Optimization method for improving page voice intention recognition result

Abstract

The invention provides an optimization method for improving a page voice intention recognition result, and belongs to the technical field of data processing, and the method comprises the steps: 1, determining system configuration data and preset voice configuration data of an intelligent voice control system, and obtaining real-time voice recognition data and historical voice recognition data; 2, object correction data, voice operation data and correction vector data are determined, and correction voice configuration data of the intelligent voice control system are determined; 3, determining real-time intention data; and 4, determining real-time operation data based on the real-time intention data, and performing real-time response to the voice control system based on the real-time operation data. The response speed and accuracy of the intelligent voice control system can be optimized, the ability to understand the voice instruction of the user is enhanced, the accuracy of voice control is improved, the voice control experience of the user is improved, manual operation of maintenance and management personnel is reduced, and the operation efficiency of the maintenance and management personnel is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to an optimization method for improving page speech intent recognition results. Background Art

[0002] Intelligent speech recognition technology has a history of several decades of development. Early speech recognition technology was mainly based on limited vocabulary and pattern recognition algorithms, and its effectiveness was limited by hardware performance and the quality of speech data. In recent years, intelligent speech control systems have gradually introduced big data analysis and machine learning technologies, and have realized functions such as interface jumps, interface element control, and information entry through intelligent speech technology, thus achieving full system function control. However, it still faces the disadvantages of low accuracy in speech intent recognition and the inability to learn and correct aliases through the front-end page.

[0003] Therefore, the present invention provides an optimization method for improving the recognition result of page speech intent. Summary of the invention

[0004] The present invention provides an optimization method for improving the page voice intention recognition result, by analyzing the determined system configuration data, preset voice configuration data and the acquired historical voice recognition data, determining the object correction data, voice operation data and correction vector data, and determining the correction voice configuration data of the intelligent voice control system, according to the real-time voice recognition data and the correction voice configuration data, determining the real-time intention data, determining the real-time operation data and executing. The response speed and accuracy of the intelligent voice control system can be optimized, the ability to understand the user's voice instructions can be enhanced, the accuracy of voice control can be improved, the user's voice control experience can be improved, the manual operation of maintenance personnel can be reduced, and the operation efficiency of maintenance personnel can be improved.

[0005] The present invention provides an optimization method for improving page speech intention recognition results, comprising: Step 1: Determine the system configuration data and preset voice configuration data of the intelligent voice control system, and obtain real-time voice recognition data and historical voice recognition data; Step 2: Based on the system configuration data, the preset voice configuration data and the historical voice recognition data, determine the object correction data, the voice operation data and the correction vector data, and determine the corrected voice configuration data of the intelligent voice control system; Step 3: Determine real-time intent data based on the real-time speech recognition data and the modified speech configuration data of the intelligent speech control system; Step 4: Determine real-time operation data based on the real-time intention data, and respond to the voice control system in real time based on the real-time operation data.

[0006] According to an optimization method for improving the page voice intention recognition result provided by the present invention, the system configuration data and preset voice configuration data of the intelligent voice control system are determined, including: Obtaining page configuration data of each page in the intelligent voice control system, wherein the page configuration data at least includes a plurality of page configuration objects, a page configuration object type of each page configuration object, an operation action of each page configuration object, and a plurality of operation action aliases of each operation action of each page configuration object; Determine system configuration data of the intelligent voice control system based on page configuration data of all pages in the intelligent voice control system; Determine a plurality of voice mapping instructions based on system configuration data of the intelligent voice control system, and determine an object mapping set for each page configuration object based on all the voice mapping instructions; The preset voice configuration data is determined based on an object mapping set of all page configuration objects.

[0007] According to an optimization method for improving the page voice intention recognition result provided by the present invention, real-time voice recognition data and historical voice recognition data are obtained, including: Get real-time speech recognition data within the current specified time period; Acquire historical speech recognition sub-data within multiple historical specified time periods before the current specified time period, wherein the historical speech recognition sub-data includes multiple historical operation instructions, the speech configuration object of each historical operation instruction, the operation label of each historical operation instruction, and the operation action of each historical operation instruction; Based on all the historical speech recognition sub-data of the historical designated time period, the historical speech recognition data is determined.

[0008] According to an optimization method for improving the page voice intention recognition result provided by the present invention, based on system configuration data, preset voice configuration data and historical voice recognition data, object correction data, voice operation data and correction vector data are determined, and the corrected voice configuration data of the intelligent voice control system is determined, including: Extract all page configuration objects in the system configuration data to determine a page configuration object set; Extracting the voice configuration objects of all historical operation instructions in all historical voice recognition sub-data of the historical specified time period in the historical voice recognition data to determine a voice configuration object set; Determine the number of page configuration objects in the page configuration object set as the number of clusters K for cluster analysis, randomly select K voice configuration objects in the voice configuration object set as initial cluster centers, perform cluster analysis on the voice configuration object set, and determine K clusters, cluster center configuration objects of each cluster, and a subset of voice configuration objects of each cluster; Extracting all operation labels of all voice configuration objects in the voice configuration object subset of each cluster group in the historical voice recognition sub-data within each specified time period based on the historical voice recognition data, and determining the operation label set of the cluster core configuration object of each cluster group; Determine a cluster center configuration object matrix of the cluster center configuration objects of each cluster family within each specified time period based on the system configuration data, the historical speech recognition data and the operation tag set of the cluster center configuration objects of each cluster family, and determine a correction success vector of the cluster center configuration objects of each cluster family; Determine a correction mapping set, an operation label set, and a correction success vector for each page configuration object; Determine object correction data based on a correction mapping set of all page configuration objects in the page configuration object set, determine voice operation data based on an operation tag set of all page configuration objects in the page configuration object set, and determine correction vector data based on correction success vectors of all page configuration objects in the page configuration object set; Corrected voice configuration data of the intelligent voice control system is determined based on the object correction data, the voice operation data, and the correction vector data.

[0009] According to an optimization method for improving page speech intent recognition results provided by the present invention, determining a cluster center configuration object matrix of a cluster center configuration object of each cluster family within each specified time period includes: Determine a cluster core configuration object matrix of the cluster core configuration objects of each cluster family within each specified time period based on the system configuration data, the historical speech recognition data and the operation tag set of the cluster core configuration objects of each cluster family; ; ; in, The cluster center configuration object matrix representing the cluster center configuration object of the a-th cluster family in the t-th historical specified time period, Indicates that the operation label in the operation label set of the cluster center configuration object of the a-th cluster group in the t-th historical specified time period is successful. represents the number of operation tags in the operation tag set of the cluster center configuration object of the a-th cluster family, They respectively represent the ith operation label and the aN2th operation label in the operation label set of the cluster center configuration object of the ath cluster family in the tth historical specified time period, represents the operation success rate of the operation tags in the operation tag set of the cluster center configuration object of the a-th cluster group in the t-th historical specified time period, represents the operation failure rate of the ith operation label and the aN2th operation label in the operation label set of the cluster center configuration object of the ath cluster group in the tth historical specified time period, The operation label of the Lth historical operation instruction of the jth voice configuration object in the ath cluster of the voice configuration object subset in the tth historical specified time period, a represents the number of all historical operation instructions based on the jth voice configuration object in the voice configuration object subset of the ath cluster, Indicates the number of voice configuration objects in the voice configuration object subset of the a-th cluster in the t-th historical specified time period.

[0010] According to an optimization method for improving the page speech intention recognition result provided by the present invention, determining the correction success vector of the cluster center configuration object of each cluster family includes: determining the correction success vector of the cluster center configuration object of each cluster family based on the cluster center configuration object matrix of the cluster center configuration object of each cluster family within all specified time periods; ; ; in, The modified success vector of the cluster center configuration object of the a-th cluster family They represent the correction success rates of the second operation label, the ith operation label, and the aN2th operation label in the operation label set of the cluster center configuration object of the ath cluster family, respectively. represents the period attenuation factor, represents the correction factor, and Nu represents the number of historical specified time periods.

[0011] According to an optimization method for improving page speech intent recognition results provided by the present invention, a correction mapping set, an operation label set, and a correction success vector for each page configuration object are determined, including: Determine a revised mapping set for each page configuration object based on an object mapping set for each page configuration object in the preset voice configuration data, all page configuration objects in the page configuration object set, cluster core configuration objects of all cluster families, and a voice configuration object sub-set; ;; ; in, Represents the similarity value between the xth page configuration object and the ath cluster in the page configuration object set. represents the correction mapping set of the xth page configuration object in the page configuration object set, represents the cluster center configuration object of the a-th cluster family, represents the jth voice configuration object in the ath cluster group's voice configuration object subset, Represents the Euclidean norm of the xth page configuration object in the page configuration object collection, represents the Euclidean norm of the cluster center configuration object of the a-th cluster family, represents the Euclidean norm of the jth voice configuration object in the subset of voice configuration objects of the ath cluster, represents the subset of voice configuration objects of the a-th cluster, represents the voice configuration object sub-set with the largest similarity value to the x-th page configuration object in the page configuration object set, An object mapping collection representing the xth page configuration object in the page configuration object collection; Determine an operation tag set for each page configuration object in the page configuration object set based on an operation tag set of a cluster corresponding to a voice configuration object subset with a maximum similarity value to each page configuration object in the page configuration object set; Based on the correction success vector of the cluster group corresponding to the voice configuration object subset with the largest similarity value to each page configuration object in the page configuration object set, a correction success vector of each page configuration object in the page configuration object set is determined.

[0012] According to an optimization method for improving the page voice intention recognition result provided by the present invention, based on the real-time voice recognition data and the modified voice configuration data of the intelligent voice control system, the real-time intention data is determined, including: The intent model is optimized based on the correction vector data in the corrected voice configuration data, and the real-time voice recognition data is analyzed based on the optimized intent model. If the analysis is successful, the predicted intent data is determined, and the predicted intent data is matched with the correction mapping set of each page configuration object in the corrected voice configuration data, and the real-time intent data is determined based on the matching result; If the analysis is unsuccessful, a hard match is performed on the real-time speech recognition data to determine the real-time intent data.

[0013] According to an optimization method for improving page voice intent recognition results provided by the present invention, real-time intent data is determined based on matching results, including: If there is a page configuration object in the modified voice configuration data that highly matches the predicted intent data, determine the real-time intent data of the real-time voice data based on the modified mapping set and the operation label set of the page configuration object that highly matches the predicted intent data; If there is no page configuration object in the revised voice configuration data that is highly matched with the predicted intent data, the revised mapping set and operation label set of the page configuration object that is most matched with the predicted intent data are revised again.

[0014] According to an optimization method for improving the page voice intention recognition result provided by the present invention, real-time operation data is determined based on real-time intention data, and a voice control system is responded to in real time based on the real-time operation data, including: Parsing the real-time intention data to determine real-time operation data, wherein the real-time operation data includes at least one or more real-time operation instructions; The voice control system executes each real-time operation instruction in the real-time operation data and determines an operation label of each real-time operation instruction; Based on all the real-time operation instructions in the real-time operation data and the operation label of each real-time operation instruction, current speech recognition sub-data of a current specified time period is determined.

[0015] Compared with the prior art, the present invention has the following beneficial effects: By analyzing the determined system configuration data, preset voice configuration data, and historical voice recognition data, the object correction data, voice operation data, and correction vector data are determined, and the corrected voice configuration data of the intelligent voice control system is determined. According to the real-time voice recognition data and the corrected voice configuration data, the real-time intention data is determined, and the real-time operation data is determined and executed. The response speed and accuracy of the intelligent voice control system can be optimized, the ability to understand user voice commands can be enhanced, the accuracy of voice control can be improved, the user voice control experience can be improved, the manual operation of maintenance personnel can be reduced, and the operation efficiency of maintenance personnel can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the present invention or the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0017] Figure 1 It is a flow chart of an optimization method for improving the page voice intent recognition result provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0018] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0019] Embodiment 1: The embodiment of the present invention provides an optimization method for improving the recognition result of page speech intent, such as Figure 1 As shown, including: Step 1: Determine the system configuration data and preset voice configuration data of the intelligent voice control system, and obtain real-time voice recognition data and historical voice recognition data; Step 2: Based on the system configuration data, the preset voice configuration data and the historical voice recognition data, determine the object correction data, the voice operation data and the correction vector data, and determine the corrected voice configuration data of the intelligent voice control system; Step 3: Determine real-time intent data based on the real-time speech recognition data and the modified speech configuration data of the intelligent speech control system; Step 4: Determine real-time operation data based on the real-time intention data, and respond to the voice control system in real time based on the real-time operation data.

[0020] In this embodiment, the system configuration data, the voice configuration data and the historical voice recognition data are combined, and the object correction data, the voice operation data and the correction vector data are generated by analysis, thereby generating the corrected voice configuration data.

[0021] In this embodiment, real-time speech recognition data and modified speech configuration data are used to identify and determine real-time intent data based on the intent model, that is, the user's real-time intent is determined by parsing the voice instructions.

[0022] In this embodiment, the real-time operation data (specific real-time operation instructions prepared by the system according to the intention data) is further determined based on the real-time intention data. Then, the intelligent voice control system will respond in real time and perform corresponding operations.

[0023] The beneficial effects of the above technical solution are as follows: by analyzing the determined system configuration data, preset voice configuration data and the acquired historical voice recognition data, the object correction data, voice operation data and correction vector data are determined, and the corrected voice configuration data of the intelligent voice control system is determined, and the real-time intention data is determined according to the real-time voice recognition data and the corrected voice configuration data, and the real-time operation data is determined and executed. The response speed and accuracy of the intelligent voice control system can be optimized, the ability to understand user voice commands can be enhanced, the accuracy of voice control can be improved, the user voice control experience can be improved, the manual operation of maintenance personnel can be reduced, and the operation efficiency of maintenance personnel can be improved.

[0024] Embodiment 2: The embodiment of the present invention provides an optimization method for improving the page voice intention recognition result, determining the system configuration data and preset voice configuration data of the intelligent voice control system, including: Obtaining page configuration data of each page in the intelligent voice control system, wherein the page configuration data at least includes a plurality of page configuration objects, a page configuration object type of each page configuration object, an operation action of each page configuration object, and a plurality of operation action aliases of each operation action of each page configuration object; Determine system configuration data of the intelligent voice control system based on page configuration data of all pages in the intelligent voice control system; Determine a plurality of voice mapping instructions based on system configuration data of the intelligent voice control system, and determine an object mapping set for each page configuration object based on all the voice mapping instructions; The preset voice configuration data is determined based on an object mapping set of all page configuration objects.

[0025] In this embodiment, the configuration data of each page includes multiple page configuration objects, each object has a type, an operation action, and multiple aliases of the operation action. For example, a page may include a configuration object of "desk lamp", the type of this object is "switch device", the operation action is "open", and the aliases of the operation action are "open", "light up", etc.

[0026] In this embodiment, the global configuration of the intelligent voice control system is determined by integrating the page configuration data of all pages, that is, all page object types and action aliases constitute the overall configuration of the system, ensuring that the system can support all voice commands and operations.

[0027] In this embodiment, according to the system configuration data, multiple voice commands are formulated to be mapped to corresponding page configuration objects. For example, "turn on the light" can be mapped to a "desk lamp" configuration object on a certain page.

[0028] In this embodiment, a voice command mapping associated with each page configuration object is created to ensure that the voice recognition system can correctly understand different commands and perform corresponding operations.

[0029] In this embodiment, a set of preset voice configurations is defined through object mapping relationships, so that the voice control system can quickly respond to user voice commands in different scenarios.

[0030] The beneficial effects of the above technical solution are: determining the system configuration data and preset voice configuration data of the intelligent voice control system can provide a data basis for determining and correcting the voice configuration data, thereby improving the accuracy and response speed of voice command recognition and improving the user interaction experience.

[0031] Embodiment 3: The embodiment of the present invention provides an optimization method for improving the page speech intention recognition result, obtaining real-time speech recognition data and historical speech recognition data, including: Get real-time speech recognition data within the current specified time period; Acquire historical speech recognition sub-data within multiple historical specified time periods before the current specified time period, wherein the historical speech recognition sub-data includes multiple historical operation instructions, the speech configuration object of each historical operation instruction, the operation label of each historical operation instruction, and the operation action of each historical operation instruction; Based on all the historical speech recognition sub-data of the historical designated time period, the historical speech recognition data is determined.

[0032] In this embodiment, the real-time voice recognition data refers to the instant recognition result of the user's voice input by the intelligent voice control system during the current specified time period, including the voice instructions issued by the user and the converted text of the voice.

[0033] In this embodiment, the historical voice recognition sub-data refers to the data of each historical operation instruction stored by the system within a certain historical specified time period in the past. Each historical operation instruction includes: voice configuration object: the specific operation object corresponding to the instruction (such as equipment, function, etc.); operation tag: the execution result of the instruction (such as success, unrecognized voice configuration object, missing voice configuration object, etc.); operation action: the specific action performed by the voice instruction (such as "open", "close", etc.).

[0034] In this embodiment, the historical speech recognition data is formed by integrating the historical speech recognition sub-data within multiple time periods to form a complete historical record.

[0035] The beneficial effects of the above technical solution are as follows: obtaining real-time speech recognition data and historical speech recognition data can provide a data basis for determining and correcting speech configuration data, provide personalized speech services, and improve user interaction experience.

[0036] Embodiment 4: The embodiment of the present invention provides an optimization method for improving the page voice intention recognition result, which determines object correction data, voice operation data and correction vector data based on system configuration data, preset voice configuration data and historical voice recognition data, and determines the corrected voice configuration data of the intelligent voice control system, including: Extract all page configuration objects in the system configuration data to determine a page configuration object set; Extracting the voice configuration objects of all historical operation instructions in all historical voice recognition sub-data of the historical specified time period in the historical voice recognition data to determine a voice configuration object set; Determine the number of page configuration objects in the page configuration object set as the number of clusters K for cluster analysis, randomly select K voice configuration objects in the voice configuration object set as initial cluster centers, perform cluster analysis on the voice configuration object set, and determine K clusters, cluster center configuration objects of each cluster, and a subset of voice configuration objects of each cluster; Extracting all operation labels of all voice configuration objects in the voice configuration object subset of each cluster group in the historical voice recognition sub-data within each specified time period based on the historical voice recognition data, and determining the operation label set of the cluster core configuration object of each cluster group; Determine a cluster center configuration object matrix of the cluster center configuration objects of each cluster family within each specified time period based on the system configuration data, the historical speech recognition data and the operation tag set of the cluster center configuration objects of each cluster family, and determine a correction success vector of the cluster center configuration objects of each cluster family; Determine a correction mapping set, an operation label set, and a correction success vector for each page configuration object; Determine object correction data based on a correction mapping set of all page configuration objects in the page configuration object set, determine voice operation data based on an operation tag set of all page configuration objects in the page configuration object set, and determine correction vector data based on correction success vectors of all page configuration objects in the page configuration object set; Corrected voice configuration data of the intelligent voice control system is determined based on the object correction data, the voice operation data, and the correction vector data.

[0037] In this embodiment, all page configuration objects are extracted from the system configuration data. Each page configuration object includes information of operation items such as equipment and functions.

[0038] In this embodiment, the voice configuration object of each historical operation instruction is extracted from the historical voice recognition data, and the voice configuration object reflects the specific operation item of the voice instruction.

[0039] In this embodiment, the number of configuration objects in the page configuration object is used as the number of clusters (K) for cluster analysis, K voice configuration objects are randomly selected as initial cluster centers, and then cluster analysis is performed on the voice configuration objects to determine K cluster centers, the cluster center configuration objects of each cluster center, and all voice configuration object subsets of the cluster center.

[0040] In this embodiment, based on the historical speech recognition data, the operation tags of all the speech configuration objects in each cluster are extracted, and these tags reflect the execution results of different historical operation instructions.

[0041] In this embodiment, based on the system configuration data and the historical speech recognition data, the operation labels of the cluster core configuration objects of each cluster family are analyzed to generate a cluster core configuration object matrix.

[0042] In this embodiment, the object correction data, the voice operation data and the correction vector data are integrated to finally generate the corrected voice configuration data.

[0043] The beneficial effects of the above technical solution are as follows: based on system configuration data, preset voice configuration data and historical voice recognition data, object correction data, voice operation data and correction vector data are determined, and the corrected voice configuration data of the intelligent voice control system is determined, which can improve the system's recognition ability for complex voice commands, and enhance voice recognition accuracy and personalized experience.

[0044] Embodiment 5: The embodiment of the present invention provides an optimization method for improving the page speech intention recognition result, which determines the cluster center configuration object matrix of the cluster center configuration object of each cluster family within each specified time period, including: Determine a cluster core configuration object matrix of the cluster core configuration objects of each cluster family within each specified time period based on the system configuration data, the historical speech recognition data and the operation tag set of the cluster core configuration objects of each cluster family; ; ; in, The cluster center configuration object matrix representing the cluster center configuration object of the a-th cluster family in the t-th historical specified time period, Indicates that the operation label in the operation label set of the cluster center configuration object of the a-th cluster group in the t-th historical specified time period is successful. represents the number of operation tags in the operation tag set of the cluster center configuration object of the a-th cluster family, They respectively represent the ith operation label and the aN2th operation label in the operation label set of the cluster center configuration object of the ath cluster family in the tth historical specified time period, represents the operation success rate of the operation tags in the operation tag set of the cluster center configuration object of the a-th cluster group in the t-th historical specified time period, represents the operation failure rate of the ith operation label and the aN2th operation label in the operation label set of the cluster center configuration object of the ath cluster group in the tth historical specified time period, The operation label of the Lth historical operation instruction of the jth voice configuration object in the ath cluster of the voice configuration object subset in the tth historical specified time period, a represents the number of all historical operation instructions based on the jth voice configuration object in the voice configuration object subset of the ath cluster, Indicates the number of voice configuration objects in the voice configuration object subset of the a-th cluster in the t-th historical specified time period.

[0045] In this embodiment, The operation label of the Lth historical operation instruction of the jth voice configuration object in the ath cluster voice configuration object subset in the tth historical specified time period and the ith operation label in the operation label set of the cluster center configuration object of the ath cluster group in the tth historical specified time period If they are the same, the value is 1.

[0046] The beneficial effects of the above technical solution are: determining the cluster center configuration object matrix of the cluster center configuration objects of each cluster family within each specified time period can provide a data basis for determining the corrected success vector of the cluster center configuration objects of all cluster families and correcting the vector data, thereby improving the success rate of real-time speech recognition data recognition.

[0047] Embodiment 6: The embodiment of the present invention provides an optimization method for improving the page speech intention recognition result, determining the correction success vector of the cluster center configuration object of each cluster family, including: determining the correction success vector of the cluster center configuration object of each cluster family based on the cluster center configuration object matrix of the cluster center configuration object of each cluster family within all specified time periods; ; ; in, The modified success vector of the cluster center configuration object of the a-th cluster family They represent the correction success rates of the second operation label, the ith operation label, and the aN2th operation label in the operation label set of the cluster center configuration object of the ath cluster family, respectively. represents the period attenuation factor, represents the correction factor, and Nu represents the number of historical specified time periods.

[0048] In this embodiment, It represents the operation success frequency of the operation tags in the operation tag set of the cluster core configuration object of the a-th cluster family, that is, the weight of the operation success rate of the operation tags in the operation tag set of the cluster core configuration object of the a-th cluster family.

[0049] In this embodiment, It represents the operation failure frequency of the i-th operation label and the aN2-th operation label in the operation label set of the cluster core configuration object of the a-th cluster family, that is, the weight of the operation failure rate of the i-th operation label and the aN2-th operation label in the operation label set of the cluster core configuration object of the a-th cluster family.

[0050] The beneficial effect of the above technical solution is: determining the corrected success vector of the cluster center configuration object of each cluster group can provide a data basis for determining the corrected vector data, thereby improving the success rate of real-time speech recognition data recognition.

[0051] Embodiment 7: The embodiment of the present invention provides an optimization method for improving the page speech intent recognition result, determining a correction mapping set, an operation label set and a correction success vector for each page configuration object, including: Determine a revised mapping set for each page configuration object based on an object mapping set for each page configuration object in the preset voice configuration data, all page configuration objects in the page configuration object set, cluster core configuration objects of all cluster families, and a voice configuration object sub-set; ;; ; in, Represents the similarity value between the xth page configuration object and the ath cluster in the page configuration object set. represents the correction mapping set of the xth page configuration object in the page configuration object set, represents the cluster center configuration object of the a-th cluster family, represents the jth voice configuration object in the ath cluster group's voice configuration object subset, Represents the Euclidean norm of the xth page configuration object in the page configuration object collection, represents the Euclidean norm of the cluster center configuration object of the a-th cluster family, represents the Euclidean norm of the jth voice configuration object in the subset of voice configuration objects of the ath cluster, represents the subset of voice configuration objects of the a-th cluster, represents the voice configuration object sub-set with the largest similarity value to the x-th page configuration object in the page configuration object set, An object mapping collection representing the xth page configuration object in the page configuration object collection; Determine an operation tag set for each page configuration object in the page configuration object set based on an operation tag set of a cluster corresponding to a voice configuration object subset with a maximum similarity value to each page configuration object in the page configuration object set; Based on the correction success vector of the cluster group corresponding to the voice configuration object subset with the largest similarity value to each page configuration object in the page configuration object set, a correction success vector of each page configuration object in the page configuration object set is determined.

[0052] In this embodiment, Represents the weight of the jth voice configuration object in the ath cluster group's voice configuration object subset in the tth historical specified time period.

[0053] In this embodiment, Represents the weight of the subset of voice configuration objects of the ath cluster group in all historical specified time periods.

[0054] In this embodiment, the value range of a is 1 to K.

[0055] In this embodiment, a subset of voice configuration objects that is most similar to each page configuration object is selected and corresponds to a cluster, and an operation label of the cluster is determined as the operation label of the corresponding page configuration object.

[0056] In this embodiment, for each page configuration object, the system extracts the correction success vector of the corresponding cluster based on the subset of voice configuration objects that is most similar to the page configuration object.

[0057] The beneficial effects of the above technical solution are: determining the correction mapping set, operation label set and correction success vector of each page configuration object can provide a data basis for determining the correction vector data, dynamically adjust and optimize the voice recognition and operation response of the intelligent voice control system, improve the accuracy and efficiency of user interaction, and enhance the personalized experience.

[0058] Embodiment 8: The embodiment of the present invention provides an optimization method for improving the page voice intention recognition result, which determines the real-time intention data based on the real-time voice recognition data and the modified voice configuration data of the intelligent voice control system, including: Optimizing the intent model based on the correction vector data in the correction voice configuration data; Perform intent analysis on the real-time speech recognition data based on the optimized intent model. If the analysis is successful, determine the predicted intent data, match the predicted intent data with the revised mapping set of each page configuration object in the revised speech configuration data, and determine the real-time intent data based on the matching result; If the analysis is unsuccessful, a hard match is performed on the real-time speech recognition data to determine the real-time intent data.

[0059] In this embodiment, the intention model is the core model used by the intelligent voice control system to parse the user's voice commands. The intention model is optimized by correcting the correction vector data in the voice configuration data. The system can correct the correction success vectors of all page configuration objects to improve the accuracy of the model, and the intention model can be optimized through the correction algorithm.

[0060] In this embodiment, after the intent model is optimized, the real-time speech recognition data is analyzed, and if the analysis is successful, the predicted intent data is determined.

[0061] In this embodiment, if intent parsing fails, the system will adopt a hard matching method, that is, directly matching the real-time speech recognition data with the page configuration object, without relying on the optimized intent model, to ensure that a response can still be given when intent parsing fails.

[0062] The beneficial effects of the above technical solution are as follows: based on real-time speech recognition data and the corrected speech configuration data of the intelligent voice control system, real-time intention data is determined, which can ensure the efficiency and reliability of speech recognition, improve the adaptability and robustness of the intelligent voice control system, and improve the user experience and accuracy of voice control.

[0063] Embodiment 9: The embodiment of the present invention provides an optimization method for improving the page voice intent recognition result, and determines the real-time intent data based on the matching result, including: If there is a page configuration object in the modified voice configuration data that highly matches the predicted intent data, determine the real-time intent data of the real-time voice data based on the modified mapping set and the operation label set of the page configuration object that highly matches the predicted intent data; If there is no page configuration object in the revised voice configuration data that is highly matched with the predicted intent data, the revised mapping set and operation label set of the page configuration object that is most matched with the predicted intent data are revised again.

[0064] In this embodiment, when there are page configuration objects in the modified voice configuration data that are highly matched with the predicted intention data, the real-time intention data of the real-time voice data is determined based on the modified mapping set and the operation tag set of these page configuration objects. This means that the system will use the matching relationship between the page configuration objects, combined with the corresponding operation tags, to confirm the user's specific intention.

[0065] In this embodiment, if a page configuration object that highly matches the predicted intent data cannot be found in the modified voice configuration data, the system will select the page configuration object that has the highest match with the predicted intent data, and further modify the modified mapping set and operation label set of the page configuration object to further optimize the matching results and ensure that the system can recognize and respond to the user's intent.

[0066] The beneficial effects of the above technical solution are: determining real-time intention data based on the matching results can ensure that the intelligent voice control system can accurately understand and respond to complex user commands, enhance the intelligence and adaptability of the system, improve the accuracy of voice control, and improve user experience.

[0067] Embodiment 10: The embodiment of the present invention provides an optimization method for improving the page voice intention recognition result, determining real-time operation data based on real-time intention data, and responding to the voice control system in real time based on the real-time operation data, including: Parsing the real-time intention data to determine real-time operation data, wherein the real-time operation data includes at least one or more real-time operation instructions; The voice control system executes each real-time operation instruction in the real-time operation data and determines an operation label of each real-time operation instruction; Based on all the real-time operation instructions in the real-time operation data and the operation label of each real-time operation instruction, current speech recognition sub-data of a current specified time period is determined.

[0068] In this embodiment, based on the real-time speech recognition data, the system analyzes and determines the user's intention and converts it into operation data (such as opening an application, adjusting the volume, etc.). The real-time operation data usually contains multiple instructions, each of which represents a specific operation, such as "play music" or "turn off the lights."

[0069] In this embodiment, the intelligent voice control system executes each real-time operation instruction in sequence according to the determined real-time operation data. After each real-time operation instruction is executed, an operation tag is determined to identify the execution result of the operation.

[0070] In this embodiment, all real-time operation instructions and their labels are aggregated to generate voice recognition sub-data for the current specified time period. The beneficial effects of the above technical solution are: determining real-time operation data based on real-time intention data, and responding to the voice control system in real time based on the real-time operation data, which can improve the real-time response capability of the intelligent voice control system, accurately understand and respond to complex user instructions, and provide a personalized and efficient voice control experience.

[0071] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0072] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0073] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An optimization method for improving the recognition result of page speech intent, characterized in that: include: Step 1: Determine the system configuration data and preset voice configuration data of the intelligent voice control system, and obtain real-time voice recognition data and historical voice recognition data; Step 2: Based on the system configuration data, the preset voice configuration data and the historical voice recognition data, determine the object correction data, the voice operation data and the correction vector data, and determine the corrected voice configuration data of the intelligent voice control system; Step 3: Determine real-time intent data based on the real-time speech recognition data and the modified speech configuration data of the intelligent speech control system; Step 4: Determine real-time operation data based on the real-time intention data, and respond to the voice control system in real time based on the real-time operation data.

2. The optimization method for improving the recognition result of page speech intent according to claim 1, characterized in that: Determine the system configuration data and preset voice configuration data of the intelligent voice control system, including: Obtaining page configuration data of each page in the intelligent voice control system, wherein the page configuration data at least includes a plurality of page configuration objects, a page configuration object type of each page configuration object, an operation action of each page configuration object, and a plurality of operation action aliases of each operation action of each page configuration object; Determine system configuration data of the intelligent voice control system based on page configuration data of all pages in the intelligent voice control system; Determine a plurality of voice mapping instructions based on system configuration data of the intelligent voice control system, and determine an object mapping set for each page configuration object based on all the voice mapping instructions; The preset voice configuration data is determined based on an object mapping set of all page configuration objects.

3. The optimization method for improving the recognition result of page speech intent according to claim 2, characterized in that: Get real-time and historical speech recognition data, including: Get real-time speech recognition data within the current specified time period; Acquire historical speech recognition sub-data within multiple historical specified time periods before the current specified time period, wherein the historical speech recognition sub-data includes multiple historical operation instructions, the speech configuration object of each historical operation instruction, the operation label of each historical operation instruction, and the operation action of each historical operation instruction; Based on all the historical speech recognition sub-data of the historical designated time period, the historical speech recognition data is determined.

4. The optimization method for improving the recognition result of page speech intent according to claim 3, characterized in that: Based on the system configuration data, the preset voice configuration data and the historical voice recognition data, the object correction data, the voice operation data and the correction vector data are determined, and the corrected voice configuration data of the intelligent voice control system is determined, including: Extract all page configuration objects in the system configuration data to determine a page configuration object set; Extracting the voice configuration objects of all historical operation instructions in all historical voice recognition sub-data of the historical specified time period in the historical voice recognition data to determine a voice configuration object set; Determine the number of page configuration objects in the page configuration object set as the number of clusters K for cluster analysis, randomly select K voice configuration objects in the voice configuration object set as initial cluster centers, perform cluster analysis on the voice configuration object set, and determine K clusters, cluster center configuration objects of each cluster, and a subset of voice configuration objects of each cluster; Extracting all operation labels of all voice configuration objects in the voice configuration object subset of each cluster group in the historical voice recognition sub-data within each specified time period based on the historical voice recognition data, and determining the operation label set of the cluster core configuration object of each cluster group; Determine a cluster center configuration object matrix of the cluster center configuration objects of each cluster family within each specified time period based on the system configuration data, the historical speech recognition data and the operation tag set of the cluster center configuration objects of each cluster family, and determine a correction success vector of the cluster center configuration objects of each cluster family; Determine a correction mapping set, an operation label set, and a correction success vector for each page configuration object; Determine object correction data based on a correction mapping set of all page configuration objects in the page configuration object set, determine voice operation data based on an operation tag set of all page configuration objects in the page configuration object set, and determine correction vector data based on correction success vectors of all page configuration objects in the page configuration object set; Corrected voice configuration data of the intelligent voice control system is determined based on the object correction data, the voice operation data, and the correction vector data.

5. The optimization method for improving the recognition result of page speech intent according to claim 4, characterized in that: A cluster center configuration object matrix is ​​determined for each cluster family in each specified time period, including: Determine a cluster core configuration object matrix of the cluster core configuration objects of each cluster family within each specified time period based on the system configuration data, the historical speech recognition data and the operation tag set of the cluster core configuration objects of each cluster family; ； ； in, The cluster center configuration object matrix representing the cluster center configuration object of the a-th cluster family in the t-th historical specified time period, Indicates that the operation label in the operation label set of the cluster center configuration object of the a-th cluster group in the t-th historical specified time period is successful. represents the number of operation tags in the operation tag set of the cluster center configuration object of the a-th cluster family, They respectively represent the ith operation label and the aN2th operation label in the operation label set of the cluster center configuration object of the ath cluster family in the tth historical specified time period, represents the operation success rate of the operation tags in the operation tag set of the cluster center configuration object of the a-th cluster group in the t-th historical specified time period, represents the operation failure rate of the ith operation label and the aN2th operation label in the operation label set of the cluster center configuration object of the ath cluster group in the tth historical specified time period, The operation label of the Lth historical operation instruction of the jth voice configuration object in the ath cluster of the voice configuration object subset in the tth historical specified time period, a represents the number of all historical operation instructions based on the jth voice configuration object in the voice configuration object subset of the ath cluster, Indicates the number of voice configuration objects in the voice configuration object subset of the a-th cluster in the t-th historical specified time period.

6. The optimization method for improving the recognition result of page speech intent according to claim 5, characterized in that: Determine the revised success vector of the cluster center configuration object for each cluster family, including: Determine a revised success vector of the cluster center configuration object of each cluster family based on the cluster center configuration object matrix of the cluster center configuration objects of each cluster family within all specified time periods; ； ； in, The modified success vector of the cluster center configuration object of the a-th cluster family They represent the correction success rates of the second operation label, the ith operation label, and the aN2th operation label in the operation label set of the cluster center configuration object of the ath cluster family, respectively. represents the period attenuation factor, represents the correction factor, and Nu represents the number of historical specified time periods.

7. The optimization method for improving the recognition result of page speech intent according to claim 6, characterized in that: Determine the correction mapping set, operation label set, and correction success vector for each page configuration object, including: Determine a revised mapping set for each page configuration object based on an object mapping set for each page configuration object in the preset voice configuration data, all page configuration objects in the page configuration object set, cluster core configuration objects of all cluster families, and a voice configuration object sub-set; ；； ； in, Represents the similarity value between the xth page configuration object and the ath cluster in the page configuration object set. represents the correction mapping set of the xth page configuration object in the page configuration object set, represents the cluster center configuration object of the a-th cluster family, represents the jth voice configuration object in the ath cluster group's voice configuration object subset, Represents the Euclidean norm of the xth page configuration object in the page configuration object collection, represents the Euclidean norm of the cluster center configuration object of the a-th cluster family, represents the Euclidean norm of the jth voice configuration object in the subset of voice configuration objects of the ath cluster, represents the subset of voice configuration objects of the a-th cluster, represents the voice configuration object sub-set with the largest similarity value to the x-th page configuration object in the page configuration object set, An object mapping collection representing the xth page configuration object in the page configuration object collection; Determine an operation tag set for each page configuration object in the page configuration object set based on an operation tag set of a cluster corresponding to a voice configuration object subset with a maximum similarity value to each page configuration object in the page configuration object set; Based on the correction success vector of the cluster group corresponding to the voice configuration object subset with the largest similarity value to each page configuration object in the page configuration object set, a correction success vector of each page configuration object in the page configuration object set is determined.

8. The optimization method for improving the recognition result of page speech intent according to claim 1, characterized in that: Based on the real-time speech recognition data and the modified speech configuration data of the intelligent speech control system, real-time intent data is determined, including: The intent model is optimized based on the correction vector data in the corrected voice configuration data, and the real-time voice recognition data is analyzed based on the optimized intent model. If the analysis is successful, the predicted intent data is determined, and the predicted intent data is matched with the correction mapping set of each page configuration object in the corrected voice configuration data, and the real-time intent data is determined based on the matching result; If the analysis is unsuccessful, a hard match is performed on the real-time speech recognition data to determine the real-time intent data.

9. The optimization method for improving the recognition result of page speech intent according to claim 8, characterized in that: Determine real-time intent data based on matching results, including: If there is a page configuration object in the modified voice configuration data that highly matches the predicted intent data, determine the real-time intent data of the real-time voice data based on the modified mapping set and the operation label set of the page configuration object that highly matches the predicted intent data; If there is no page configuration object in the revised voice configuration data that is highly matched with the predicted intent data, the revised mapping set and operation label set of the page configuration object that is most matched with the predicted intent data are revised again.

10. The optimization method for improving the recognition result of page speech intent according to claim 1, characterized in that: Determine real-time operation data based on the real-time intent data, and respond to the voice control system in real time based on the real-time operation data, including: Parsing the real-time intention data to determine real-time operation data, wherein the real-time operation data includes at least one or more real-time operation instructions; The voice control system executes each real-time operation instruction in the real-time operation data and determines an operation label of each real-time operation instruction; Based on all the real-time operation instructions in the real-time operation data and the operation label of each real-time operation instruction, current speech recognition sub-data of a current specified time period is determined.

Images