Intention instruction generation method and device, equipment and storage medium

Abstract

The invention provides an intention instruction generation method and device, equipment and a storage medium, relates to the technical field of computers, and is used for solving the problem of relatively low intention recognition accuracy. The method comprises the steps that firstly, a type keyword is determined from an initial intention instruction, and the type keyword is used for reflecting the type of the initial intention instruction; thirdly, classifying the initial intention instruction based on the type keyword to obtain a target classification result; and determining a target template corresponding to the target classification result based on the target classification result, and generating a target intention instruction based on the target template.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular, to a method, apparatus, device, and storage medium for generating intent instructions. Background Art

[0002] With the development of artificial intelligence technology, the human-computer dialogue system has become a key technology in the field of artificial intelligence in different application scenarios. For example, in the operation and management scenario of the computing power network, the human-computer dialogue system can identify the user's intent to improve the management efficiency of network resources.

[0003] For the operation and management scenario of the computing power network, the traditional method of identifying the user's intent has low accuracy in intent recognition and data formatting, resulting in limited intelligence and inability to perform corresponding operations on the network according to the user's intent. Therefore, how to improve the accuracy of intent recognition has become an urgent technical problem to be solved. Summary of the Invention

[0004] This application provides a method, apparatus, device, and storage medium for generating intent instructions to solve the problem of low accuracy in intent recognition.

[0005] To achieve the above object, this application adopts the following technical solutions:

[0006] In a first aspect, this application provides a method and system for generating intent instructions. The method includes: an intent instruction generation device (hereinafter referred to as the "generation device") determines type keywords from the initial intent instruction, where the type keywords are used to reflect the category of the initial intent instruction. Classify the initial intent instruction based on the type keywords to obtain a target classification result. Based on the target classification result, determine a target template corresponding to the target classification result, and generate a target intent instruction based on the target template.

[0007] The technical solutions provided by this application at least bring the following beneficial effects: By extracting type keywords that can reflect the core intent from the initial intent instruction input by the user, the initial intent instruction can be classified to obtain a target classification result. Then, determine the corresponding target template according to the target classification result, and generate a target intent instruction based on the target template. In this way, by extracting type keywords from the user input and classifying the initial intent instruction based on these keywords, the actual needs of the user can be understood more accurately, and the accuracy of intent recognition can be improved. Moreover, this method can achieve automated processing, which can improve the response speed and processing efficiency of the system.

[0008] Optionally, the method of "classifying the initial intent instruction based on the type keyword to obtain the target classification result" includes: constructing first intent information based on the type keyword; obtaining second intent information, where the second intent information is the intent information constructed based on the type keyword in the historical intent instruction; and obtaining the target classification result based on the first intent information and the second intent information.

[0009] Optionally, the method of "obtaining the target classification result based on the first intent information and the second intent information" includes: determining the intent similarity between the first intent information and the second intent information; if the intent similarity is less than or equal to the preset intent similarity threshold, classifying the first intent information to obtain the target classification result; if the intent similarity is greater than the preset intent similarity threshold, obtaining the first classification result corresponding to the second intent information and using the first classification result as the target classification result.

[0010] Optionally, the method of "determining the target template corresponding to the target classification result based on the target classification result" includes: determining at least one first similarity, where the first similarity is the similarity between the target classification result and the first preset classification result in the preset classification set; the preset classification set includes at least one preset first-level classification result, the preset first-level classification result includes at least one preset second-level classification result, and the first preset classification result is constructed by one preset first-level classification result and one preset second-level classification result; one first similarity corresponds to one first preset classification result; if there is a first target similarity greater than the preset similarity threshold among the at least one first similarity, using the first template corresponding to the first target second-level classification result as the target template, where the first target second-level classification result is the preset second-level classification result in the first preset classification result corresponding to the first target similarity.

[0011] Optionally, the method further includes: if all of the at least one first similarity are less than or equal to the preset similarity threshold, determining at least one second similarity, where the second similarity is the similarity between the target first-level classification result and the preset first-level classification result, and one second similarity corresponds to one preset first-level classification result; and determining the target template corresponding to the target classification result according to the second similarity.

[0012] Optionally, the method of "determining the target template corresponding to the target classification result according to the second similarity" includes: If there is a second target similarity greater than the preset similarity threshold in at least one second similarity, determine the second target first-level classification result, and the second target first-level classification result is the preset first-level classification result corresponding to the second target similarity. Determine at least one third similarity, where the third similarity is the similarity between the target second-level classification result and the second target second-level classification result, and the second target second-level classification result is at least one preset second-level classification result included in the second target first-level classification result. If there is a third target similarity greater than the preset similarity threshold in at least one third similarity, use the second template corresponding to the third target second-level classification result as the target template, and the third target second-level classification result is the second target second-level classification result with the third similarity greater than the preset similarity threshold.

[0013] Optionally, the method of "determining the target template corresponding to the target classification result according to the second similarity" further includes: If at least one second similarity is less than or equal to the preset similarity threshold, determine at least one fourth similarity, where the fourth similarity is the similarity between the target second-level classification result and the preset second-level classification result, and one fourth similarity corresponds to one preset second-level classification result. If there is a fourth target similarity greater than the preset similarity threshold in at least one fourth similarity, use the third template corresponding to the fourth target second-level classification result as the target template, and the fourth target second-level classification result is the preset second-level classification result corresponding to the fourth target similarity.

[0014] In a second aspect, the present application provides a device for generating an intent instruction, and the device includes: an acquisition module and a processing module.

[0015] The processing module is configured to determine a type keyword from the initial intent instruction, and the type keyword is used to reflect the category of the initial intent instruction. The processing module is further configured to classify the initial intent instruction based on the type keyword to obtain a target classification result. The processing module is further configured to determine the target template corresponding to the target classification result based on the target classification result, and generate a target intent instruction based on the target template.

[0016] Optionally, the processing module is specifically configured to construct first intent information based on the type keyword. The acquisition module is configured to acquire second intent information, where the second intent information is intent information constructed based on the type keyword in the historical intent instruction. The processing module is further configured to obtain a target classification result based on the first intent information and the second intent information.

[0017] Optionally, the processing module is specifically configured to determine the intention similarity between the first intention information and the second intention information. If the intention similarity is less than or equal to a preset intention similarity threshold, the processing module is further configured to classify the first intention information to obtain a target classification result. If the intention similarity is greater than the preset intention similarity threshold, the processing module is further configured to obtain the first classification result corresponding to the second intention information and use the first classification result as the target classification result.

[0018] Optionally, the processing module is specifically configured to determine at least one first similarity, where the first similarity is the similarity between the target classification result and the first preset classification result in the preset classification set. The preset classification set includes at least one preset first-level classification result, the preset first-level classification result includes at least one preset second-level classification result, and the first preset classification result is constructed from one preset first-level classification result and one preset second-level classification result. One first similarity corresponds to one first preset classification result. If there is a first target similarity greater than the preset similarity threshold among the at least one first similarity, the processing module is further configured to use the first template corresponding to the first target second-level classification result as the target template, where the first target second-level classification result is the preset second-level classification result in the first preset classification result corresponding to the first target similarity.

[0019] Optionally, if all of the at least one first similarity are less than or equal to the preset similarity threshold, the processing module is specifically configured to determine at least one second similarity, where the second similarity is the similarity between the target first-level classification result and the preset first-level classification result, and one second similarity corresponds to one preset first-level classification result. The processing module is further configured to determine the target template corresponding to the target classification result according to the second similarity.

[0020] Optionally, if there is a second target similarity greater than the preset similarity threshold among the at least one second similarity, the processing module is specifically configured to determine the second target first-level classification result, where the second target first-level classification result is the preset first-level classification result corresponding to the second target similarity. The processing module is further configured to determine at least one third similarity, where the third similarity is the similarity between the target second-level classification result and the second target second-level classification result, and the second target second-level classification result is at least one of the preset second-level classification results included in the second target first-level classification result. If there is a third target similarity greater than the preset similarity threshold among the at least one third similarity, the processing module is further configured to use the second template corresponding to the third target second-level classification result as the target template, where the third target second-level classification result is the second target second-level classification result with the third similarity greater than the preset similarity threshold.

[0021] Optionally, if at least one second similarity is less than or equal to a preset similarity threshold, the processing module is specifically configured to determine at least one fourth similarity, where the fourth similarity is the similarity between the target secondary classification result and the preset secondary classification result, and one fourth similarity corresponds to one preset secondary classification result. If there is a fourth target similarity greater than the preset similarity threshold among the at least one fourth similarity, the processing module is further configured to use the third template corresponding to the fourth target secondary classification result as the target template, where the fourth target secondary classification result is the preset secondary classification result corresponding to the fourth target similarity.

[0022] In a third aspect, the present application provides a device for generating intent instructions. The device includes a processor and a memory. The processor is coupled to the memory. The memory is used to store one or more programs, and the one or more programs include computer execution instructions. When the device for generating intent instructions runs, the processor executes the computer execution instructions stored in the memory to implement the method for generating intent instructions described in the first aspect or any optional one of the first aspect.

[0023] In a fourth aspect, the present application provides a computer-readable storage medium. Instructions are stored in the computer-readable storage medium. When the instructions run on a computer, the computer is caused to execute the method for generating intent instructions described in the first aspect or any optional one of the first aspect.

[0024] In a fifth aspect, the present application provides a computer program product, which is applied to a server. The computer program product includes computer instructions. When the computer instructions run on the server, the server implements the method for generating intent instructions described in the first aspect or any optional one of the first aspect.

[0025] In the above solutions, the technical problems that can be solved and the technical effects achieved by the intent instruction generation device, device, computer storage medium, or computer program product can be referred to the technical problems solved and technical effects in the first aspect above, and will not be elaborated here. Description of the Drawings

[0026] Figure 1 It is a schematic diagram of a communication system provided by an embodiment of the present application;

[0027] Figure 2 It is a schematic diagram of a device for the method of generating intent instructions provided by an embodiment of the present application;

[0028] Figure 3 It is a schematic diagram of an example of a category of common intents provided by an embodiment of the present application;

[0029] Figure 4 It is a schematic diagram of an example of a common intent category and a filling model provided by an embodiment of the present application;

[0030] Figure 5 A flowchart of a method for generating an intent instruction provided by an embodiment of the present application;

[0031] Figure 6 A flowchart of another method for generating an intent instruction provided by an embodiment of the present application;

[0032] Figure 7 A structural schematic diagram of a device for generating an intent instruction provided by an embodiment of the present application;

[0033] Figure 8 A structural schematic diagram of a device for generating an intent instruction provided by an embodiment of the present application;

[0034] Figure 9 A conceptual partial view of a computer program product provided by an embodiment of the present application. Detailed implementation manners

[0035] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0036] In this article, the character " / " generally indicates an "or" relationship between the associated objects before and after. For example, A / B can be understood as A or B.

[0037] The terms "first" and "second" in the description and claims of the present application are used to distinguish different objects, rather than to describe a specific order of the objects.

[0038] In addition, the terms "including" and "having" and any variations thereof mentioned in the description of the present application are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or modules is not limited to the listed steps or modules, but optionally further includes other unlisted steps or modules, or optionally further includes other steps or modules inherent to these processes, methods, products, or devices.

[0039] In addition, in the embodiments of the present application, words such as "exemplary" or "for example" are used to represent examples, illustrations, or explanations. Any embodiment or design solution described as "exemplary" or "for example" in the present application should not be construed as being more preferred or having more advantages than other embodiments or design solutions. Rather, the use of words such as "exemplary" or "for example" is intended to present concepts in a specific manner.

[0040] Before introducing the method for generating intention instructions provided in the embodiments of the present application in detail, the implementation environment and application scenarios of the embodiments of the present application will be introduced first.

[0041] First, the application scenarios of the embodiments of the present application will be introduced.

[0042] With the development of artificial intelligence technology, the human-machine dialogue system has become a key technology in the field of artificial intelligence in different application scenarios. For example, in the scenario of computing power network operation and management, the human-machine dialogue system can identify the intentions of users to improve the management efficiency of network resources.

[0043] Currently, in order to identify the intentions of users, sample text data corresponding to at least two intention prediction networks can be obtained respectively, where the sample text data corresponding to each intention prediction network is different. Then, the sample text data is input into a general feature extraction network to obtain sample text features, and the sample text features are input into the corresponding intention prediction network to obtain the sample intentions predicted by the intention prediction network. Next, the intention prediction network is trained according to the sample intentions predicted by the intention prediction network and the intention supervision data of the sample text data. Then, the general feature extraction network is trained according to the sample intentions predicted by at least two intention prediction networks and the intention supervision data of the sample text data.

[0044] However, the existing computing power network operation and management system has a low degree of intelligence, lacks intention recognition and formatting capabilities, and has low accuracy in intention recognition and data formatting, resulting in limited intelligence and inability to perform corresponding operations on the network according to user intentions, resulting in low automation.

[0045] In summary, how to improve the accuracy of intention recognition has become an urgent technical problem to be solved.

[0046] To solve the above problems, the embodiments of the present application provide a method for generating intention instructions. The method for generating intention instructions provided in the embodiments of the present application is applied to the scenario of intention recognition. By extracting type keywords that can reflect the core intention from the initial intention instructions input by the user, the initial intention instructions can be classified to obtain the target classification result. Then, the corresponding target template is determined according to the target classification result, and the target intention instruction is generated based on the target template. In this way, by extracting type keywords from the user input and classifying the initial intention instructions based on these keywords, the actual needs of the user can be understood more accurately, and the accuracy of intention recognition can be improved. Moreover, this method can achieve automated processing, which can improve the response speed and processing efficiency of the system.

[0047] Next, the implementation environment of the embodiments of the present application will be introduced.

[0048] As shown Figure 1 in the figure, it is a schematic diagram of a communication system provided by an embodiment of the present application. The communication system may include: an acquisition device 101 and a generation device 102. Among them, the acquisition device 101 may be connected to the generation device 102 in a wired / wireless manner.

[0049] Specifically, the acquisition device 101 may be used to obtain an initial intention instruction. The acquisition device 101 may also be used to obtain intention information constructed from type keywords in historical intention instructions. And the acquisition device 101 may also send the intention information and the initial intention instruction to the generation device 102.

[0050] It should be noted that the embodiment of the present application does not limit the type of the acquisition device. For example, the acquisition device may be a voice acquisition device, or for another example, the acquisition device may be a text acquisition device, or for another example, the acquisition device may be an image and video acquisition device.

[0051] The generation device 102 may be used for classification and generation of instructions. The generation device 102 may be used to receive the intention instructions collected by the acquisition device 101, and determine keywords that can reflect the category of the initial intention instruction from them. The generation device 102 may also be used to classify the intention instructions. The generation device 102 may also be used to determine a template corresponding to the classification result based on the classification result, and generate an intention instruction based on the template.

[0052] In the embodiment of the present application, the acquisition device 101 may be deployed in the generation device 102, and the generation device 102 independently completes the generation of intention instructions.

[0053] In some embodiments, as shown Figure 2 in the figure, the generation device 102 may include: an intention template library 201, an intention filling model 202, an intelligent interaction module 203, a data preprocessing module 204, an intention classification module 205, an intention category matching module 206, and an intention template filling module 207. Among them, the intention template library 201 may be connected to the intention filling model 202 in a wired / wireless manner, the intelligent interaction module 203 may be connected to the data preprocessing module 204 in a wired / wireless manner, the data preprocessing module 204 may be connected to the intention classification module 205 in a wired / wireless manner, the intention category matching module 206 may be connected to the intention classification module 205, the intention template filling module 207, and the intention template library 201 in a wired / wireless manner, and the intention template filling module 207 may be connected to the intention category matching module 206 and the intention filling model 202 in a wired / wireless manner.

[0054] The intent template library 201 can be used to provide intent categories, intent entries, and intent templates for common intents, where the intent categories and intent entries are used for intent matching. The intent categories and intent entries are constructed by means of artificial intelligence (AI) clustering and expert review. AI clustering is to cluster texts such as operation and maintenance manuals and operation instructions through an AI clustering algorithm to form preliminary common intent categories and intent entries. Then, after expert review, common intent categories and intent entries are formed. Then, they are stored in the form of a knowledge graph to form an intent template library.

[0055] As Figure 3 shown, the intent categories include first-level classifications and second-level classifications. The first-level classifications include, but are not limited to, resource management, performance guarantee, service quality management, and security management. The second-level classifications are the subordinate subdivisions of the first-level classifications. Specifically: the second-level classifications under resource management include computing power resource planning, resource allocation and scheduling, resource inventory and update, etc.; the second-level classifications under performance guarantee include performance monitoring, network optimization, fault handling, etc.; the second-level classifications under service quality management include quality index setting, customer feedback collection and processing, continuous improvement, etc.; the second-level classifications under security management include construction of a security protection system, formulation and update of security policies, security audit and emergency response, etc.

[0056] Among them, there are multiple intent entries under the second-level classifications. Computing power resource planning includes Intent Entry 1, which can be referred to as shown in 1-4 below: 1. Based on historical business data, use algorithms such as regression analysis and time series to predict the computing power requirements for the next year and update them quarterly. 2. Divide the priority of computing power requirements according to the proportion of business income and growth slope and make a visual chart. 3. Build a simple computing power simulation model, input business parameters, and quickly estimate the computing power usage in different scenarios. 4. Review the characteristics of computing power consumption during past business peaks, summarize the rules, and reserve specific computing power in advance for the next peak.

[0057] Resource allocation and scheduling includes Intent Item 2, which can be referred to as shown in 5 - 14 below: 5. Install intelligent monitoring software to capture the computing power utilization rate of the business system every 5 minutes and allocate it according to preset rules. 6. Conduct multiple rounds of stress testing on each business link in the week before a major e - commerce promotion to accurately allocate computing power during the promotion period. 7. Evaluate the effect of the intelligent scheduling algorithm monthly and optimize the algorithm parameters in combination with business feedback. 8. Set a "protection threshold" for computing power for key projects. When approaching the lower limit, give an early warning and prioritize resource replenishment. 9. Use container technology to achieve fine - grained computing power allocation and dynamically adjust according to service load. 10. Record the computing power scheduling log each time, including time, source task, target task, and scheduling volume for easy traceability. 11. When local computing power is tense, suspend or migrate low - priority background tasks to ensure key interaction computing power. 12. Analyze the business load change curve weekly and layout the adjustment of computing power allocation in advance. 13. Create a computing power allocation dashboard so that operation and maintenance personnel can intuitively see the real - time occupancy of each business. 14. For emergency and sudden tasks, open a fast - approval channel to complete computing power allocation within 15 minutes.

[0058] Resource inventory and update includes Intent Item 3, which can be referred to as shown in 15 - 19 below: 15. Conduct a full - scale inventory of computing power devices every quarter and register the brand, model, purchase time, performance parameters, etc. 16. Randomly inspect the performance of some devices monthly, compare with the nominal values, and mark devices with performance degradation. 17. List the inventory of idle devices every six months according to the idle duration and utilization rate of the devices and plan for processing. 18. Build a device update reminder model to prompt the update time based on the age and maintenance frequency. 19. Compare the energy consumption bills of new and old devices, calculate the energy - saving accounts, and facilitate the replacement.

[0059] Performance monitoring includes Intent Item 4, which can be referred to as shown in a - i below: a. Use a network probe to collect key link bandwidth, latency, and packet loss data per second and transmit it to the monitoring center. b. Set custom alarms for the software - defined networking (SDN) controller monitoring device. When the threshold is exceeded, push notifications to the operation and maintenance mobile phones. c. Generate a network performance daily report every day and visually display the fluctuations of key indicators. d. Deploy lightweight monitoring agents (a software component or program running on specific devices or nodes) at the network edge to complete the monitoring coverage. e. Check the integrity of the monitoring data every two weeks and repair abnormal missing points. f. For important business links, add backup monitoring to prevent misjudgment of single - point failures. g. Use big data tools to dig out potential risk trends in the monitoring data weekly. h. Build a network performance baseline library and set standard performance ranges according to the business. i. Real - time monitor the hardware status of network devices and correlate with performance changes.

[0060] Network optimization includes intent item five, which can be referred to as shown in the following jq: j. When bandwidth utilization exceeds 75%, trigger the link aggregation script and evaluate the effect afterwards. k. Collect user-side network experience every month and optimize routing strategies in a targeted manner. l. Evaluate cache strategies every quarter and adjust capacity and update frequency according to popularity. m. In case of local jamming, quickly deploy sniffers to locate and repair faulty links. n. During low business peak periods, fine-tune the network architecture to improve stability. o. Make a performance comparison report before and after optimization to accumulate experience. p. Customize exclusive network optimization packages for important customers. q. After the architecture is upgraded, track performance for a week to prevent new problems.

[0061] Fault handling includes intention item six, which can be referred to as shown in the following rz: r. Prepare a five-level network fault plan, determine the level within 30 seconds, and cut the backup link in 3 minutes for level I. S. Organize a 24-hour standby emergency team, and members should have mobile phones available to respond at any time. t. Review the troubleshooting within 24 hours, write a personal review, and summarize the problems. u. Simulate a serious fault scenario drill once every six months to improve emergency response capabilities. v. Build a fault knowledge base, enter phenomena, causes, and solutions for easy retrieval. w. Key equipment is equipped with redundant modules, and faults are automatically switched. x. In case of large-scale faults, initiate external announcements to inform users of the recovery time. y. Update the emergency plan regularly and adjust the process with new equipment and business. z. After the fault is repaired, track network performance for a week and check for secondary problems.

[0062] The quality indicator setting includes seven intent items, which can be referred to as follows (1)-(10): (1) Software as a service (SaaS) sets indicators such as 99.9% availability and response time of less than 2 seconds. (2) Platform as a service (PaaS) specifies monthly indicators for resource allocation accuracy and platform stability. (3) Customize indicators based on the industry. (4) Evaluate rationality half-yearly and fine-tune in combination with business. (5) Assign indicators to positions and sign a letter of responsibility. (6) Establish a benchmarking system and compare with competitors on a monthly basis to find gaps. (7) Invite experts to discuss and ensure that the indicators are scientific. (8) Set trial period indicators for new services and convert them to regular indicators after the trial period. (9) Create a visual dashboard to track compliance with the standards. (10) Set stricter standards for high-value customers.

[0063] Customer feedback collection and processing includes intent item eight, which can be referred to as follows (11)-(14): (11) Feedback entry is set up in a prominent place on the official website and application (APP), and reply is given within 24 hours. (12) Customer service hotline is on duty 24 hours a day, 7 days a week, complaint is recorded within 5 minutes and transferred within 1 hour. (13) Satisfaction survey is conducted monthly, covering 80% of active customers. (14) Social media feedback is monitored and timely interaction is carried out.

[0064] Continuous improvement includes Intention Item Nine, which can be referred to as shown in (15)-(18) below: (15) Analyze data monthly, draw a fishbone diagram to find the root causes of short - boards, and formulate a plan. (16) Optimize products and processes based on churn analysis to reduce the churn rate. (17) Establish a tracking mechanism, update progress weekly, and ensure timely completion. (18) Compare the satisfaction and net promoter score (NPS) before and after to quantify the effectiveness.

[0065] The construction of the security protection system includes Intention Item Ten, which can be referred to as shown in A - G below: A. Evaluate the firewall policy and update the rules according to new attacks and business. B. Deploy a honeypot system to divert attack data and optimize protection. C. Install intrusion detection on critical entrances and servers and update the intelligence database daily. D. Classify data regularly by level and allocate different protections. E. Conduct a penetration test on the protection system every six months to find vulnerabilities. F. Use virtualization security technology to isolate in a multi - tenant environment. G. Strengthen the authentication and encryption mechanisms for remote access.

[0066] The formulation and update of security policies include Intention Item Eleven, which can be referred to as shown in H - K below: H. Formulate policies two weeks before the launch of new services and go live after review. I. Issue a patch policy for new threats within 3 days. J. Establish a version management system to record revision details. K. Create a visualization guide for employees to refer to.

[0067] Security auditing and emergency response include Intention Item Twelve, which can be referred to as shown in L - P below: L. The emergency response team responds within 10 minutes for an event, stop losses and trace the source according to the pre - plan. M. Write a report one week after the incident is quelled, including the process, losses, and measures. N. Establish a case library to store event materials for training and reference. O. Evaluate the process efficiency, shorten the response time, and improve efficiency. P. Analyze the causes of repeated events and strengthen prevention.

[0068] In addition, the intention template library 201 also contains a number of intention templates. The intention template is a formatted intention formed according to a set template for the intention item, which is convenient for unified reading and processing of intentions. The fields of the template generally include intention category, target, object, scope, tool / method, frequency, start time, end time, recipient of execution results, etc.

[0069] Optionally, the intention categories and intention items in the intention template library 201 can be constructed by means of AI clustering and expert review. Cluster texts such as operation and maintenance manuals and operation instructions through the AI clustering algorithm to form preliminary common intention categories and intention items. Then, after expert review, form common intention categories and intention items. Then, store them in the form of a knowledge graph to form an intention template library.

[0070] The intention filling model 202 can be used to fill the intention template. The intention filling model 202 can be constructed according to the secondary classification, and there is an intention filling model under each secondary classification.

[0071] Optionally, by using the options of the intent template as tags, annotate the intent entries under the secondary classification. After the annotation is completed, use the AI classification algorithm to train the intent entries and tags to form an intent filling model. After the intent filling model is trained, input the intent entries under the secondary classification respectively, and the intent templates corresponding to the intent entries will be formed. After manual verification, the tags of the intent entries can be adjusted, and the intent filling model can be optimized. Common intent categories and filling models are as Figure 4 shown. For the secondary classification of computing power resource planning under the primary classification of resource management, it corresponds to the computing power resource planning intent filling model. For the secondary classification of resource allocation and scheduling under the primary classification of resource management, it corresponds to the resource allocation and scheduling intent filling model. For the secondary classification of resource inventory and update under the primary classification of resource management, it corresponds to the resource inventory and update intent filling model. For the secondary classification of performance monitoring under the primary classification of performance guarantee, it corresponds to the performance monitoring intent filling model. For the secondary classification of network optimization under the primary classification of performance guarantee, it corresponds to the network optimization intent filling model. For the secondary classification of fault handling under the primary classification of performance guarantee, it corresponds to the fault handling intent filling model. For the secondary classification of quality index setting under the primary classification of service quality management, it corresponds to the quality index setting intent filling model. For the secondary classification of customer feedback collection and processing under the primary classification of service quality management, it corresponds to the customer feedback collection and processing intent filling model. For the secondary classification of continuous improvement under the primary classification of service quality management, it corresponds to the continuous improvement intent filling model. For the secondary classification of security protection system construction under the primary classification of security management, it corresponds to the security protection system construction intent filling model. For the secondary classification of security policy formulation and update under the primary classification of security management, it corresponds to the security policy formulation and update intent filling model. For the secondary classification of security audit and emergency response under the primary classification of security management, it corresponds to the security audit and emergency response intent filling model.

[0072] Exemplarily, for the entry "Based on historical business data, use algorithms such as regression analysis and time series to predict the computing power demand for the next year and update it quarterly" under the secondary classification of "computing power resource planning" in the primary classification of "resource management", it can be annotated as: {Intent category: "Resource management - Computing power resource planning", Target: "Predict the computing power demand for the next year", Object: "Computing power demand", Scope: "Historical business data", Tool / Method: "Use algorithms such as regression analysis and time series", Frequency: "Default 1 time", Start time: "Immediately (default)", End time: "(1 time) default", Receiver of execution result: "User (default)"}.

[0073] The intelligent interaction module 203 can be used to receive the original intention input by the user (i.e., the initial intention instruction), convert it into text output, and output the result externally. The intelligent interaction module 203 can be used to filter the original intention to make the text content more concise and regular. For example, when inputting by voice, it will filter the ambient noise, and for another example, when inputting by text, it will clean the content input by the user to remove the noise information in the text. Among them, the content of the filtering process includes: hypertext markup language (HTML) tags, special symbols, redundant spaces, and stop words (such as meaningless words like modal particles and auxiliary words).

[0074] It should be noted that the embodiments of this application do not limit the input method for the user to input the original intention to the intelligent interaction module. For example, the input method for the user to input the original intention to the intelligent interaction module can be voice input, and for another example, the input method for the user to input the original intention to the intelligent interaction module can be text input, and for another example, the input method for the user to input the original intention to the intelligent interaction module can be gesture input.

[0075] The data preprocessing module 204 can be used to segment the input original intention, and cut the text of the original intention into multiple independent words. The data preprocessing module 204 can also be used to generate an intention keyword list according to the multiple independent words divided.

[0076] It should be noted that for the original intention in Chinese format, word segmentation can be carried out through dictionaries, statistics, or machine learning word segmentation algorithms; for the original intention in English format, word segmentation can be carried out through spaces.

[0077] Exemplarily, if the original intention is: Please optimize the performance of the fourth generation (4G) mobile communication technology base station. Then the original intention will be divided into "Please", "for", "4G", "base station", "carry out", "performance", "optimize". Then, perform part-of-speech recognition and entity annotation on each segmented word, and the data preprocessing module can identify the verb "optimize", nouns "4G", "base station", "performance", etc., to form an intention keyword list. Then the intention keyword list can be {verb1: "optimize", noun1: "base station", noun2: "performance", noun3: "4G"}.

[0078] The intent classification module 205 can be used to splice together the core keywords (i.e., type keywords) in the intent keyword list formed by the data preprocessing module 204 in a way of concatenation symbols (such as "-" or "+", etc.) (i.e., based on the type keywords, construct the first intent information). The intent classification module 205 can also be used to obtain the historical records input by the user, compare the historical intent with the original intent, and select the 20% keywords that best reflect the user intent for splicing among the historical intents with a similarity exceeding 80% (i.e., obtain the second intent information). The intent classification module 205 can also be used to input the type keywords into the pre-trained intent classification model to form an intent classification result (i.e., the target classification result), including Figure 1 level classification (i.e., the target first-level classification result) and second-level classification (i.e., the target second-level classification result).

[0079] Exemplarily, the original intent is: starting from 0:00 on January 1, 2025, use a network probe to collect key link bandwidth, delay, and packet loss data per second and transmit it to the monitoring center. Then the intent classification result is network performance guarantee - performance monitoring, where network performance guarantee is the first-level classification and performance monitoring is the second-level classification.

[0080] The intent category matching module 206 can be used to match the intent classification result with the intent categories in the intent template library 201, and perform traversal similarity calculation and sorting on the intents in the intent template library 201.

[0081] Optionally, the intent category matching module 206 can first calculate the similarity according to the intent "first-level classification - second-level classification" respectively, sort the similarities of each intent category, and select the one with the highest similarity and greater than or equal to 0.8 as the matching intent category. If each similarity is less than 0.8, it means that there is no corresponding intent category in the intent template library. At this time, calculate the similarity according to the intent "first-level classification" respectively, sort the similarities of each Figure 1 level classification, and select the one with the highest similarity and greater than or equal to 0.8 as the matching Figure 1 level classification. If each similarity is less than 0.8, it means that there is no corresponding Figure 1 level classification in the intent template library. At this time, it is necessary to call the interface of the intent template library to add a new Figure 1 level classification. Otherwise, calculate the similarity according to the "second-level classification" under the already matched Figure 1 level classification respectively, sort the similarities of each Figure 2 level classification, and select the one with the highest similarity and greater than or equal to 0.8 as the matching Figure 2 level classification. If each similarity is less than 0.8, it means that there is no corresponding Figure 2 level classification in the intent template library. If the similarities of each first-level classification in the intent matching are all less than 0.8, call the interface of the intent template library to add a new Figure 1For the primary classification, if the similarity of each secondary classification in the intention matching is less than 0.8, the interface of the intention template library is called to add a new Figure 2 primary classification.

[0082] The intention template filling module 207 can be used to call the intention filling model under the secondary classification to fill the intention text and form a formatted intention (that is, based on the target classification result, determine the target template corresponding to the target classification result, and generate a target intention instruction based on the target template).

[0083] Exemplarily, the original intention is: starting from 0:00 on January 1, 2025, use a network probe to collect the key link bandwidth, delay, and packet loss data per second and transmit it to the monitoring center. Then the formatted intention is {intention category: network performance guarantee - performance monitoring, target: collect data, object: key link, scope: bandwidth, delay, packet loss data, tool / method: network probe, frequency: per second, start time: 0:00 on January 1, 2025, end time: long term, execution result recipient: monitoring center}.

[0084] As a possible implementation, the intelligent interaction module 203 can receive the original intention input by the user, perform noise filtering, and send the filtered intention to the data preprocessing module 204. The data preprocessing module 204 can segment the intention to generate an intention keyword list and send it to the intention classification module 205. The intention classification module 205 can splice the keywords to obtain the spliced keywords (i.e., type keywords), and input them into the intention classification model for classification, and send the classification result to the intention category matching module 206 (that is, classify the initial intention instruction based on the type keywords to obtain the target classification result). The intention category matching module 206 can match the intention classification with the intention categories in the intention template library 201 and pass the matching result to the intention template filling module 207. If the matching is successful, the intention filling model outputs an intention template (i.e., the target template), and the intention template filling module 202 calls the model to output and generate a formatted template for the intention (that is, based on the target classification result, determine the target template corresponding to the target classification result, and generate a target intention instruction based on the target template). If the matching fails, the intention classification is added to the intention template library 201, and a corresponding intention filling model is generated to re-perform intention filling to output and generate a formatted template.

[0085] After introducing the application scenarios and implementation environments of the embodiments of the present application, the method for generating intention instructions provided by the embodiments of the present application will be described in detail below in combination with the above implementation environments.

[0086] The methods in the following embodiments can all be implemented in the above application scenarios and implementation environments. The embodiments of the present application will be specifically described below in combination with the accompanying drawings of the specification.

[0087] Figure 5 The flowchart of a method for generating an intent instruction provided by an embodiment of the present application. As Figure 5 shown, the method may include: S501 - S503.

[0088] S501. Determine type keywords from the initial intent instruction.

[0089] Among them, the initial intent instruction is the original intent instruction input by the user, and the type keyword is used to reflect the category of the initial intent instruction, and is the core word in the initial intent instruction that can represent the user's intent.

[0090] As a possible implementation, the generating device may interact with the base station corresponding to the built network to obtain the measurement report of the built network collected by the base station.

[0091] Exemplarily, if the initial intent instruction is: Optimize the performance of base station A to improve network quality, then the type keywords are base station, performance optimization, and network quality.

[0092] S502. Classify the initial intent instruction based on the type keywords to obtain the target classification result.

[0093] As a possible implementation, the generating device may pre - store classification rules. Then, the generating device may match the type keywords with the pre - stored classification rules to find the most suitable category. Then, the generating device may obtain the target classification result according to the matching result.

[0094] Exemplarily, if the type keywords of the initial intent instruction 1 are "base station" and "performance", and the type keywords of the initial intent instruction 2 are "base station", "performance", and "financial report". If classification rule A is that the type keywords contain "base station" and "performance", then it is classified as communication network optimization, and classification rule B is that the type keywords contain "base station", "performance", and "financial report", then it is classified as financial management. Then the target classification result corresponding to the initial intent instruction 1 is communication network optimization, and the target classification result corresponding to the initial intent instruction 1 is financial management.

[0095] In some embodiments, the generating device may construct the first intent information according to the type keywords. Then, the generating device may obtain the second intent information, where the second intent information is the intent information constructed based on the type keywords in the historical intent instructions. Then, the generating device may determine the intent similarity between the first intent information and the second intent information. When the intent similarity is less than or equal to the preset intent similarity threshold, the generating device may classify the first intent information to obtain the target classification result.

[0096] Alternatively, when the intention similarity is greater than a preset intention similarity threshold, the generating device may obtain the first classification result corresponding to the second intention information and use the first classification result as the target classification result.

[0097] Exemplarily, if the historical intention instructions include: historical intention instruction 1 and historical intention instruction 2, where the type keywords of historical intention instruction 1 are "base station", "user experience", and "network congestion", then the second intention information of historical intention instruction 1 may be base station-user experience-network congestion, and the corresponding first classification result is communication network optimization. The type keywords of historical intention instruction 2 are "base station", "coverage", and "signal quality", then the second intention information of historical intention instruction 2 may be base station-coverage-signal quality, and the corresponding first classification result is base station coverage optimization.

[0098] If the type keywords of the initial intention instruction are "base station", "performance optimization", and "network blockage", then the first intention information may be base station-performance optimization-network blockage. Moreover, the intention similarity between the first intention information and the second intention information is 0.3, the intention similarity between the first intention information and the second intention information is 0.8, and the preset intention similarity threshold is 0.6. Then, the target classification result is communication network optimization.

[0099] That is to say, the type keywords are extracted from the initial intention instruction input by the user and the first intention information is constructed. Then, the second intention information related to the type keywords is obtained from the historical intention instructions. By calculating the intention similarity between the first intention information and the second intention information, the relationship between the two is judged, and according to the relationship between the intention similarity and the preset intention similarity threshold, the target classification result is generated in different ways. In this way, the historical data can be effectively utilized to improve the classification efficiency and accuracy.

[0100] S503. Based on the target classification result, determine the target template corresponding to the target classification result, and generate a target intention instruction based on the target template.

[0101] As a possible implementation, the generating device may pre-store the preset templates corresponding to the preset classification results. The generating device may compare the target classification result with the preset classification results according to the target classification result. If the target classification result and the preset classification result match each other, the generating device may use the preset template corresponding to the preset classification result as the target template and generate a target intention instruction based on the target template.

[0102] As another possible implementation, the generating device may construct a target template for the target classification result and generate a target intention instruction based on the target template.

[0103] Exemplarily, if the initial intent instruction is: Please optimize the performance of Base Station A to solve the network congestion problem. The target classification result is communication network optimization, and the target template corresponding to the target classification result is: Optimize the [problem description] of [base station number] to improve [optimization target]. Then, the base station number is Base Station A, the problem description is the network congestion problem, and the optimization target is network optimization. The target intent instruction is: Optimize the network congestion problem of Base Station A to improve network performance.

[0104] The technical solutions provided by the above embodiments at least bring the following beneficial effects: By extracting type keywords that can reflect the core intent from the initial intent instruction input by the user, the initial intent instruction can be classified to obtain the target classification result. Then, based on the target classification result, the corresponding target template is determined, and the target intent instruction is generated based on the target template. In this way, by extracting type keywords from the user input and classifying the initial intent instruction based on these keywords, the actual needs of the user can be understood more accurately, and the accuracy of intent recognition can be improved. Moreover, this method can achieve automated processing, which can improve the response speed and processing efficiency of the system.

[0105] It should be noted that in the process where the above generation device determines the target template corresponding to the target classification result based on the target classification result and generates the target intent instruction based on the target template. The target classification result may include: the target first-level classification result and the target second-level classification result. By combining the classification results of these two levels, the generation device can perform more refined processing on the target classification result. This method can enable the generation device to more accurately match the appropriate target template, thereby generating a target intent instruction that better meets the user's needs and improving the accuracy of intent recognition processing.

[0106] In some embodiments, as Figure 6 shown, in this intent instruction generation method, S503 may include: S601 - S605.

[0107] S601. Determine at least one first similarity.

[0108] Among them, the first similarity is the similarity between the target classification result and the first preset classification result in the preset classification set.

[0109] It should be noted that the preset classification set includes at least one preset first-level classification result, the preset first-level classification result includes at least one preset second-level classification result, the first preset classification result is constructed by a preset first-level classification result and a preset second-level classification result, and one first similarity corresponds to one first preset classification result.

[0110] As a possible implementation, the generating device may pre-store a preset classification set, which contains multiple first preset classification results. The generating device may compare the similarity between the target classification result and the first preset classification results to determine at least one first similarity.

[0111] Exemplarily, the preset classification set includes the first preset classification result 1, the first preset classification result 2, and the first preset classification result 3. Among them, the first preset classification result 1 is resource management - computing power resource planning, the first preset classification result 2 is resource management - resource inventory and update, and the first preset classification result 3 is performance guarantee - network optimization. The target classification result is resource processing - resource summary and planning. The first similarity between the first preset classification result 1 and the target classification result is 0.8, the first similarity between the first preset classification result 2 and the target classification result is 0.6, and the first similarity between the first preset classification result 1 and the target classification result is 0.2.

[0112] S602. Determine whether there is a first target similarity greater than the preset similarity threshold among at least one first similarity.

[0113] Among them, the preset similarity threshold is the second largest value among at least one second similarity, and the preset similarity threshold is greater than 0.8.

[0114] That is to say, the first target similarity is the maximum value among at least one first similarity greater than 0.8.

[0115] It should be noted that the specific value of the preset similarity threshold can be flexibly set according to actual needs. This embodiment only provides an exemplary solution for reference.

[0116] As a possible implementation, the generating device may pre-store a preset similarity threshold. The generating device may compare at least one first similarity with the preset similarity threshold to determine whether there is a first target similarity among at least one first similarity.

[0117] In some embodiments, if it is determined that there is a first target similarity greater than the preset similarity threshold among at least one first similarity, the generating device may execute S603.

[0118] S603. Use the first template corresponding to the first target secondary classification result as the target template.

[0119] Among them, the first target secondary classification result is the preset secondary classification result in the first preset classification result corresponding to the first target similarity.

[0120] Combined with the above example, the first similarity between the first preset classification result 1 and the target classification result is 0.8, and the first preset classification result 1 is Resource Management - Computing Resource Planning. Then the first target similarity is 0.8, the first target secondary classification result is Computing Resource Planning, and the target template is the first template corresponding to Computing Resource Planning.

[0121] That is to say, by comparing the similarity (i.e., the first similarity) between the preset classification result formed by combining the first-level classification and the second-level classification in the preset classification set and the target classification result, the preset template corresponding to the preset classification result that is most similar to the target classification result can be used as the target template. This can ensure more accurate selection of the template and improve the efficiency of template matching.

[0122] In some other embodiments, if it is determined that at least one first similarity is less than or equal to the preset similarity threshold, the generating device may execute S604.

[0123] S604. Determine at least one second similarity.

[0124] Among them, the second similarity is the similarity between the target first-level classification result and the preset first-level classification result, and one second similarity corresponds to one preset first-level classification result.

[0125] S605. Determine the target template corresponding to the target classification result according to the second similarity.

[0126] As a possible implementation, the generating device may pre-store the preset similarity threshold. The generating device may compare at least one second similarity with the preset similarity threshold, and determine the target template corresponding to the target classification result according to the determined size relationship between the second similarity and the preset similarity threshold.

[0127] In some embodiments, when there is a second target similarity greater than the preset similarity threshold among at least one second similarity, the generating device may determine the second target first-level classification result, and the second target first-level classification result is the preset first-level classification result corresponding to the second target similarity. Then, the generating device may determine at least one third similarity. Then, the generating device may compare the third similarity with the preset similarity threshold. If there is a third target similarity greater than the preset similarity threshold among at least one third similarity, the generating device may use the second template corresponding to the third target secondary classification result as the target template.

[0128] Among them, the third similarity is the similarity between the target secondary classification result and the second target secondary classification result, and the second target secondary classification result is at least one preset secondary classification result included in the second target first-level classification result. The third target secondary classification result is the second target secondary classification result with the third similarity greater than the preset similarity threshold.

[0129] It should be noted that for the second similarity, the preset similarity threshold is the second largest value among at least one second similarity, and the preset similarity threshold is greater than 0.8. That is to say, the second target similarity is the maximum value among at least one second similarity greater than 0.8. For the third similarity, the preset similarity threshold is the second largest value among at least one third similarity. That is to say, the third target similarity is the maximum value among at least one third similarity.

[0130] It should be noted that the specific value of the preset similarity threshold can be flexibly set according to actual needs, and this embodiment only provides an exemplary solution for reference.

[0131] Exemplarily, if the preset first-level classification results include: preset first-level classification result A, preset first-level classification result B, and preset first-level classification result C, where the fourth similarity between the target first-level classification result and preset first-level classification result A is 0.5, the fourth similarity between the target first-level classification result and preset first-level classification result B is 0.85, and the fourth similarity between the target first-level classification result and preset first-level classification result C is 0.9, then the second target similarity is 0.9, and preset first-level classification result C is the second target first-level classification result. Preset first-level classification result C includes: preset second-level classification result 1, preset second-level classification result 2, and preset second-level classification result 3. Among them, the fourth similarity between the target second-level classification result and preset second-level classification result 1 is 0.4, the fourth similarity between the target second-level classification result and preset second-level classification result 2 is 0.2, and the fourth similarity between the target second-level classification result and preset second-level classification result 3 is 0.6. Then the third target similarity is 0.6, and the second template corresponding to preset second-level classification result 3 is the target template.

[0132] That is to say, if the similarity (i.e., the second similarity) between the target classification result and a certain preset first-level classification is greater than the preset threshold, then select this first-level classification as the target first-level classification. Then, calculate the similarity (i.e., the third similarity) between the target second-level classification result and all preset second-level classifications under the target first-level classification. If the similarity of a certain second-level classification is greater than the preset threshold, then associate its template as the target template. In this way, by first determining the first-level classification and then focusing on the second-level classification, the scope can be gradually narrowed to ensure that the finally selected template highly matches the target classification. And through the two-level similarity threshold judgment, irrelevant classifications are effectively filtered, improving the stability and reliability of the processing process.

[0133] This classification method based on multi-level similarity calculation not only improves the accuracy of classification, but also provides a reliable basis for subsequent target template selection and intent instruction generation.

[0134] In some other embodiments, when at least one second similarity is less than or equal to a preset similarity threshold, the generating device may determine at least one fourth similarity. Then, the generating device may compare the fourth similarity with the preset similarity threshold. If there is a fourth target similarity greater than the preset similarity threshold among at least one fourth similarity, the generating device may use the third template corresponding to the fourth target secondary classification result as the target template.

[0135] Alternatively, if at least one fourth similarity is less than or equal to the preset similarity threshold, the generating device may generate a corresponding target template based on the target classification result.

[0136] Wherein, the fourth similarity is the similarity between the target secondary classification result and the preset secondary classification result, and one fourth similarity corresponds to one preset secondary classification result. The fourth target secondary classification result is the preset secondary classification result corresponding to the fourth target similarity.

[0137] It should be noted that for the fourth similarity, the preset similarity threshold is the second largest value among at least one fourth similarity, and the preset similarity threshold is greater than 0.8. That is to say, the fourth target similarity is the maximum value among at least one fourth similarity greater than 0.8.

[0138] It should be noted that the specific value of the preset similarity threshold can be flexibly set according to actual needs, and this embodiment only provides an exemplary solution for reference.

[0139] Exemplarily, if the preset secondary classification results include: preset secondary classification result A, preset secondary classification result B, and preset secondary classification result C, wherein the fourth similarity between the target secondary classification result and preset secondary classification result A is 0.5, the fourth similarity between the target secondary classification result and preset secondary classification result B is 0.7, and the fourth similarity between the target secondary classification result and preset secondary classification result C is 0.9, then the fourth target similarity is 0.9, and the third template corresponding to preset secondary classification result C is the target template.

[0140] That is to say, if the similarities (i.e., the second similarities) between the target classification result and all preset primary classifications do not exceed the threshold, then calculate the similarities (i.e., the fourth similarities) between the target secondary classification result and all preset secondary classifications. If a certain fourth similarity is greater than the threshold, select the template associated with the corresponding secondary classification as the target template; if all fourth similarities still do not exceed the threshold, directly generate a new template based on the target classification result. In this way, in the case where primary classification matching fails, secondary classification matching can be further attempted to ensure that existing template resources are utilized as much as possible. And when existing classifications and templates cannot meet the requirements, new templates can be flexibly generated to adapt to complex or unknown scenarios.

[0141] It can be understood that if the similarity between the target classification result and all preset classifications (i.e., the first similarity) does not exceed the threshold, then calculate the similarity between the target first-level classification and each preset first-level classification (i.e., the second similarity), and determine the final target template according to the second similarity. In this way, when the first similarity fails to meet the conditions, the matching range can be expanded by calculating the second similarity, increasing the probability of finding a suitable template.

[0142] In summary, the method and device for generating intent instructions in this application can understand the information of the user's input, including information understanding methods such as analyzing the grammatical structure of the input content, understanding the true meaning of the sentence, and identifying the intent behind the user's input. Moreover, this application can classify the analyzed and understood information into corresponding intent categories and then perform formatting processing according to the set format to form intent instructions. In addition, the method and device for generating intent instructions in this application also support the context awareness function, that is, during the continuous conversation process, it will refer to the previous conversation content to understand the current input and can understand the input according to the scenario where the user is located. In this way, the device for generating intent instructions can use intelligent technology to classify, identify, and format the operation management intent, realizing the automated management of network operation intent and promoting the development of network operation management towards self-configuration, self-optimization, and self-repair. It helps to promote network operation management towards automation and intelligence.

[0143] The above mainly introduced the solution provided in the embodiment of this application from the perspective of computer devices. It can be understood that in order for a computer device to implement the above functions, it includes the corresponding hardware structure and / or software module for executing each function. Those skilled in the art should easily realize that, combined with the method steps for generating intent instructions of each example described in the embodiment of this application disclosed, this application can be implemented in the form of hardware or a combination of hardware and computer software. Whether a certain function is executed in the way of hardware or computer software driving the hardware depends on the specific application and design constraint conditions of the technical solution. Professional technicians can use different methods to implement the described function for each specific application, but such implementation should not be considered to exceed the scope of this application.

[0144] The embodiment of this application also provides a device for generating intent instructions. The device for generating intent instructions can be a computer device, or the CPU in the above computer device, or the processing module in the above computer device for generating intent instructions, or the client in the above computer device for generating intent instructions.

[0145] The embodiments of the present application may divide the generating device of the intent instruction into functional modules or functional units according to the above method examples. For example, each functional module or functional unit may be corresponding to each function, or two or more functions may be integrated into one processing module. The above integrated module may be implemented in the form of hardware, or in the form of a software functional module or functional unit. Among them, the division of modules or units in the embodiments of the present application is illustrative, and is only a logical function division. There may be other division methods in actual implementation.

[0146] As Figure 7 shown, it is a schematic structural diagram of a generating device of an intent instruction provided by an embodiment of the present application. The generating device of the intent instruction is used to execute Figure 5 , Figure 6 the intent instruction generating method shown. The generating device 700 of the intent instruction may include: an obtaining module 701 and a processing module 702.

[0147] The processing module 702 is configured to determine a type keyword from the initial intent instruction, where the type keyword is used to reflect the category of the initial intent instruction. The processing module 702 is further configured to classify the initial intent instruction based on the type keyword to obtain a target classification result. The processing module 702 is further configured to determine a target template corresponding to the target classification result based on the target classification result, and generate a target intent instruction based on the target template.

[0148] Optionally, the processing module 702 is specifically configured to construct first intent information based on the type keyword. The obtaining module 701 is configured to obtain second intent information, where the second intent information is intent information constructed based on the type keyword in the historical intent instruction. The processing module 702 is further configured to obtain a target classification result based on the first intent information and the second intent information.

[0149] Optionally, the processing module 702 is specifically configured to determine the intent similarity between the first intent information and the second intent information. If the intent similarity is less than or equal to a preset intent similarity threshold, the processing module 702 is further configured to classify the first intent information to obtain a target classification result. If the intent similarity is greater than the preset intent similarity threshold, the processing module 702 is further configured to obtain a first classification result corresponding to the second intent information, and use the first classification result as the target classification result.

[0150] Optionally, the processing module 702 is specifically configured to determine at least one first similarity, where the first similarity is the similarity between the target classification result and the first preset classification result in the preset classification set. The preset classification set includes at least one preset first-level classification result, the preset first-level classification result includes at least one preset second-level classification result, and the first preset classification result is constructed by one preset first-level classification result and one preset second-level classification result. One first similarity corresponds to one first preset classification result. If there is a first target similarity greater than the preset similarity threshold among the at least one first similarity, the processing module 702 is further configured to use the first template corresponding to the first target second-level classification result as the target template, where the first target second-level classification result is the preset second-level classification result in the first preset classification result corresponding to the first target similarity.

[0151] Optionally, if all of the at least one first similarity are less than or equal to the preset similarity threshold, the processing module 702 is specifically configured to determine at least one second similarity, where the second similarity is the similarity between the target first-level classification result and the preset first-level classification result, and one second similarity corresponds to one preset first-level classification result. The processing module 702 is further configured to determine the target template corresponding to the target classification result according to the second similarity.

[0152] Optionally, if there is a second target similarity greater than the preset similarity threshold among the at least one second similarity, the processing module 702 is specifically configured to determine the second target first-level classification result, where the second target first-level classification result is the preset first-level classification result corresponding to the second target similarity. The processing module 702 is further configured to determine at least one third similarity, where the third similarity is the similarity between the target second-level classification result and the second target second-level classification result, and the second target second-level classification result is at least one preset second-level classification result included in the second target first-level classification result. If there is a third target similarity greater than the preset similarity threshold among the at least one third similarity, the processing module 702 is further configured to use the second template corresponding to the third target second-level classification result as the target template, where the third target second-level classification result is the second target second-level classification result with the third similarity greater than the preset similarity threshold.

[0153] Optionally, if all of the at least one second similarity are less than or equal to the preset similarity threshold, the processing module 702 is specifically configured to determine at least one fourth similarity, where the fourth similarity is the similarity between the target second-level classification result and the preset second-level classification result, and one fourth similarity corresponds to one preset second-level classification result. If there is a fourth target similarity greater than the preset similarity threshold among the at least one fourth similarity, the processing module 702 is further configured to use the third template corresponding to the fourth target second-level classification result as the target template, where the fourth target second-level classification result is the preset second-level classification result corresponding to the fourth target similarity.

[0154] Figure 8 It is a schematic structural diagram of a device for generating intent instructions shown according to an exemplary embodiment. The device may include a processor 802, and the processor 802 is used to execute application program code, thereby implementing the method for generating intent instructions in this application.

[0155] The processor 802 may be a CPU, a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits for controlling the execution of the program in this application solution.

[0156] As Figure 8 shown, the device for generating intent instructions may further include a memory 803. Among them, the memory 803 is used to store the application program code for executing this application solution and is controlled by the processor 802 for execution.

[0157] The memory 803 may be a read-only memory (ROM) or other types of static storage devices that can store static information and instructions, a random access memory (RAM) or other types of dynamic storage devices that can store information and instructions, or may also be an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compact discs, laser discs, optical discs, digital versatile discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto. The memory 803 may exist independently and be connected to the processor 802 through a bus 804. The memory 803 may also be integrated with the processor 802.

[0158] As Figure 8 shown, the device for generating intent instructions may further include a communication interface 801. Among them, the communication interface 801, the processor 802, and the memory 803 may be coupled to each other. For example, they are coupled to each other through a bus 804. The communication interface 801 is used to interact with other devices, for example, to support the information interaction between the device for generating intent instructions and other devices.

[0159] It should be noted that Figure 8 the device structure shown in Figure 8In addition to the components shown, the generating device of the intent instruction may include more or fewer components than those shown, or combine certain components, or have a different component arrangement.

[0160] In actual implementation, all the functions implemented by the processing module 702 can be Figure 8 implemented by the processor 802 shown calling the program code in the memory 803.

[0161] This application also provides a computer-readable storage medium. Instructions are stored on the computer-readable storage medium. When the instructions in the computer-readable storage medium are executed by the processor of the computer device, the computer can execute the intent instruction generation method provided in the above-described embodiments. For example, the computer-readable storage medium may be the memory 803 including instructions, and the above instructions can be executed by the processor 802 of the computer device to complete the above method. Optionally, the computer-readable storage medium may be a non-transitory computer-readable storage medium. For example, the non-transitory computer-readable storage medium may be ROM, RAM, CD-ROM, magnetic tape, floppy disk, and optical data storage devices, etc.

[0162] Figure 9 Exemplarily shown is a conceptual partial view of a computer program product provided by an embodiment of this application. The computer program product includes a computer program for executing a computer process on a computing device.

[0163] In one embodiment, the computer program product is provided using a signal-bearing medium 900. The signal-bearing medium 900 may include one or more program instructions that, when run by one or more processors, can provide the functions or partial functions described above for Figure 5 、 Figure 6 Therefore, for example, referring to the embodiment shown in Figure 5 , one or more features of S501 to S503 may be borne by one or more instructions associated with the signal-bearing medium 900. In addition, Figure 9 the program instructions in also describe example instructions.

[0164] In some examples, the signal-bearing medium 900 may include a computer-readable medium 901, such as but not limited to, a hard disk drive, a compact disc (CD), a digital video disc (DVD), a digital tape, a memory, a read-only memory (ROM), or a random access memory (RAM), etc.

[0165] In some embodiments, the signal-bearing medium 900 may include a computer-readable recording medium 902, such as, but not limited to, a memory, a read / write (R / W) CD, R / W, DVD, and the like.

[0166] In some embodiments, the signal-bearing medium 900 may include a communication medium 903, such as, but not limited to, digital and / or analog communication media (e.g., fiber optic cables, waveguides, wired communication links, wireless communication links, etc.).

[0167] The signal-bearing medium 900 may be conveyed by a wireless form of the communication medium 903. One or more program instructions may be, for example, computer-executable instructions or logic-implemented instructions.

[0168] In some examples, such as for Figure 7 the generating device of the intention instructions described may be configured to provide various operations, functions, or actions in response to one or more program instructions via the computer-readable medium 901, the computer-readable recording medium 902, and / or the communication medium 903.

[0169] Through the description of the above embodiments, those skilled in the art can clearly understand that for the convenience and conciseness of description, only the above division of each functional module is used as an example. In actual applications, the above functions can be allocated to different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above.

[0170] In several embodiments provided in the present application, it should be understood that the disclosed device and method can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of modules or units is only a logical functional division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some interfaces. The indirect couplings or communication connections of the device or unit can be in electrical, mechanical or other forms.

[0171] The units described as separate components may or may not be physically separated. The components shown as units may be one physical unit or multiple physical units, that is, they can be in one place, or they can be distributed to multiple different places. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0172] In addition, in each embodiment of the present application, each functional unit can be integrated into a processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of a software functional unit.

[0173] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a readable storage medium. Based on this understanding, the technical solution of the embodiments of the present application, in essence, or the part that contributes to the prior art, or all or part of this technical solution can be embodied in the form of a software product. This software product is stored in a storage medium and includes several instructions for causing a device (which can be a single-chip microcomputer, a chip, etc.) or a processor to execute all or part of the steps of the methods of the various embodiments of the present application. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a ROM, a RAM, a magnetic disk, or an optical disc.

[0174] The above is only the specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions within the technical scope disclosed in the present application should be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for generating an intention instruction, characterized in that: The method comprises: Determine a type keyword from the initial intention instruction, where the type keyword is used to reflect the category of the initial intention instruction; Classifying the initial intention instruction based on the type keyword to obtain a target classification result; Based on the target classification result, a target template corresponding to the target classification result is determined, and a target intention instruction is generated based on the target template.

2. The method according to claim 1, characterized in that The classifying the initial intention instruction based on the type keyword to obtain a target classification result includes: Based on the type keyword, construct first intent information; Acquire second intent information, where the second intent information is intent information constructed based on a type keyword in the historical intent instruction; The target classification result is obtained based on the first intent information and the second intent information.

3. The method according to claim 2, characterized in that The obtaining a target classification result based on the first intent information and the second intent information includes: Determining intent similarity between the first intent information and the second intent information; If the intention similarity is less than or equal to a preset intention similarity threshold, classifying the first intention information to obtain the target classification result; If the intention similarity is greater than the preset intention similarity threshold, a first classification result corresponding to the second intention information is obtained, and the first classification result is used as the target classification result.

4. The method according to any one of claims 1 to 3, characterized in that The target classification result includes: a target primary classification result and a target secondary classification result; based on the target classification result, determining a target template corresponding to the target classification result includes: Determine at least one first similarity, where the first similarity is the similarity between the target classification result and a first preset classification result in a preset classification set; the preset classification set includes at least one preset first-level classification result, the preset first-level classification result includes at least one preset second-level classification result, and the first preset classification result is constructed by one preset first-level classification result and one preset second-level classification result; one first similarity corresponds to one first preset classification result; If there is a first target similarity greater than a preset similarity threshold among the at least one first similarity, the first template corresponding to the first target secondary classification result is used as the target template, and the first target secondary classification result is a preset secondary classification result among the first preset classification results corresponding to the first target similarity.

5. The method according to claim 4, characterized in that The method further comprises: If the at least one first similarity is less than or equal to the preset similarity threshold, then determining at least one second similarity, where the second similarity is the similarity between the target first-level classification result and the preset first-level classification result, and one second similarity corresponds to one preset first-level classification result; According to the second similarity, a target template corresponding to the target classification result is determined.

6. The method according to claim 5, characterized in that The step of determining the target template corresponding to the target classification result according to the second similarity includes: If there is a second target similarity greater than the preset similarity threshold among the at least one second similarity, determining a second target primary classification result, where the second target primary classification result is a preset primary classification result corresponding to the second target similarity; Determine at least one third similarity, where the third similarity is a similarity between the target secondary classification result and the second target secondary classification result, where the second target secondary classification result is at least one preset secondary classification result included in the second target primary classification result; If there is a third target similarity greater than the preset similarity threshold among the at least one third similarity, the second template corresponding to the third target secondary classification result is used as the target template, and the third target secondary classification result is the second target secondary classification result whose third similarity is greater than the preset similarity threshold.

7. The method according to claim 5, characterized in that The step of determining the target template corresponding to the target classification result according to the second similarity further includes: If at least one second similarity is less than or equal to the preset similarity threshold, then determining at least one fourth similarity, the fourth similarity being the similarity between the target secondary classification result and the preset secondary classification result, and one fourth similarity corresponds to one preset secondary classification result; If there is a fourth target similarity greater than the preset similarity threshold among the at least one fourth similarity, the third template corresponding to the fourth target secondary classification result is used as the target template, and the fourth target secondary classification result is the preset secondary classification result corresponding to the fourth target similarity.

8. A device for generating an intention instruction, characterized in that: The device comprises: A processing module, used to determine a type keyword from the initial intention instruction, wherein the type keyword is used to reflect the category of the initial intention instruction; The processing module is further used to classify the initial intention instruction based on the type keyword to obtain a target classification result; The processing module is further used to determine a target template corresponding to the target classification result based on the target classification result, and generate a target intention instruction based on the target template.

9. A device for generating an intention instruction, characterized in that: include: Processor and memory; The processor is coupled to the memory; The memory is used to store one or more programs, and the one or more programs include computer-executable instructions. When the device for generating the intention instruction is running, the processor executes the computer-executable instructions stored in the memory to enable the device for generating the intention instruction to perform the method as described in any one of claims 1-7.

10. A computer-readable storage medium, wherein instructions are stored in the computer-readable storage medium, characterized in that: When a computer executes the instructions, the computer performs the method according to any one of claims 1 to 7.

11. A computer program product, characterized in that The computer program product comprises computer program instructions, which implement the method according to any one of claims 1 to 7 when the computer program instructions are executed.

Images