages from the finished product library. The process of constructing atomic-grained knowledge plugin packages can be seen in Figure 7. The corpus pipeline shown in Figure 7 can package knowledge content and its meta-model description information into atomic-grained knowledge plugins. This design allows knowledge content to be refined to the smallest reusable unit, facilitating flexible combination and configuration later. Examples include: atomic-grained knowledge plugin packages for help manuals and atomic-grained knowledge plugin packages for field fault handling. The configuration information for the atomic-grained knowledge plugin packages is shown in Table 2 above.

[0126] Step 3: Construct an app-level knowledge combination plugin package. Based on actual user scenarios and product needs, a plugin package pipeline combines atomic-level knowledge plugins into larger-grained knowledge combination plugin packages that meet specific application requirements, simplifying plugin release and updates. For example, a knowledge combination plugin package for Q&A customer service (including an atomic-level knowledge plugin package for help manuals and an atomic-level knowledge plugin package for field fault handling). This includes defining query strategy algorithms and optimizing multi-database query strategies, such as an atomic-level knowledge base priority query strategy. The configuration information for the combined knowledge plugin package is shown in Table 3 above.

[0127] Step 4: Providing GUI and API Interfaces: To facilitate user operation and system integration, this application embodiment provides both a graphical user interface (GUI) and an application programming interface (API) to enable the addition, deletion, and querying of knowledge plugins. Users can configure plugins online through the GUI, while the system upgrade component can automatically import and update plugins via the API. Referring to Figure 6-2, combined knowledge plugin packages can be imported during installation and upgrade processes, and users can also initiate the addition and updating of combined knowledge plugin packages through the GUI according to their needs.

[0128] There are two ways to import plugins: First, the granular knowledge plugin package is automatically imported with the model software installation, and then manually imported and updated. This method is convenient and ensures the consistency of the atomic-level knowledge plugin package. Second, the atomic-level knowledge plugin package can be automatically imported with the model installation, and can be manually imported and updated later. Users need to pay attention to the atomic-level knowledge plugin package. When it is necessary to select the corresponding atomic-level knowledge plugin, a scenario for importing and updating the atomic-level knowledge plugin package individually can be provided through the GUI.

[0129] Step 5: The knowledge plugin management module saves corpora in different formats to the data storage module (i.e., the external knowledge base for large model applications) according to the description files of the knowledge plugins. Refer to Figures 6-2 and 8. As shown in Figure 8, embedded data is saved to the vector database, and graph data is saved to the graph database.

[0130] By following the steps above, the knowledge of the plug-in can be updated flexibly.

[0131] Step Six: When a user submits a question request to the large model application, the large model application management module performs knowledge retrieval based on the user's request, using imported data and query strategies, as shown in Figure 6-2.

[0132] It should be noted that the numbers in Figures 6-1 and 6-2 represent the order in which the steps are executed, but this is only an example and not a limitation.

[0133] This application primarily addresses the challenges faced by large-scale model applications when using external knowledge bases, namely the mismatch between knowledge update speed and software version release speed, and the difficulty in flexibly combining different knowledge content. Through this application, rapid knowledge updates, flexible configuration, and combination can be achieved, thereby significantly improving the knowledge processing efficiency and flexibility of large-scale model applications.

[0134] In large-scale model applications, knowledge is the core content. This application's embodiment allows the plugin itself to be flexibly released as an independently distributed product. Due to the high value of the plugin package, its sale can be controlled through a license. Releasing it as a tested, finished knowledge product to enterprise users offers greater stability and efficiency compared to online knowledge import. This application's embodiment effectively solves the problem of how to securely and efficiently import custom corpora.

[0135] Referring to Figure 9, this application embodiment also provides a data acquisition device 900, including a receiving module 910, an acquisition module 920, and a feedback module 930. The receiving module 910 is used to receive a target request; the acquisition module 920 is used to acquire knowledge data corresponding to the target request from a target knowledge plugin package; the feedback module 930 is used to provide feedback on the knowledge data; wherein the target knowledge plugin package includes an atomic-level knowledge plugin package for knowledge data acquisition, and the atomic-level knowledge plugin package includes a plugin identifier, knowledge directory information, and knowledge content.

[0136] In the data acquisition device provided in this application embodiment, since the target knowledge plugin package includes an atomic-level knowledge plugin package for knowledge data acquisition, and the atomic-level knowledge plugin package includes a plugin identifier, knowledge directory information and knowledge content, knowledge data can be flexibly acquired from the atomic-level knowledge plugin package, thus solving the problem of insufficient flexibility in acquiring knowledge data in related technologies.

[0137] It should be noted that the embodiments of the data acquisition device in this specification and the embodiments of the knowledge data acquisition method in this specification are based on the same inventive concept. Therefore, the implementation of this embodiment can refer to the implementation of the corresponding data acquisition method above, and the repeated parts will not be described again.

[0138] Referring to Figure 10, this embodiment of the application also provides a knowledge base construction apparatus 1000, including an acquisition module 1010 and a processing module 1020. The acquisition module 1010 is used to: acquire knowledge corpus for a target application scenario; the processing module 1020 is used to: process the knowledge corpus to obtain P atomic-level knowledge plugin packages, each of the P atomic-level knowledge plugin packages including a plugin identifier, knowledge directory information, and the knowledge content itself; and obtain a knowledge base based on the P atomic-level knowledge plugin packages; where P is a positive integer.

[0139] In the knowledge base construction apparatus provided in this application embodiment, since the target knowledge plugin package includes an atomic-level knowledge plugin package for knowledge data acquisition, and the atomic-level knowledge plugin package includes a plugin identifier, knowledge directory information and knowledge content, knowledge data can be flexibly acquired from the atomic-level knowledge plugin package, thus solving the problem of insufficient flexibility in acquiring knowledge data in related technologies.

[0140] It should be noted that the embodiments of the knowledge base construction device in this specification and the embodiments of the knowledge base construction method in this specification are based on the same inventive concept. Therefore, the implementation of this embodiment can refer to the implementation of the corresponding knowledge base construction method mentioned above, and the repeated parts will not be described again.

[0141] Figure 11 is a schematic diagram of an electronic device provided in an embodiment of this application. As shown in Figure 11, the electronic device 1100 provided in this embodiment may include a processor 1110 and a memory 1120. The memory 1120 stores a computer program, which, when executed, implements the steps in the data acquisition method (e.g., the data acquisition method shown in any of Figures 1 to 4) or the knowledge base construction method (e.g., the knowledge base construction method shown in Figure 5) provided in this embodiment.

[0142] Memory is used to store programs or data. Memory may be, but is not limited to, Random Access Memory (RAM), Read Only Memory (ROM), Programmable Read-Only Memory (PROM), Erasable Programmable Read-Only Memory (EPROM), Electrically Erasable Programmable Read-Only Memory (EEPROM), etc.

[0143] This application also provides a computer-readable storage medium storing a program or instructions that, when executed by a processor, implement the various processes of the above-described caching method embodiments based on a content delivery network and achieve the same technical effect. To avoid repetition, these will not be described again here.

[0144] This application also provides a computer program product, which is stored in a storage medium and executed by at least one processor to implement the various processes of the above method embodiments and achieve the same technical effect. To avoid repetition, it will not be described again here.

[0145] The processor is the processor in the electronic device described in the above embodiments. The readable storage medium includes computer-readable storage media, such as computer read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk.

[0146] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element. Furthermore, it should be noted that the scope of the methods and apparatuses in the embodiments of this application is not limited to performing functions in the order shown or discussed, but may also include performing functions substantially simultaneously or in the reverse order, depending on the functions involved. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.

[0147] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a computer software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0148] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit and scope of the claims, and all of these forms are within the protection scope of this application.

Claims

A knowledge data acquisition method, comprising: Receive target request; Obtain the knowledge data corresponding to the target request from the target knowledge plugin package; Feedback of the aforementioned knowledge data; The target knowledge plugin package includes an atomic-level knowledge plugin package for knowledge data acquisition, wherein the atomic-level knowledge plugin package includes a plugin identifier, knowledge directory information, and knowledge content. According to the method of claim 1, wherein, The target knowledge plugin package is a plugin package in the external knowledge base of the target model, and the target knowledge plugin package matches the application scenario of the target model. The method according to claim 2, wherein, The target knowledge plugin package is a plugin package originating from the finished product library. The content of the plugin package in the finished product library cannot be modified during the runtime of the external knowledge base. The method according to any one of claims 1-3, wherein, The target knowledge plugin package contains a combined knowledge plugin package, which is obtained by combining the atomic granularity knowledge plugin packages. The combined knowledge plugin package includes a plugin identifier, a list of the included atomic granularity knowledge plugin packages, and a query strategy. The method according to claim 4, wherein, The atomic-level knowledge plugin package is a reusable knowledge plugin package. According to the method of claim 1, wherein, The atomic-level knowledge plugin package contains identity verification information, and each plugin package in the target knowledge plugin package is a plugin package whose identity verification information has been verified. The method according to claim 2, wherein, Prior to receiving the target request, the method further includes: Construct M atomic-level knowledge plugin packages; Based on the M atomic-level knowledge plugin packages, construct N combined knowledge plugin packages; Based on at least a portion of the M atomic-level knowledge plugin packages and the N combined knowledge plugin packages, the target knowledge plugin package is obtained. Where M and N are both positive integers. The method according to claim 7, wherein, The process of obtaining the target knowledge plugin package based on at least a portion of the M atomic-level knowledge plugin packages and the N combined knowledge plugin packages includes: Obtain the application scenarios of the target model; Obtain K combined knowledge plugin packages that match the application scenario from the N combined knowledge plugin packages, and The K combined knowledge plugin packages are identified as the target knowledge plugin packages; Where K is a positive integer less than or equal to N. The method according to claim 2, wherein, After feeding back the knowledge data, the method further includes: An update operation is performed on the target knowledge plugin package to obtain an updated target knowledge plugin package; The updated target knowledge plugin package is used for subsequent acquisition of knowledge data. The method according to claim 9, wherein, The update operation is performed during the runtime of the add-on knowledge base. The method according to claim 9 or 10, wherein, The update operation on the target knowledge plugin package includes: The plugin package to be operated on is determined, which is a plugin package to be deleted from the target knowledge plugin package, a plugin package to be replaced in the target knowledge plugin package, or a plugin package to be added to the target knowledge plugin package. If the plugin package to be operated on is a plugin package to be deleted in the target knowledge plugin package, then the plugin package to be operated on in the target knowledge plugin package is deleted. If the plugin package to be operated is the plugin package to be replaced in the target knowledge plugin package, obtain the replacement package of the plugin package to be operated, the replacement package is a finished product from the finished product library and cannot be modified during the runtime of the plug-in knowledge library; use the replacement package to replace the plugin package to be operated in the target knowledge plugin package; If the plugin package to be operated is a plugin package to be added to the target knowledge plugin package, obtain the plugin package to be operated, wherein the plugin package to be operated is a finished product originating from the finished product library and which is not allowed to be modified during the runtime of the external knowledge library; add the plugin package to be operated to the target knowledge plugin package. The method according to claim 11, wherein, The process of determining the plugin package to be operated on includes: Receive target information input by the target user in the graphical user interface, the target information being used to indicate the plug-in package to be operated; Based on the target information, the plugin package to be operated is determined. A knowledge base construction method, comprising: Acquire knowledge corpora specific to the target application scenario; The knowledge corpus is processed to obtain P atomic-level knowledge plugin packages, each of which includes a plugin identifier, knowledge directory information, and the knowledge content itself. Based on the P atomic-level knowledge plugin packages, a knowledge base is obtained; P is a positive integer. The method according to claim 13, wherein, The knowledge base obtained based on the P atomic-level knowledge plugin packages includes: The Q atomic granularity knowledge plugin packages from the P atomic granularity knowledge plugin packages are combined to obtain a combined knowledge plugin package; where Q is a positive integer and Q is less than or equal to P; A knowledge base is obtained based on the aforementioned combined knowledge plugin package. The method according to claim 13, wherein, The method further includes: During the runtime of the knowledge base, the knowledge base is updated to obtain an updated knowledge base. An electronic device includes a processor and a memory, the memory storing a program or instructions that run on the processor, the program or instructions, when executed by the processor, implement the steps of the method as described in any one of claims 1-15. A computer-readable storage medium storing a program or instructions that, when executed, implement the steps of the method as described in any one of claims 1-15. A computer program product comprising a computer program that, when executed by a processor, implements the steps of the method according to any one of claims 1-15.

Images