Intention Processing Method, Device, Electronic Device, and Readable Storage Medium

Abstract

The present application provides a method, apparatus, electronic device, and readable storage medium for intent processing. The method includes: analyzing the obtained buried point data and the initial user intent to determine the user intent to be processed, where the initial user intent is used to represent the original needs of the user; determining a configuration scheme for the intent network according to the user intent to be processed. By analyzing the obtained buried point data and the initial user intent to determine the user intent to be processed, the input information of the user is simplified, the deficiencies of the initial user intent can be compensated by the buried point data, the understanding of the user intent can be expanded, and the ambiguity of the user intent can be eliminated; according to the user intent to be processed, determining a configuration scheme for the intent network, establishing an association relationship between the user intent to be processed and the configuration information of the intent network, and improving the accuracy of intent translation in the intent network.

Description

Technical Field

[0001] This application relates to the field of communication network technologies, and particularly to an intent processing method, apparatus, electronic device, and readable storage medium. Background Art

[0002] An Intent-Based Network (IBN) is a closed-loop network architecture that is built and operated based on human service intents under the condition of mastering its own "holographic state", realizing the automatic conversion from user intents to specific infrastructure, and being able to monitor the overall performance of the network, identify problems occurring in the network, and automatically solve the problems without manual intervention. IBN includes functions such as intent translation and verification, automatic implementation, perception of network status, reliability, and dynamic optimization / repair. Among them, intent translation realizes the conversion of the intent expressed in the natural language of the user (User says) into an intent recognizable by the network, which is a key link to ensure IBN.

[0003] Currently, in IBN solutions, due to the uncertainty of user input, the system's parsing of user intents may be ambiguous. Coupled with problems such as the diversity of networking structures and the complexity of network configuration parameters, it is difficult to guarantee the accuracy of network configuration schemes, reducing the user experience. Summary of the Invention

[0004] This application provides an intent processing method, apparatus, electronic device, and readable storage medium.

[0005] An embodiment of this application provides an intent processing method, which includes: analyzing the obtained buried point data and the initial user intent to determine the user intent to be processed, where the initial user intent is used to represent the original needs of the user; and determining the configuration scheme of the intent network according to the user intent to be processed.

[0006] An embodiment of this application provides an intent processing apparatus, including: an analysis module, configured to analyze the obtained buried point data and the initial user intent to determine the user intent to be processed, where the initial user intent is used to represent the original needs of the user; and a processing module, configured to determine the configuration scheme of the intent network according to the user intent to be processed.

[0007] An embodiment of this application provides an electronic device, including: one or more processors; a memory having stored thereon one or more programs, which when executed by the one or more processors, cause the one or more processors to implement any one of the intent processing methods in the embodiments of this application.

[0008] An embodiment of the present application provides a readable storage medium, which stores a computer program. When the computer program is executed by a processor, any one of the intention processing methods in the embodiments of the present application is implemented.

[0009] According to the intention processing method, device, electronic device and readable storage medium of the embodiments of the present application, by analyzing the obtained buried point data and the initial user intention, the user intention to be processed is determined, simplifying the user's input information. The deficiency of the initial user intention can be made up by the buried point data, expanding the understanding of the user intention and eliminating the ambiguity of the user intention. According to the user intention to be processed, the configuration scheme of the intention network is determined, and the association relationship between the user intention to be processed and the configuration information of the intention network is established, ensuring the accuracy of the configuration of the intention network and improving the accuracy of intention translation in the intention network to enhance the user experience satisfaction.

[0010] More descriptions are provided in the accompanying drawings, specific implementation manners and claims regarding the above embodiments and other aspects of the present application and their implementation manners. Description of the Drawings

[0011] Figure 1 The system framework diagram of the IBN in the present application is shown.

[0012] Figure 2 The flowchart of the intention processing method in an embodiment of the present application is shown.

[0013] Figure 3 The structural diagram of the event metadata and user metadata in the embodiments of the present application is shown.

[0014] Figure 4 The structural diagram of the knowledge graph in the embodiments of the present application is shown.

[0015] Figure 5 The flowchart of the intention processing method in another embodiment of the present application is shown.

[0016] Figure 6 The block diagram of the components of the intention processing device provided by an embodiment of the present application is shown.

[0017] Figure 7 The block diagram of the components of the intention processing device in another embodiment of the present application is shown.

[0018] Figure 8 The flowchart of the method for the intention processing device based on data buried points to process the user intention in the embodiments of the present application is shown.

[0019] Figure 9 The flowchart of the method for complementing the user intention based on the buried point data in the embodiments of the present application is shown.

[0020] Figure 10 A schematic diagram showing the intent features obtained by parsing user information in an embodiment of the present application.

[0021] Figure 11 A flowchart showing a method for translating a user intent to be processed into a configuration scheme of an intent network in an embodiment of the present application.

[0022] Figure 12 A structural diagram showing an exemplary hardware architecture of an electronic device capable of implementing an intent processing method and apparatus according to an embodiment of the present application. Detailed implementation manners

[0023] To make the objectives, technical solutions and advantages of the present application clearer and more understandable, the embodiments of the present application will be described in detail below with reference to the accompanying drawings. It should be noted that, without conflict, the embodiments in the present application and the features in the embodiments may be combined arbitrarily with each other.

[0024] Figure 1 A system framework diagram of the IBN in the present application is shown. As Figure 1 shown, the IBN may include the following devices: a cloudified control and management integrated platform 10, an intent engine 20, a Software-defined Networking (SDN) controller 30, network infrastructure 40, and a database 50.

[0025] Among them, the user 11 inputs the intent information of the user 11 into the cloudified control and management integrated platform 10 by operating the cloudified control and management integrated platform 10, so that the cloudified control and management integrated platform 10 can interact with the SDN controller 30 and the intent engine 20 through a standard interface to monitor the operation status of the IBN network in real time. For example, software such as an "intelligent assistant" is used to assist the user 11 in completing the input of intent information.

[0026] The intent engine 20 performs semantic parsing on the user intent input by the cloudified control and management integrated platform 10 through a Northbound Interface (NBI), converts the information input by the user 11 into a user intent; through intent translation and verification, converts the user intent into a network policy and checks its integrity.

[0027] The SDN controller 30 is an application in a software-defined network, responsible for traffic control to ensure the intelligent operation of the network. The SDN controller interacts with the network infrastructure device 40 based on protocols such as the Open Networking Foundation's OpenFlow protocol, outputs the user instructions input by the cloudified management and control integrated platform 10 to the network infrastructure device 40, and forwards the network intent configuration solution input by the intent engine 20 to the cloudified management and control integrated platform 10, so that the cloudified management and control integrated platform 10 can present the network intent configuration solution to the user 11. The user 11 updates the network intent configuration solution according to their personalized needs, and feeds back the updated network intent configuration solution through the cloudified management and control integrated platform 10 to be sent to the device by the SDN controller 30, thus completing the complete process of intent input, translation, automatic implementation, and deployment and distribution.

[0028] The network infrastructure device 40 is used to transmit information in the IBN. For example, the network infrastructure device 40 can be an optical fiber device to achieve fast transmission of information in the IBN, etc.

[0029] The database 50 is used to store the natural language information input by the user 11 to assist the intent engine 20 in translating the user intent.

[0030] However, in the prior art, for the translation of the user intent by the intent engine 20, mainly the feature information in the natural language information of the user 11 is directly extracted, without referring to other information related to the user 11, which easily leads to poor translation accuracy of the natural language information of the user 11, making the user 11 unable to obtain the network service they want through the IBN network. For example, when the IBN solution is applied in a communication network, due to the complexity of the network structure and service logic, the intent translation network device used may translate the instruction information incorrectly when obtaining a certain service processing intent instruction input by the user through voice, resulting in the user being unable to obtain the expected physical configuration solution, thereby reducing the user experience satisfaction.

[0031] To make the objectives, technical solutions, and advantages of the present application clearer, the following will further describe the embodiments of the present application in detail with reference to the accompanying drawings.

[0032] Figure 2 A flowchart showing the intent processing method according to an embodiment of the present application is shown. This intent processing method can be applied to an intent processing device. As Figure 2 shown, the intent processing method in the embodiments of the present application may include the following steps.

[0033] Step 110: Analyze the obtained buried point data and the initial user intent to determine the user intent to be processed.

[0034] The initial user intent is used to represent the user's original needs. For example, the initial user intent is "user A activates a 10GE service" to express what user A wants to do.

[0035] In some specific implementations, the tracking data is data extracted from log tracking information. For example, by collecting and organizing the log tracking information, the tracking data related to the initial user intent is extracted.

[0036] In some specific implementations, the acquired buried data and initial user intent are analyzed to determine the user intent to be processed, including: step 111, extracting the first intention feature set in the initial user intent; step 112, determining the intention feature set to be analyzed based on the first intention feature set and the full set of preset intention features; step 113, determining the user intent to be supplemented based on the buried data and the intention feature set to be analyzed; step 114, filling the user intent to be supplemented into the initial user intent to obtain the user intent to be processed.

[0037] For example, by comparing the first intention feature set with the full set of preset intention features, the intention feature set to be analyzed is determined, and the intention features shared by the intention feature set to be analyzed and the buried data are extracted, and based on this part of the shared intention features, the user intent to be supplemented is determined; the user intent to be supplemented is filled into the initial user intent to ensure the completeness and accuracy of the user intent, and to improve the accuracy of the subsequent translation of the user intent.

[0038] Step 120: Determine a configuration scheme for the intent-based network based on the user intent to be processed.

[0039] For example, through intent translation, the user intent to be processed is translated into the configuration plan of the intent network to ensure the accuracy of the configuration of the intent network.

[0040] In some specific implementations, a configuration scheme of the intent network is determined based on the user intent to be processed, including: step 121, determining a set of initial configuration schemes based on the knowledge graph, a preset reasoning algorithm and the user intent to be processed; step 122, calculating the weighted value of each initial configuration scheme in the set of initial configuration schemes based on the preset business type; step 123, determining the configuration scheme of the intent network based on the weighted value of each initial configuration scheme and the set of initial configuration schemes.

[0041] By calculating the weighted values of the initial configuration schemes corresponding to different preset service types, the screening of each initial configuration scheme can be further refined. When the user intention to be processed is used to represent a specific service type among the preset service types, the initial configuration scheme corresponding to the specific service type can be accurately screened out, accelerating the screening of each initial configuration scheme in the initial configuration scheme set, improving the generation efficiency of the configuration scheme of the intention network, and enhancing the user experience satisfaction.

[0042] In this embodiment, by analyzing the obtained buried point data and the initial user intention, the user intention to be processed is determined, simplifying the user's input information. The deficiency of the initial user intention can be made up by the buried point data, expanding the understanding of the user intention, and eliminating the ambiguity of the user intention. According to the user intention to be processed, the configuration scheme of the intention network is determined, and the association relationship between the user intention to be processed and the configuration information of the intention network is established, ensuring the configuration accuracy of the intention network and improving the accuracy of intention translation in the intention network, so as to enhance the user experience satisfaction.

[0043] In a specific implementation, determining the user intention to be supplemented according to the buried point data and the intention feature set to be analyzed in step 113 includes the following steps:

[0044] Step 1131, analyze the event metadata and user metadata in the buried point data to obtain the buried point information.

[0045] It should be noted that the buried point data includes two different models, for example, the model of event metadata and the model of user metadata. Figure 3 The structural schematic diagrams of the event metadata and user metadata in the embodiments of the present application are shown. Among them, the event metadata 210 includes a preset event 211, a virtual event 212, and a custom event 213; the user metadata 220 includes a preset attribute 221 and a custom attribute 222. The attribute information of the preset event 211 in the event metadata 210 may include a user identifier 2111, an event identifier 2112, location information 2113, behavior information 2114, context information 2115, and custom attribute information 2116. The preset attribute 221 in the user metadata 220 includes a user identifier 2211, natural attribute information 2212, user usage habit information 2213, and account attribute information 2214. The user usage habit information 2213 may be information such as the user's habit of using the alarm monitoring function for a long time; the natural attribute information 2212 may be information such as the user's age and gender.

[0046] Among them, event metadata is data used to characterize a user's behavior, user metadata is data used to characterize a user's attribute features, the buried point information includes a buried point intention feature set and a buried point event set, and the buried point intention feature set includes buried point intention features. By classifying the buried point information and extracting different types of information, the processing of the buried point data can be accelerated, facilitating the subsequent processing of the buried point information and improving the processing efficiency.

[0047] Step 1132, when it is determined that there is a matching event in the buried point event set that matches the preset mapping relationship, extract the matching event in the buried point event set to generate a matching event set.

[0048] Among them, the preset mapping relationship is the mapping relationship between a preset event and a preset intention feature. For example, the preset event is "alarm query", and the preset intention features include "alarm identifier, alarm type, and alarm generation time"; by comparing the events in the buried point event set with the preset, when there is an "alarm query" in the buried point event set, it means that there is a matching event in the buried point event set, and this matching event is the "alarm query". Thus, according to the preset mapping relationship, it can be quickly determined that the preset intention features corresponding to this matching event include "alarm identifier, alarm type, and alarm generation time". This improves the extraction speed of the intention features of the matching event and accelerates the screening of the matching event.

[0049] Step 1133, based on the matching event set, the buried point intention feature set, and the intention feature set to be analyzed, determine the user intention to be supplemented.

[0050] In some specific implementations, determining the user intention to be supplemented based on the matching event set, the buried point intention feature set, and the intention feature set to be analyzed includes: extracting a second intention feature from the buried point intention feature set according to the matching event in the matching event set, where the second intention feature is the buried point intention feature corresponding to the matching event; searching for the intention feature set to be analyzed according to the second intention feature to obtain a search result; when it is determined that there is an intention feature in the intention feature set to be analyzed that is the same as the second intention feature in the search result, calculate the feature value corresponding to the second intention feature; based on the second intention feature and its corresponding feature value, determine the user intention to be supplemented.

[0051] For example, a machine learning algorithm (MLA) can be used to calculate the eigenvalue corresponding to the second intent feature, enabling the intent processing device to more intuitively understand the buried point intent feature and ensuring the correctness of processing the buried point intent features in the buried point intent feature set. Then, based on the second intent feature and its corresponding eigenvalue, the user intent to be supplemented is determined through intent translation to ensure the integrity of the user intent, so that the subsequent analysis of the user intent can be accurate and more in line with the user's needs.

[0052] By extracting the matching events that match the preset mapping relationship from the buried point event set, the processing of the buried point data is accelerated, and the data processing efficiency is improved; then, the second intent feature corresponding to the matching event is extracted from the buried point intent feature set, and the eigenvalue of the second intent feature is calculated, enabling the intent processing device to more intuitively understand the buried point intent feature and improving the processing accuracy of the device; the user intent to be supplemented is determined through intent translation to ensure the integrity of the user intent, so that the subsequent analysis of the user intent can be accurate and more in line with the user's needs.

[0053] In some specific implementations, filling the user intent to be supplemented into the initial user intent to obtain the user intent to be processed in step 114 includes: extracting the second intent feature in the user intent to be supplemented; calculating the eigenvalue corresponding to the second intent feature; filling the second intent feature and its corresponding eigenvalue into the first intent feature set in the initial user intent to generate the intent feature set to be processed; obtaining the event to be processed corresponding to the intent feature in the intent feature set to be processed; and determining the user intent to be processed based on the intent feature set to be processed and the event to be processed.

[0054] For example, if the user inputs "User A wants to open a 10GE service", the initial intent features of the user can be expressed as {person, service type, layer rate, time}, where the person is "User A", the service type is "customer layer", the layer rate is "10GE", and the time is "September 24, 2020". By filling the second intent feature {source end, destination end, protection type, intelligence level} and its corresponding eigenvalue into the first intent feature set in the initial user intent, the intent feature set to be processed is generated, and the intent feature set to be processed can be expressed as {person, service type, layer rate, source end, destination end, protection type, intelligence level, time}.

[0055] The above set of pending intent features is used to determine the user intent to be processed, expanding the user intent so that the intent processing device can more fully understand the user's intended intention and ensure the integrity of the user intent. This provides more reliable data for the subsequent translation of the pending user intent into the configuration scheme of the intent network, improving the accuracy of the intent network translation.

[0056] In some specific implementations, determining a set of initial configuration schemes based on the knowledge graph, a preset reasoning algorithm, and the user intent to be processed in step 121 includes: generating a set of first configuration schemes based on the knowledge graph and the preset reasoning algorithm, wherein the knowledge graph includes entity elements, attribute elements, and relationship elements; training the schemes in the set of first configuration schemes based on the user intent to be processed to obtain training results; sorting the training results based on a scoring function to obtain a first sorting result, wherein the scoring function is a function determined based on the correlation between entity elements and relationship elements; and determining a set of initial configuration schemes based on the first sorting result.

[0057] For example, Figure 4 The schematic diagram of the structure of the knowledge graph in the embodiment of the present application is shown. Among them, a knowledge graph can be represented as a triple (subject, relationship, object). For example, Figure 4 As shown, Daming's son is Xiaoming, represented by the triple (Daming, son, Xiaoming). Subjects and objects are collectively referred to as entities. Relationships are irreversible, meaning that the subject and object cannot be reversed. A collection of knowledge graphs, linked together, forms a graph. Each node is an entity, and each edge is a relationship or fact. Each edge represents the connection between a subject and an object. Therefore, a knowledge graph is a directed graph.

[0058] Using a preset inference algorithm, reasoning training is performed on each entity in the knowledge graph and its corresponding relationships to generate a set of first configuration solutions. The user intent to be processed is then used as an input parameter to train the solutions in the first set of configuration solutions to obtain training results that are more consistent with the user intent to be processed. The training results are sorted according to a scoring function to obtain a first sorted result. The top-ranked training results in the first sorted result are used as initial configuration solutions to obtain multiple initial configuration solutions. For example, the top 10 training results in the first sorted result are extracted to generate a set of initial configuration solutions, thereby improving the configuration solutions of the intent network to better match the user intent to be processed and improving the accuracy of translating user intent into configuration solutions of the intent network.

[0059] In some specific implementations, calculating the weighted value of each initial configuration scheme in the set of initial configuration schemes based on the preset service type in step 122 includes:

[0060] Step 1221: Extract the configuration parameter information and the corresponding initial event information of each initial configuration scheme in sequence.

[0061] Step 1222: Determine the target event metric information corresponding to the preset service type according to the preset service type.

[0062] It should be noted that the target event metric information corresponding to different service types is also different. Among them, the preset service type can be service issuance or the processing of alarm information. The above are only examples of the preset service type, and can be specifically set according to actual needs. Other unspecified preset service types are also within the protection scope of this application and will not be elaborated here.

[0063] Step 1223: Calculate the weighted value of each initial configuration scheme according to the target event metric information and the initial event information.

[0064] For example, compare the target event metric information with the initial event information. When the relevant information in the initial event information meets the target event metric, the weighted value of the corresponding initial configuration scheme will be higher; when the relevant information in the initial event information is far from the target event metric, the weighted value of the corresponding initial configuration scheme will be lower. Ensure that different initial configuration schemes can determine the weighted value of the initial configuration scheme according to the degree to which the corresponding initial event information is away from the target event metric information, so as to distinguish the advantages and disadvantages of different initial configuration schemes, facilitate the screening of the initial configuration scheme, and obtain a configuration scheme of the intent network that better meets the user's needs.

[0065] In some specific implementations, calculating the weighted value of each initial configuration scheme according to the target event metric information and the initial event information in Step 1223 includes: in the case where the preset service type is service issuance, determining the conversion rate of successful service issuance of each initial configuration scheme according to the first target event metric information corresponding to service issuance and each initial event information; determining the weighted value of each initial configuration scheme according to the conversion rate of successful service issuance of each initial configuration scheme; in the case where the preset service type is alarm reduction, determining the alarm retention rate of each initial configuration scheme according to the second target event metric information corresponding to alarm reduction and each initial event information; determining the weighted value of each initial configuration scheme according to the alarm retention rate of each initial configuration scheme.

[0066] For example, the alarm reduction levels are divided into level 1, level 2, and level 3. The lower the level number, the higher the corresponding alarm retention rate. For example, when the alarm reduction level is level 1, the corresponding alarm retention rate is 30%; when the alarm reduction level is level 2, the corresponding alarm retention rate is 60%; when the alarm reduction level is level 3, the corresponding alarm retention rate is 75%. That is, users have different tolerances for alarm reduction.

[0067] By sequentially extracting the configuration parameter information of each initial configuration scheme and the initial event information corresponding to the configuration parameter information; determining the target event metric information corresponding to the preset service type according to the preset service type; calculating the weighted values of each initial configuration scheme according to the target event metric information and the initial event information, and using the weighted values to reflect the advantages and disadvantages of the configuration schemes corresponding to different service types, ensuring the configuration accuracy of the intent network and improving the user experience satisfaction.

[0068] In some specific implementations, determining the configuration scheme of the intent network according to the weighted values of each initial configuration scheme and the set of initial configuration schemes in step 123 includes: re-sorting each initial configuration scheme in the set of initial configuration schemes according to the first sorting result and the weighted values of each initial configuration scheme to obtain a second sorting result; determining the configuration scheme of the intent network according to the second sorting result.

[0069] By using the weighted values of each initial configuration scheme as the reference values for the second sorting, the second sorting can more accurately match the configuration schemes of the intent network corresponding to different service types, improving the configuration accuracy of the intent network.

[0070] Figure 5 The flowchart of the intent processing method in another embodiment of the present application is shown. This intent processing method can be applied to an intent processing device. As Figure 5 shown, the intent processing method in the embodiment of the present application may include the following steps.

[0071] Step 410, obtain user information according to the intent start event identifier and the intent end event identifier.

[0072] Among them, the user information includes the user's operation information and input information. For example, the user's operation information can be the operation of opening a web page and browsing certain web pages; the user's input information can be information such as the user name and password entered when the user logs in to certain web pages. When the user enters the intent start page (taking the operation of "the user enters the intent start page" as the intent start event identifier), a session of the user intent is started, and a new intent session is generated. When the user enters a certain intent end page (taking the operation of "the user enters the intent end page" as...), the intent processing device will select the user's operation information and input information between the intent start event identifier and the intent end event identifier as the user information of the current intent session to facilitate the extraction of user information.

[0073] Step 420: Determine the initial user intent based on the user's operation information and input information.

[0074] Among them, the input information includes intent text and / or voice information, and the initial user intent is an intent language recognizable by a machine.

[0075] For example, when the input information is intent text, through technical means such as command parsing in machine learning, the intent text is converted into the initial user intent, that is, an intent language recognizable by a machine. When the user information is voice information, the voice information needs to be first converted into text information, and then according to the processing method of the text information, the text information is converted into the initial user intent. Ensure that for different forms of input information, it can be converted into an intent voice recognizable by a machine, facilitating further processing by the intent processing device.

[0076] Step 430: Obtain the buried point data from the data server according to the intent start event identifier and the buried point port information of the data server.

[0077] It should be noted that the buried point data therein is data related to the initial user intent. When the user starts an intent session (for example, taking the intent start event identifier as the identifier for starting to extract the buried point data), the buried point data stored in the data server will be synchronously obtained through the buried point port of the data server (for example, the port number is 4500, etc.). The user intent is supplemented through this buried point data, enabling the user intent to be more fully expressed as information recognizable by a machine, and improving the accuracy of the configuration scheme for translating the user intent into an intent network.

[0078] Step 440: Analyze the obtained buried point data and the initial user intent to determine the user intent to be processed.

[0079] Step 450: Determine the configuration scheme of the intent network according to the user intent to be processed.

[0080] It should be noted that steps 440 to 450 in this embodiment are the same as steps 110 to 120 in the previous embodiment, and will not be elaborated here.

[0081] In this embodiment, by relying on the intent start event identifier and the intent end event identifier, user information is obtained; and by relying on the intent start event identifier and the data server's buried point port information, buried point data is obtained from the data server; the obtained buried point data and the initial user intent are analyzed to determine the user intent to be processed, making the user intent to be processed more sufficient and capable of expressing the user's needs more accurately, so as to ensure the integrity and accuracy of the user intent. According to the user intent to be processed, a configuration scheme for the intent network is determined. Ensure the accuracy of the configuration of the intent network, improve the scalability of the intent network, and improve user experience satisfaction.

[0082] In some specific implementations, after the step of determining the configuration scheme of the intent network according to the user intent to be processed, it further includes: feeding back the configuration scheme of the intent network to the user; obtaining the modification parameters determined by the user based on the configuration scheme of the intent network; according to the modification parameters and the configuration scheme of the intent network, updating the configuration scheme of the intent network, and the configuration scheme of the intent network includes network configuration parameters.

[0083] By obtaining the modification parameters determined by the user based on the configuration scheme of the intent network, the user can update the configuration scheme of the intent network according to personalized needs, fully ensuring the accuracy of the configuration parameters of the intent network.

[0084] In some specific implementations, after the step of updating the configuration scheme of the intent network according to the modification parameters and the configuration scheme of the intent network, it further includes: configuring the intent network according to the updated configuration scheme of the intent network; in the case of determining that the configuration of the intent network is successful, deleting the cached initial user intent and buried point data.

[0085] By using the updated configuration scheme of the intent network to configure the intent network, it is ensured that the configuration of the intent network better meets the user's needs and is closer to the user intent; in the case of determining that the configuration of the intent network is successful, deleting the cached initial user intent and buried point data saves storage space, facilitates the next processing of user intent and corresponding buried point data, avoids data confusion, and improves data processing efficiency.

[0086] Next, the apparatus according to the embodiments of the present application will be introduced in detail with reference to the accompanying drawings. Figure 6 The block diagram showing the composition of the intent processing apparatus provided by an embodiment of the present application is as follows Figure 6 As shown, the intent processing apparatus may include the following modules:

[0087] An analysis module 510 is configured to analyze the acquired logged data and the initial user intention to determine the user intention to be processed, where the initial user intention is used to represent the original needs of the user; a processing module 520 is configured to determine a configuration scheme of the intention network according to the user intention to be processed.

[0088] In some specific implementations, the analysis module 510 includes: an extraction sub-module configured to extract a first set of intention features from the initial user intention; a first determination sub-module configured to determine a set of intention features to be analyzed according to the first set of intention features and a preset full set of intention features; a second determination sub-module configured to determine the user intention to be supplemented according to the logged data and the set of intention features to be analyzed; and a filling sub-module configured to fill the user intention to be supplemented into the initial user intention to obtain the user intention to be processed.

[0089] In some specific implementations, the second determination sub-module is configured to analyze the event metadata and user metadata in the logged data to obtain logged information, where the event metadata is data used to represent the behavior of the user, the user metadata is data used to represent the attribute characteristics of the user, the logged information includes a set of logged intention features and a set of logged events, and the set of logged intention features includes logged intention features; in the case where a matching event that matches a preset mapping relationship exists in the set of logged events, extract the matching event in the set of logged events to generate a set of matching events, where the preset mapping relationship is a mapping relationship between a preset event and a preset intention feature; and determine the user intention to be supplemented according to the set of matching events, the set of logged intention features, and the set of intention features to be analyzed.

[0090] In some specific implementations, the second determination sub-module's determining the user intention to be supplemented according to the set of matching events, the set of logged intention features, and the set of intention features to be analyzed includes: extracting a second intention feature from the set of logged intention features according to the matching event in the set of matching events, where the second intention feature is a logged intention feature corresponding to the matching event; searching for the set of intention features to be analyzed according to the second intention feature to obtain a search result; in the case where it is determined that an intention feature identical to the second intention feature exists in the set of intention features to be analyzed in the search result, calculating the feature value corresponding to the second intention feature; and determining the user intention to be supplemented according to the second intention feature and its corresponding feature value.

[0091] In some specific implementations, the filling sub-module is used to extract the second intention feature in the user intention to be supplemented; calculate the feature value corresponding to the second intention feature; fill the second intention feature and its corresponding feature value into the first intention feature set in the initial user intention to generate a to-be-processed intention feature set; obtain the to-be-processed event corresponding to the intention feature in the to-be-processed intention feature set; and determine the to-be-processed user intention based on the to-be-processed intention feature set and the to-be-processed event.

[0092] In some specific implementations, the processing module 520 includes: an initial configuration scheme determination sub-module, which is used to determine a set of initial configuration schemes according to the knowledge graph, the preset inference algorithm, and the to-be-processed user intention; a weighting value determination sub-module, which is used to calculate the weighting value of each initial configuration scheme in the set of initial configuration schemes according to the preset business type; and a configuration scheme determination sub-module, which is used to determine the configuration scheme of the intention network according to the weighting value of each initial configuration scheme and the set of initial configuration schemes.

[0093] In some specific implementations, the initial configuration scheme determination sub-module is used to generate a set of first configuration schemes based on the knowledge graph and the preset inference algorithm, where the knowledge graph includes entity elements, attribute elements, and relationship elements; train the schemes in the set of first configuration schemes according to the to-be-processed user intention to obtain a training result; sort the training result according to the scoring function to obtain a first sorting result, where the scoring function is a function determined according to the association degree between the entity elements and the relationship elements; and determine the set of initial configuration schemes according to the first sorting result.

[0094] In some specific implementations, the configuration scheme determination sub-module is used to re-sort each initial configuration scheme in the set of initial configuration schemes according to the first sorting result and the weighting value of each initial configuration scheme to obtain a second sorting result; and determine the configuration scheme of the intention network according to the second sorting result.

[0095] In some specific implementations, the weighting value determination sub-module is used to sequentially extract the configuration parameter information of each initial configuration scheme and the initial event information corresponding to the configuration parameter information; determine the target event metric information corresponding to the preset business type according to the preset business type; and calculate the weighting value of each initial configuration scheme according to the target event metric information and the initial event information.

[0096] In some specific implementations, the weighted value determination sub-module calculates the weighted values of each initial configuration scheme based on the target event metric information and the initial event information, including: when it is determined that the preset service type is service issuance, determining the conversion rate of successful service issuance for each initial configuration scheme based on the first target event metric information corresponding to service issuance and each initial event information; determining the weighted value of each initial configuration scheme based on the conversion rate of successful service issuance for each initial configuration scheme; when it is determined that the preset service type is alarm reduction, determining the alarm retention rate of each initial configuration scheme based on the second target event metric information corresponding to alarm reduction and each initial event information; determining the weighted value of each initial configuration scheme based on the alarm retention rate of each initial configuration scheme.

[0097] In some specific implementations, the intent processing device further includes: a user information acquisition module, configured to acquire user information based on the intent start event identifier and the intent end event identifier, where the user information includes the user's operation information and the user's input information; a determination module, configured to determine the initial user intent based on the user's operation information and the user's input information, where the input information includes intent text and / or voice information, and the initial user intent is an intent language recognizable by a machine; a buried point data acquisition module, configured to acquire buried point data from a data server based on the intent start event identifier and the buried point port information of the data server.

[0098] In some specific implementations, the intent processing device further includes: a feedback module, configured to feedback the configuration scheme of the intent network to the user; a parameter modification module, configured to acquire the modification parameters determined by the user based on the configuration scheme of the intent network; an update module, configured to update the configuration scheme of the intent network based on the modification parameters and the configuration scheme of the intent network, where the configuration scheme of the intent network includes network configuration parameters.

[0099] In some specific implementations, the intent processing device further includes: a configuration module, configured to configure the intent network based on the updated configuration scheme of the intent network; a deletion module, configured to delete the cached initial user intent and buried point data when it is determined that the configuration of the intent network is successful.

[0100] In some specific implementations, the buried point data in the intent processing device is data extracted from log buried point information.

[0101] According to the intention processing device of the embodiment of the present application, the analysis module analyzes the obtained buried point data and the initial user intention to determine the user intention to be processed, simplifies the user's input information, can make up for the deficiencies of the initial user intention through the buried point data, expands the understanding of the user intention, and eliminates the ambiguity of the user intention; the processing module is used to determine the configuration scheme of the intention network according to the user intention to be processed, and establish the association relationship between the user intention to be processed and the configuration information of the intention network, ensuring the configuration accuracy of the intention network and improving the accuracy of intention translation in the intention network to enhance the user experience satisfaction.

[0102] It should be clear that the present application is not limited to the specific configurations and processes described and illustrated in the above embodiments. For the convenience and brevity of description, the detailed description of known methods is omitted here, and the specific working processes of the systems, modules and units described above can refer to the corresponding processes in the foregoing method embodiments, which will not be elaborated here.

[0103] Figure 7 The block diagram showing the composition of the intention processing device in another embodiment of the present application is as Figure 7 shown, the intention processing device based on data buried point includes: an intention feedback module 610, an intention parsing module 620, a buried point analysis module 630 and a session control module 640 connected in sequence. Among them, the buried point analysis module 630 includes a preprocessing unit 631, a result recombination unit 632, an event analysis unit 633, an index evaluation unit 634 and a data extraction unit 635 connected in sequence.

[0104] The intention feedback module 610 is used to send information such as the user intention output by the intention parsing module 620 and the configuration scheme of the intention network corresponding to the user intention to the display interface, so that the user can obtain the configuration scheme of the intention network through the display interface.

[0105] The intention parsing module 620 is used to convert the information input by the user (for example, intention text and / or voice information, etc.) into the initial user intention, that is, the intention language that can be recognized by the machine. For example, the intention parsing module 620 obtains the original intention text input by the user and the user's behavior events, analyzes the user's behavior events, and uses text preprocessing functions to eliminate invalid text inputs, and converts valid text inputs into initial user intentions through machine learning techniques such as command parsing and semantic slot filling. The intention parsing module 620 calls the buried point analysis module 630 to obtain the user intention to be processed after the semantic slot is filled and returned by the result reorganization unit 632 in the buried point analysis module 630 to ensure the integrity and accuracy of the user intention. Based on the user intention to be processed and the weighted value input by the indicator evaluation unit 634 in the buried point analysis module 630, the configuration plan for the intention network is determined, and the configuration plan of the intention network is fed back to the intention feedback module 610.

[0106] It should be noted that the intent parsing module 620 uses a preset reasoning algorithm to process the knowledge graph and the user intent to be processed (for example, through the generation of intent strategies, network orchestration and verification, etc.) to generate a set of initial configuration schemes. Combined with the weighted value input by the indicator evaluation unit 634, each initial configuration scheme in the set of initial configuration schemes is evaluated and optimized to determine the configuration scheme of the intent network that is ultimately recommended to the user. The configuration scheme of the intent network is then output to the intent feedback module 610 so that the intent feedback module 610 feeds back to the user to obtain the modification parameters determined by the user based on the configuration scheme of the intent network; based on the modification parameters and the configuration scheme of the intent network, the configuration scheme of the intent network is updated. The configuration accuracy of the intent network is ensured.

[0107] The tracking point analysis module 630 is used to read tracking point data from the data server, analyze the tracking point data, determine the user intent to be supplemented, and feed the supplemented user intent back to the intent analysis module 620. At the same time, based on different business types, it calculates the weighted values of each initial configuration plan and feeds this weighted value back to the intent analysis module 620.

[0108] Preprocessing unit 631 is configured to perform preprocessing operations such as feature extraction and attribute comparison on the initial user intent input by intent parsing module 620 to obtain a first intent feature set, where the first intent feature set includes multiple first intent features. The first intent features are compared with intent features in the full set of preset intent features to determine the intent feature set to be analyzed.

[0109] The result recombination unit 632 is used to obtain the set of buried-point intention features input by the event analysis unit 633 and the feature values corresponding to each second intention feature. The set of buried-point intention features includes second intention features. Process the second intention features and their corresponding feature values, convert them into user intentions to be supplemented, and send the user intentions to be supplemented to the intention parsing module 620.

[0110] The event analysis unit 633 is used to analyze the set of buried-point intention features, the set of buried-point events, and the preset mapping relationship input by the data extraction unit 635 to generate a set of matching events (for example, when it is determined that there are matching events in the set of buried-point events that match the preset mapping relationship, extract the matching events in the set of buried-point events to generate a set of matching events); the set of buried-point intention features includes buried-point intention features, and the preset mapping relationship is the mapping relationship between preset events and preset intention features; according to the matching events in the set of matching events, extract the second intention features from the set of buried-point intention features, where the second intention features are the buried-point intention features corresponding to the matching events; search for the set of intention features to be analyzed input by the preprocessing unit 631 according to the second intention features to obtain a search result; when it is determined that there are intention features in the set of intention features to be analyzed that are the same as the second intention features, calculate the feature values corresponding to the second intention features; output the second intention features and their corresponding feature values to the result recombination unit 632.

[0111] The metric evaluation unit 634 is used to obtain the set of buried-point intention features, the set of buried-point events, and the set preset mapping relationship from the data extraction unit 635. Combine the user intentions to be processed input by the intention parsing module 620 for comprehensive analysis, and generate corresponding weighted values according to different business types; and feedback the weighted values to the intention parsing module 620.

[0112] For example, when it is determined that the preset business type is business issuance, set the first target event metric information corresponding to business issuance to the conversion rate of the page where business issuance is successful. Through the funnel analysis method and evaluation algorithm, determine the conversion rate of the page where business issuance is successful within the preset duration (for example, within 10 minutes). When it is determined that the preset business type is alarm reduction, within the preset duration, use association analysis to analyze users who have performed event operations such as alarm confirmation and / or alarm clearing, so that when users express the user intention of "quickly reduce" alarms during the alarm peak period, evaluate the alarm retention rates corresponding to different alarm reduction rules. (For example, set that the smallest or largest alarm volume is in the Mth week, and the second target event metric information corresponding to alarm reduction is the correlation relationship between "alarm retention in the Mth week" and the "alarm reduction" event). Determine the conversion trend based on the conversion rate of the page where business issuance is successful or the alarm retention rate as the weighted value of the configuration scheme of the intention network.

[0113] A data extraction unit 635 is configured to process the buried-point data and convert the buried-point data into buried-point information, where the buried-point information includes a set of buried-point intention features and a set of buried-point events. A preset mapping relationship is set, and the preset mapping relationship is a mapping relationship between preset events and preset intention features.

[0114] For example, the buried-point data (such as event metadata and user metadata, etc.) is read from a data server, and then the buried-point data is parsed to obtain JavaScript Object Notation (JSON) data; in the JSON data, user identifiers and buried-point events corresponding to the preset mapping relationship are extracted according to preset conditions, and the buried-point events are filtered, de-duplicated, etc. to generate processed JSON data; the processed JSON data is sent to a metric evaluation unit 634 for analysis and processing. JSON data is easy for humans to read and write, and is also easy for machines to parse and generate, effectively improving the network transmission efficiency.

[0115] A session control module 640 is configured to maintain the processing flow of the user intention to be processed, manage the context dialogue information, etc., so as to track the user behavior and events.

[0116] In this embodiment, the information input by the user is converted into an initial user intention through an intention parsing module, so that the machine can recognize the user intention that the user hopes to express. Each unit in the buried-point analysis module analyzes and processes the buried-point data, and obtains the user intention to be processed filled with semantic slots returned by the result recombination unit in the buried-point analysis module to ensure the integrity and accuracy of the user intention. According to the user intention to be processed and the weighted value input by the metric evaluation unit in the buried-point analysis module, a configuration scheme for the intention network is determined, and the configuration scheme of the intention network is fed back to the intention feedback module, so that the user can obtain a matching configuration scheme of the intention network in time, which is convenient for the user to quickly and accurately configure the intention network, improves the accuracy of intention translation in the intention network, and enhances the user experience satisfaction.

[0117] Figure 8 The flowchart shows the method for processing the user intention by the intention processing device based on data burying in the embodiment of the present application. As Figure 8 shown, the intention processing device based on data burying adopts the following steps to process the user intention.

[0118] Step 701, in response to an intention session request sent by the user, obtain user information.

[0119] Among them, the user information includes the operation information of the user and the input information of the user.

[0120] For example, when the user enters the intent start page (taking the operation of "the user enters the intent start page" as the start event (Start_Event)), the session of the user intent is started, and a new intent session is generated. At the same time, the switch for obtaining buried point data is turned on (for example, in the way of a Software Development Kit (SDK), the acquisition of buried point data is started), and the buried point information corresponding to the current intent session is automatically collected; when the user enters a certain (or multiple) intent end page (taking the operation of "the user enters the intent end page" as the end event (End_Event)), or when the duration of the current intent session exceeds the preset session duration (for example, 6 hours, etc.), the current intent session is closed. It should be noted that different intent sessions are distinguished by session identifiers and business domain identifiers to facilitate the extraction and processing of relevant information of different intent sessions.

[0121] Step 702, obtain the buried point data related to the user information.

[0122] For example, in the way of semi-automated buried point, part of the manual work is standardized to make an SDK, and this SDK is embedded in the product; when the user information includes the start event, the buried point analysis module reads the buried point data from the data server by embedding the SDK.

[0123] Among them, the buried point data includes event metadata and user metadata. The event metadata includes any one or several of preset events, virtual events, and custom events; the user metadata includes any one or several of user identifiers (UserID), user attribute information (such as information about the user's age and gender, etc.), user usage habit information (such as long-term use of alarm monitoring, etc.), and account attribute information. The preset events include any one or several of user identifiers, event identifiers, location information, behavior information, context information, and event attribute information. The custom event can be an event obtained through the Track_Event interface of the SDK and is the main data for analyzing user behavior and buried point data.

[0124] It should be noted that the event metadata and the user metadata are associated by a unique identifier (such as UserID, or device identifier, etc.). According to the differences in the calling methods of the interfaces, this unique identifier is also different.

[0125] In some specific implementations, depending on the business type intended by the user, the custom event can be any one or more of the behavioral events such as the browsing operation of a topology object (ViewTopo_Event), the submission of the connection configuration operation between network elements (LinkCfg_Event), the single board installation operation (SetupBoard_Event) and the network element configuration operation (NetCfg_Event). The custom event can also include attribute information of the event such as the preset connection type, network element name, connection name, operation time, etc. Then, it is stored in the data server in the form of event metadata for call analysis by the intent processing device to facilitate subsequent processing. It should be noted that the above is only an example of custom events. Other unspecified custom events are also within the scope of protection of this application and can be set according to specific circumstances. They will not be repeated here.

[0126] Step 703: Analyze the user information to obtain the initial user intention.

[0127] When it is determined that the user information is intent text, a text preprocessing function is used to eliminate invalid text information and obtain valid text information. The valid text information is then converted into the initial user intent, that is, the machine-recognizable intent language, through machine learning techniques such as command parsing or semantic slot filling.

[0128] If the user information is determined to be voice information, the original voice signal is converted into text information through Automatic Speech Recognition (ASR) technology, and then the text information is converted into a framed semantic representation through Natural Language Processing (NLU) technology. For example, the Natural Language Understanding Intelligent Service (LUIS) platform for developers is used to convert text information into a framed semantic representation.

[0129] In step 704, the user intent to be supplemented and related to the user information in the tracking data is added to the initial user intent to obtain the user intent to be processed.

[0130] For example, call a preprocessing function to perform preprocessing operations such as cleaning and validity verification on the buried point data, and determine the buried point intention feature set and the buried point event set of the buried point data. In the case where a matching event that matches the preset mapping relationship exists in the determined buried point event set, extract the matching events in the buried point event set to generate a matching event set; wherein, the buried point intention feature set includes buried point intention features, and the preset mapping relationship is the mapping relationship between preset events and preset intention features; according to the matching events in the matching event set, extract the second intention features from the buried point intention feature set, where the second intention features are the buried point intention features corresponding to the matching events; search for the intention feature set to be analyzed input by the preprocessing unit 631 according to the second intention features to obtain a search result; in the case where it is determined that the search result is that there is an intention feature identical to the second intention feature in the intention feature set to be analyzed, calculate the feature value corresponding to the second intention feature, and then fill the second intention feature and its corresponding feature value into the initial user intention to obtain the user intention to be processed.

[0131] Step 705, determine a configuration plan for the intention network according to the user intention to be processed.

[0132] In a semantic network, a knowledge graph is composed of three elements: entities, attributes, and relationships. The knowledge graph can be represented as a triple (head entity, relationship, tail entity). Among them, the head entity and the tail entity are collectively referred to as entities, and the entities include attribute information. Adopt the Translating Embedding (TransE) algorithm to perform knowledge reasoning based on the knowledge graph to obtain a set of initial configuration plans; calculate the weighted values of each initial configuration plan in the set of initial configuration plans according to the preset service type; sort each initial configuration plan according to the weighted values of each initial configuration plan to obtain a sorting result; determine a configuration plan for the intention network according to this sorting result. Among them, the intention network is essentially a combination of network features. For example, various combinations of each topological node, tunnel policy, etc. form the intention network.

[0133] Step 706, feedback the configuration plan of the intention network to the user, obtain the modification parameters determined by the user based on the configuration plan of the intention network, and update the configuration plan of the intention network according to the modification parameters and the configuration plan of the intention network.

[0134] It should be noted that after obtaining the configuration scheme of the intent network, there may still be some deficiencies, or there may be multiple different configuration schemes at the same time. At this time, it is necessary to feedback multiple different configuration schemes to the user for the user to select according to their own needs, or the user can modify the relevant parameters in the configuration scheme according to personalized requirements. When obtaining the modified parameters determined by the user based on the configuration scheme of the intent network, the intent processing device will re-correct the configuration scheme of the intent network selected by the user according to the modified parameters to obtain the final configuration scheme of the intent network, which fully guarantees the accuracy of the configuration parameters of the intent network.

[0135] Step 707, configure the intent network using the updated configuration scheme of the intent network.

[0136] It should be noted that after completing the configuration of the intent network and receiving the configuration success message feedback from the underlying network, the intent processing device needs to clear the relevant cache data of this intent session (for example, the cached initial user intent and the buried point data related to the initial user intent, etc.).

[0137] In this embodiment, by converting the information input by the user into the initial user intent, the machine can recognize the user intent that the user hopes to express. The user intent to be supplemented related to the user information in the buried point data is supplemented into the initial user intent to obtain the user intent to be processed, so as to ensure the integrity and accuracy of the user intent. According to the preset service type, calculate the weighted values of each initial configuration scheme in the set of initial configuration schemes; sort each initial configuration scheme according to the weighted values of each initial configuration scheme to obtain a sorting result; determine the configuration scheme of the intent network according to the sorting result; and feedback the configuration scheme of the intent network to the user, which is convenient for the user to quickly and accurately configure the intent network, establish the association relationship between the user intent to be processed and the configuration information of the intent network, ensure the accuracy of the configuration of the intent network, improve the accuracy of intent translation in the intent network, and improve the user experience satisfaction.

[0138] Figure 9 The flowchart showing the method for complementing user intent based on buried point data in the embodiments of the present application is as follows. As Figure 9 shown, it includes the following steps.

[0139] Step 801, obtain the initial user intent in the user information and extract the first intent feature set in the initial user intent.

[0140] In the user information, the user's intent is usually expressed in a way similar to "what I want to do" or "how I want to be". For example, Figure 10Schematic diagram showing the intent features obtained by parsing user information in an embodiment of the present application. When the user inputs "I want to open a 10GE service", it can be parsed into intent features as shown in Figure 10 . Among them, the domain = {service}; the intent (Intent) = {open service}; the semantic slot = {person, service type, layer rate, time}. The person is "Wang Moumou", the service type is "customer layer", the layer rate is "10GE", and the time is "A year B month C day" (for example, September 24, 2020). By viewing Figure 10 , the initial user intent in the user information and the first intent feature set (i.e., the semantic slot) in the initial user intent can be clearly obtained.

[0141] Step 802: Determine the intent feature set to be analyzed according to the first intent feature set and the preset full set of intent features.

[0142] For example, by searching the configuration file, it is determined that the intent feature set corresponding to the domain = {service} and the intent = {open service} is the preset full set of intent features; then, the intent features in the semantic slot (i.e., {person, service type, layer rate, time}) are compared with the preset full set of intent features to obtain the intent feature set to be analyzed. It should be noted that the number of intent features in the preset full set of intent features is greater than the number of intent features in the first intent feature set. For example, the preset full set of intent features may include 16 different intent features, while the first intent feature set only includes 4 intent features: person, service type, layer rate, and time. Then, the number of intent features in the intent feature set to be analyzed is 12.

[0143] Step 803: Analyze the event metadata and user metadata in the buried point data to obtain the buried point information.

[0144] Among them, the event metadata is data used to represent the user's behavior, the user metadata is data used to represent the user's attribute characteristics, and the buried point information includes a buried point intent feature set and a buried point event set. The buried point intent feature set includes buried point intent features.

[0145] Step 804: Determine whether there is a matching event in the buried point event set that matches the preset mapping relationship.

[0146] Among them, the preset mapping relationship is the mapping relationship between the preset event and the preset intent feature. When it is determined that there is an event in the buried point event set that is the same as the preset event, it means that there is a matching event in the buried point event set that matches the preset mapping relationship; otherwise, when it is determined that there is no event in the buried point event set that is the same as the preset event, it means that there is no matching event in the buried point event set that matches the preset mapping relationship.

[0147] It should be noted that in the case where a matching event that matches the preset mapping relationship exists in the set of buried point events, step 805 is executed; otherwise, in the case where no matching event that matches the preset mapping relationship exists in the set of buried point events, it means that the initial user intention obtained at this time is sufficient to represent the user's expected information, and there is no need to add some intention features in the buried point data to the initial user intention. At this time, step 811 is executed to determine the user intention to be processed based on the user's initial user intention.

[0148] Step 805: Extract the matching events in the set of buried point events to generate a set of matching events.

[0149] Step 806: Extract the second intention features from the set of buried point intention features according to the matching events in the set of matching events.

[0150] Among them, the second intention features are the buried point intention features corresponding to the matching events. For example, the second intention features can be intention features such as {source end, destination end, protection type, intelligence level}. Among them, the source end represents the source address of the service initiation end or the identifier of the source end device, etc.; the destination end represents the destination address of the service receiving end or the identifier of the destination end device, etc.; the intelligence level represents the degree of possession of capabilities such as dynamically allocating and flexibly controlling bandwidth, quickly generating services, providing protection and recovery, and dynamically expanding capacity in the intention network; the protection type represents the protection method for the services in the intention network, such as linear protection, ring network protection, etc.

[0151] Step 807: Search for the set of intention features to be analyzed according to the second intention features to obtain the search result.

[0152] Step 808: In the case where it is determined that there are intention features in the set of intention features to be analyzed that are the same as the second intention features in the search result, calculate the feature values corresponding to the second intention features.

[0153] For example, through the algorithm of machine learning, calculate the feature values of the second intention features to obtain the feature values corresponding to the second intention features. Or, through the intention features of the source end and the intention features of the destination end, perform data analysis on the corresponding network element configuration events and create network element link events to determine the feature values corresponding to the second intention features. For the intention features for which the feature values cannot be obtained through calculation, perform data cleaning and discard them.

[0154] Step 809: Determine the user intention to be supplemented according to the second intention features and their corresponding feature values.

[0155] Step 810: Fill the user intention to be supplemented into the initial user intention to obtain the user intention to be processed.

[0156] For example, the second intent feature corresponding to the user intent to be supplemented and its corresponding feature value are filled into the semantic slots corresponding to the initial user intent, generating the semantic slots corresponding to the user intent to be processed = {person, business type, layer rate, source end, sink end, protection type, intelligence level, time}. The user intent to be processed is determined through each intent feature in the semantic slots corresponding to the user intent to be processed, so as to ensure the integrity and accuracy of the user intent.

[0157] It should be noted that when the voice information or text information input by the user is ambiguous (for example, the text information input by the user is "I want to establish a protected service"), by analyzing the text information input by the user, multiple intent features corresponding to the semantic feature "protected" will be obtained (for example, linear protection or ring network protection, etc.). Through multiple similar intent features, it is impossible to clearly know which type of protected service the user exactly needs to establish, nor can we know the corresponding business type (for example, the user may need to establish a traditional protected service, or may need to establish an intelligent recovery type protected service). Therefore, when parsing the text information input by the user and there are multiple similar intent features in the obtained semantic slots, it is necessary to conduct a detailed analysis on the user's operation events (for example, the content of the page browsed by the user, and / or, the object operated on) to determine the corresponding relationship between the operation events and the intent features, so as to further determine the user intent to be supplemented according to this corresponding relationship.

[0158] Step 811, output the user intent to be processed.

[0159] In this embodiment, by comparing the initial user intent with the preset complete set of intent features, the set of intent features to be analyzed is determined. Then, the buried point intent features corresponding to the matching events that match the preset mapping relationship in the buried point event set are extracted (i.e., the second intent features); the second intent features are supplemented into the set of intent features corresponding to the initial user intent to generate the user intent to be processed, ensuring the integrity of the user intent. Based on the user intent to be processed, the configuration scheme of the intent network is determined, improving the translation accuracy of the intent network.

[0160] Figure 11 The flowchart showing the method of translating the user intent to be processed into the configuration scheme of the intent network in the embodiments of the present application is as follows. As Figure 11 shown, it includes the following steps.

[0161] Step 1001, obtain the user intent to be processed.

[0162] Among them, the user intent to be processed includes the initial user intent and the user intent to be supplemented. To ensure the integrity of the user intent and improve the accurate understanding of the user intent.

[0163] Step 1002: Determine a set of initial configuration solutions based on the knowledge graph, a preset inference algorithm, and the user intent to be processed.

[0164] For example, according to the values of domain and Intent in the frame semantics of the user intent, partition the knowledge graph to generate a training set and a test set; use the user intent to be processed as an input parameter for training to obtain a training result; evaluate and score the training result according to a scoring function (such as a loss function, etc.) to obtain the scores of each training result, and then obtain the ranking corresponding to each training result. The set of initial configuration solutions can be determined according to this ranking.

[0165] It should be noted that the scoring function is used to characterize the association degree between the entities and relationships in the triple. Through this scoring function, the entity vector and the relationship vector can be continuously adjusted to make the training result more in line with the user intent to be processed.

[0166] Step 1003: Extract the configuration parameter information and the initial event information corresponding to the configuration parameter information of each initial configuration solution in sequence.

[0167] Step 1004: Determine the target event metric information corresponding to the preset business type according to the preset business type.

[0168] For example, when the preset business type is business issuance, the target event metric information corresponding to business issuance can be the conversion rate of successful business issuance; when the preset business type is alarm reduction, the target event metric information corresponding to alarm reduction can be the alarm retention rate. By determining the target event metric information corresponding to different preset business types, the specific expected standards can be clarified, which is convenient for subsequent processing and improves processing efficiency.

[0169] Step 1005: Calculate the weighted value of each initial configuration solution according to the target event metric information and the initial event information.

[0170] For example, in the case where the preset business type is determined to be business issuance, determine the conversion rate of successful business issuance of each initial configuration solution according to the first target event metric information corresponding to business issuance and each initial event information; determine the weighted value of each initial configuration solution according to the conversion rate of successful business issuance of each initial configuration solution.

[0171] When the domain is "business" and the preset business type is business distribution, first establish a funnel analysis model based on various events in the business distribution process, the conversion cycle of business distribution, the preset filtering conditions, the user to be analyzed (for example, an Optical Transport Network (OTN) user), and the funnel analysis algorithm. According to this funnel analysis model, analyze the first target event metric information corresponding to business distribution and each initial event information to determine the conversion rate of successful business distribution for each initial configuration plan, and then reflect the conversion trend through this conversion rate.

[0172] For example, when it is determined that the preset business type is alarm reduction, determine the alarm retention rate of each initial configuration plan based on the second target event metric information corresponding to alarm reduction and each initial event information; determine the weighted value of each initial configuration plan based on the alarm retention rate of each initial configuration plan. Among them, alarm reduction includes alarm reduction levels, and different alarm reduction levels correspond to different alarm retention rates. For example, the alarm reduction levels are divided into level one, level two, and level three. The lower the level number, the higher the corresponding alarm retention rate. For example, when the alarm reduction level is level one, the corresponding alarm retention rate is 30%; when the alarm reduction level is level two, the corresponding alarm retention rate is 60%; when the alarm reduction level is level three, the corresponding alarm retention rate is 75%. That is, users have different tolerances for alarm reduction.

[0173] Step 1006, determine the configuration plan of the intent network based on the weighted value of each initial configuration plan and the set of initial configuration plans.

[0174] For example, re - sort each initial configuration plan in the set of initial configuration plans according to the first sorting result and the weighted value of each initial configuration plan to obtain a second sorting result; determine the configuration plan of the intent network according to the second sorting result.

[0175] In this embodiment, since it is difficult for a static knowledge graph to describe the distribution and changes of user intentions, during the knowledge reasoning process of the knowledge graph, the operation behaviors of different users and the behavior differences of the same user at different times cannot reflect their impacts on the configuration scheme of the intention network. There is a deviation between the configuration scheme of the intention network that the user expects to obtain and the actually obtained configuration scheme of the intention network. By sequentially extracting the configuration parameter information of each initial configuration scheme and the initial event information corresponding to the configuration parameter information; determining the target event metric information corresponding to the preset service type according to the preset service type; calculating the weighted values of each initial configuration scheme based on the target event metric information and the initial event information, and sorting the initial configuration schemes again according to the weighted values, a more accurate configuration scheme of the intention network can be obtained. This makes the configuration scheme of the intention network that the user expects to obtain and the actually obtained configuration scheme of the intention network more consistent, enables the actually obtained configuration scheme of the intention network to be better corrected, improves the correlation accuracy between the user intention to be processed and the configuration information of the intention network, ensures the configuration accuracy of the intention network, and improves the accuracy of intention translation in the intention network, so as to improve the user experience satisfaction.

[0176] Figure 12 The structural diagram showing an exemplary hardware architecture of an electronic device capable of implementing the intention processing method and apparatus according to an embodiment of the present application.

[0177] As Figure 12 shown, the electronic device 1100 includes an input device 1101, an input interface 1102, a central processing unit 1103, a memory 1104, an output interface 1105, and an output device 1106. Among them, the input interface 1102, the central processing unit 1103, the memory 1104, and the output interface 1105 are connected to each other through a bus 1107. The input device 1101 and the output device 1106 are respectively connected to the bus 1107 through the input interface 1102 and the output interface 1105, and then connected to other components of the electronic device 1100.

[0178] Specifically, the input device 1101 receives external input information and transmits the input information to the central processing unit 1103 through the input interface 1102; the central processing unit 1103 processes the input information based on the computer-executable instructions stored in the memory 1104 to generate output information, temporarily or permanently stores the output information in the memory 1104, and then transmits the output information to the output device 1106 through the output interface 1105; the output device 1106 outputs the output information to the outside of the electronic device 1100 for the user to use. The electronic device 1100 can be used to execute the intention processing method described in the above embodiment.

[0179] In one embodiment, Figure 12The electronic device shown can be implemented as an intent processing system, which may include: a memory configured to store a program; a processor configured to run the program stored in the memory to execute the intent processing method described in the above embodiments.

[0180] As described above, the above are only exemplary embodiments of the present application and are not used to limit the protection scope of the present application. Generally, various embodiments of the present application can be implemented in hardware or dedicated circuits, software, logic, or any combination thereof. For example, some aspects can be implemented in hardware, while other aspects can be implemented in firmware or software that can be executed by a controller, a microprocessor, or other computing devices, although the present application is not limited thereto.

[0181] Embodiments of the present application can be implemented by a data processor of a mobile device executing computer program instructions, for example, in a processor entity, or by hardware, or by a combination of software and hardware. The computer program instructions can be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-related instructions, microcode, firmware instructions, status setting data, or source code or object code written in any combination of one or more programming languages.

[0182] Any block diagram of a logical process in the drawings of the present application can represent program steps, or can represent interconnected logical circuits, modules, and functions, or can represent a combination of program steps and logical circuits, modules, and functions. The computer program can be stored in a memory. The memory can have any type suitable for the local technical environment and can be implemented using any suitable data storage technology, such as, but not limited to, read-only memory (ROM), random access memory (RAM), optical memory devices and systems (digital versatile disc DVD or CD optical disc), etc. The computer-readable medium can include a non-transitory storage medium. The data processor can be any type suitable for the local technical environment, such as, but not limited to, a general-purpose computer, a dedicated computer, a microprocessor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), and a processor based on a multi-core processor architecture.

[0183] By way of exemplary and non-limiting examples, a detailed description of the exemplary embodiments of the present application has been provided above. However, considering the accompanying drawings and the claims, various modifications and adjustments to the above embodiments will be obvious to those skilled in the art without departing from the scope of the present application. Therefore, the proper scope of the present application will be determined according to the claims.

Claims

1. An intention processing method, characterized in that The method includes: Analyze the obtained buried point data and the initial user intention to determine the user intention to be processed, where the initial user intention is used to represent the original needs of the user; Determine the configuration scheme of the intention network according to the user intention to be processed; The analyzing the obtained buried point data and the initial user intention to determine the user intention to be processed includes: Extract the first intention feature set in the initial user intention; Determine the intention feature set to be analyzed according to the first intention feature set and the preset full set of intention features; Determine the user intention to be supplemented according to the buried point data and the intention feature set to be analyzed; Fill the user intention to be supplemented into the initial user intention to obtain the user intention to be processed; The determining the user intention to be supplemented according to the buried point data and the intention feature set to be analyzed includes: Analyze the event metadata and user metadata in the buried point data to obtain buried point information, where the event metadata is data used to represent the behavior of the user, the user metadata is data used to represent the attribute characteristics of the user, the buried point information includes a buried point intention feature set and a buried point event set, and the buried point intention feature set includes buried point intention features; In the case where it is determined that there is a matching event in the buried point event set that matches the preset mapping relationship, extract the matching event in the buried point event set to generate a matching event set, where the preset mapping relationship is a mapping relationship between a preset event and a preset intention feature; Determine the user intention to be supplemented according to the matching event set, the buried point intention feature set, and the intention feature set to be analyzed.

2. The method according to claim 1, wherein The determining the user intention to be supplemented according to the matching event set, the buried point intention feature set, and the intention feature set to be analyzed includes: Extract the second intention feature from the buried point intention feature set according to the matching event in the matching event set, where the second intention feature is the buried point intention feature corresponding to the matching event; Search for the intention feature set to be analyzed according to the second intention feature to obtain a search result; In the case where it is determined that the search result is that there is an intention feature in the intention feature set to be analyzed that is the same as the second intention feature, calculate the feature value corresponding to the second intention feature; Determine the user intention to be supplemented according to the second intention feature and its corresponding feature value.

3. The method according to claim 1, wherein The filling the user intention to be supplemented into the initial user intention to obtain the user intention to be processed includes: Extract the second intention feature in the user intention to be supplemented; Calculate the feature value corresponding to the second intention feature; Fill the second intention feature and its corresponding feature value into the first intention feature set in the initial user intention to generate an intention feature set to be processed; Obtain the event to be processed corresponding to the intention feature in the intention feature set to be processed; Determine the user intention to be processed according to the intention feature set to be processed and the event to be processed.

4. The method according to claim 1, wherein Determining a configuration solution for the intent network according to the user intent to be processed includes: Determining a set of initial configuration solutions according to the knowledge graph, the preset inference algorithm, and the user intent to be processed; Calculating the weighted values of each initial configuration solution in the set of initial configuration solutions according to the preset service type; Determining the configuration solution of the intent network according to the weighted values of each initial configuration solution and the set of initial configuration solutions.

5. The method according to claim 4, characterized in that, Determining a set of initial configuration solutions according to the knowledge graph, the preset inference algorithm, and the user intent to be processed includes: Generating a set of first configuration solutions based on the knowledge graph and the preset inference algorithm, where the knowledge graph includes entity elements, attribute elements, and relationship elements; Training the solutions in the set of first configuration solutions according to the user intent to be processed to obtain a training result; Sorting the training results according to a scoring function to obtain a first sorting result, where the scoring function is a function determined according to the correlation degree between the entity elements and the relationship elements; Determining the set of initial configuration solutions according to the first sorting result.

6. The method according to claim 5, wherein Determining the configuration solution of the intent network according to the weighted values of each initial configuration solution and the set of initial configuration solutions includes: Re-sorting each initial configuration solution in the set of initial configuration solutions according to the first sorting result and the weighted values of each initial configuration solution to obtain a second sorting result; Determining the configuration solution of the intent network according to the second sorting result.

7. The method according to claim 4, wherein Calculating the weighted values of each initial configuration solution in the set of initial configuration solutions according to the preset service type includes: Sequentially extracting the configuration parameter information of each initial configuration solution and the initial event information corresponding to the configuration parameter information; Determining the target event metric information corresponding to the preset service type according to the preset service type; Calculating the weighted values of each initial configuration solution according to the target event metric information and the initial event information.

8. The method according to claim 7, wherein Calculating the weighted values of each initial configuration solution according to the target event metric information and the initial event information includes: When it is determined that the preset service type is service issuance, determining the conversion rate of successful service issuance of each initial configuration solution according to the first target event metric information corresponding to the service issuance and each initial event information; Determining the weighted values of each initial configuration solution according to the conversion rate of successful service issuance of each initial configuration solution; When it is determined that the preset service type is alarm reduction, determining the alarm retention rate of each initial configuration solution according to the second target event metric information corresponding to the alarm reduction and each initial event information; Determining the weighted values of each initial configuration solution according to the alarm retention rate of each initial configuration solution.

9. The method according to claim 1, wherein Before the step of analyzing the obtained buried point data and the initial user intent to determine the user intent to be processed, it further includes: Acquire user information according to the intention start event identifier and the intention end event identifier, wherein the user information includes the user's operation information and the user's input information; Determining the initial user intention based on the user's operation information and the user's input information, wherein the input information includes intention text and / or voice information, and the initial user intention is machine-recognizable intention language; The embedded point data is obtained from the data server based on the intention start event identifier and the embedded point port information of the data server.

10. The method according to claim 1, wherein After the step of determining a configuration scheme of the intent-based network based on the user intent to be processed, the method further includes: Feedback the configuration plan of the intent network to the user; Obtaining modification parameters determined by the user based on the configuration scheme of the intent-based network; The configuration scheme of the intentional network is updated according to the modification parameters and the configuration scheme of the intentional network, where the configuration scheme of the intentional network includes network configuration parameters.

11. The method according to claim 10, wherein After the step of updating the configuration scheme of the INN based on the modified parameters and the configuration scheme of the INN, the method further includes: Configuring the intent-based network according to the updated configuration scheme of the intent-based network; When it is determined that the intent network is configured successfully, the cached initial user intent and the buried data are deleted.

12. The method according to any one of claims 1 to 11, characterized in that, The buried point data is data extracted from the log buried point information.

13. An intention processing device, characterized in that, It includes: An analysis module is used to analyze the acquired tracking data and initial user intent to determine the user intent to be processed, wherein the initial user intent is used to represent the user's original needs; A processing module, configured to determine a configuration scheme of the intent-based network based on the user intent to be processed; The analysis module includes: an extraction submodule for extracting a first set of intent features from the initial user intent; a first determination submodule for determining an intent feature set to be analyzed based on the first set of intent features and a full set of preset intent features; a second determination submodule for determining a user intent to be supplemented based on the embedded data and the set of intent features to be analyzed; and a filling submodule for filling the user intent to be supplemented into the initial user intent to obtain a user intent to be processed; The second determination submodule is used to analyze the event metadata and user metadata in the buried point data to obtain buried point information, wherein the event metadata is data used to characterize the user's behavior, the user metadata is data used to characterize the user's attribute characteristics, and the buried point information includes a buried point intention feature set and a buried point event set, and the buried point intention feature set includes a buried point intention feature; when it is determined that there is a matching event that matches the preset mapping relationship in the buried point event set, the matching event in the buried point event set is extracted to generate a matching event set, wherein the preset mapping relationship is a mapping relationship between the preset event and the preset intention feature; based on the matching event set, the buried point intention feature set and the intention feature set to be analyzed, the user intention to be supplemented is determined.

14. An electronic device comprising: one or more processors; A memory storing one or more programs which, when executed by the one or more processors, cause the one or more processors to implement the intended processing method according to any one of claims 1-12.

15. A readable storage medium, characterized in that, The readable storage medium stores a computer program which, when executed by a processor, implements the intended processing method according to any one of claims 1-12.

Images