Methods and devices for voice question and answer processing

Abstract

This invention discloses a method and apparatus for processing voice question-and-answer, relating to the field of data processing technology, with the main objective of addressing the poor processing performance of existing voice question-and-answer systems. The method includes: acquiring voice data collected for power grid dispatching services, and converting the voice data to obtain voice-text data; performing vector conversion on the voice-text data to obtain question vector data, and determining at least one business dimension of the question vector data, wherein the business dimension is obtained by parsing the collected power grid service data based on the power grid dispatching dimension; integrating the question vector data and the business dimension based on their dimensional correlation to generate a voice feedback result, and outputting the voice feedback result based on a pre-configured virtual playback object.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method and apparatus for processing voice question and answer. Background Technology

[0002] As the application of artificial intelligence (AI) technology continues to deepen across various industries, heavy industry enterprises are also beginning to use AI to assist in their production operations. For example, in power grid dispatching, corresponding operations need to be performed based on the power grid's production and operation, the power grid dispatching system, and personnel operations, which requires the input of relevant information.

[0003] Currently, the entry of information such as power grid dispatching typically relies on manual review, confirmation, planning, and scheduling. However, manual operation leads to slow response times in power grid dispatching operations, making it impossible to quickly execute relevant business requests to the power grid as needed. This results in long power grid dispatching work cycles, with the processing of related information consuming a significant amount of time. Therefore, there is an urgent need for a voice question-and-answer processing method to solve these problems. Summary of the Invention

[0004] In view of this, the present invention provides a method and apparatus for processing voice question and answer, the main purpose of which is to solve the problem of poor processing effect of existing voice question and answer systems.

[0005] According to one aspect of the present invention, a method for processing voice question answering is provided, comprising:

[0006] Acquire voice data collected for power grid dispatching services, and convert the voice data to obtain voice-text data;

[0007] The voice and text data are vectorized to obtain question vector data, and at least one business dimension of the question vector data is determined. The business dimension is obtained by parsing the collected power grid business data based on the power grid scheduling dimension.

[0008] The voice feedback result is generated by integrating the question vector data with the dimensional correlation of the business dimension, and the voice feedback result is output based on the pre-configured virtual playback object.

[0009] Furthermore, before acquiring the voice data collected for power grid dispatching services, the method further includes:

[0010] Retrieve power grid business data collected from different sources, and associate the power grid business data with the corresponding business dimension according to the power grid scheduling dimension matched by the collection source;

[0011] The power grid business data is stored according to the business dimensions.

[0012] Furthermore, storing the power grid business data according to the business dimension includes:

[0013] The power grid business data is converted into business vector data and stored as business vector data. The business vector data tv is determined based on the functional relationship between the number of times words appear in the power grid business data, the total number of words, the total number of data, and the quantification value of each power grid business data under different business dimensions.

[0014] Furthermore, the method also includes:

[0015] Establish correlations between multiple business dimensions, wherein the correlation is expressed as:

[0016] D=(∑(if(|tv i -tv j |≤α,1,0))) / S;

[0017] Among them, TV i TV j For each business vector data under two different business dimensions i and j, α is a preset vector threshold parameter, and S is the total number of business vectors in the business vector data dimension of the two different business dimensions i and j; if D≥β, then there is a correlation between the two different business dimensions i and j, otherwise there is no correlation; where β is a preset correlation threshold parameter.

[0018] Furthermore, the vector transformation of the speech-text data to obtain question vector data includes:

[0019] The speech text information is segmented into words, and the total number of keywords obtained after segmentation is counted.

[0020] Based on the total number of iterations, a vector transformation is performed to obtain the problem vector data;

[0021] Wherein, the problem vector data qv is represented as:

[0022] N k The frequency of each word k in the spoken text information, tv f,n For each business dimension n, there is business vector data; X is the total number of business vector data under each business dimension; B is the total number of business dimensions; and M is the total number of keywords in the voice text information.

[0023] Furthermore, the step of integrating the question vector data with the dimensional correlation of the business dimension to generate the voice feedback result includes:

[0024] The problem vector data is compared with the business vector data corresponding to the target business dimension in the business dimension according to the following constraints:

[0025] TV n For each business dimension n, u is a preset vector threshold parameter, and X is the total number of business vector data under each business dimension.

[0026] When the constraints are matched, the data is integrated based on the words corresponding to the business vector data to obtain the voice feedback result;

[0027] If the constraint is not matched, the problem vector data is re-compared with the business vector data corresponding to other business dimensions in the business dimension to determine the business vector data that matches the constraint.

[0028] Furthermore, the output of the voice feedback result based on the pre-configured virtual playback object includes:

[0029] Retrieve at least one configured virtual playback object, and play the voice feedback result according to at least one of the playback voice mode and playback scene mode corresponding to the target virtual playback object selected by the user.

[0030] According to another aspect of the present invention, a voice question-and-answer processing apparatus is provided, comprising:

[0031] The acquisition module is used to acquire voice data collected for power grid dispatching services and convert the voice data to obtain voice-text data.

[0032] The determination module is used to perform vector conversion on the voice text data to obtain question vector data, and to determine at least one business dimension of the question vector data, wherein the business dimension is obtained by parsing the collected power grid business data based on the power grid scheduling dimension;

[0033] The output module is used to integrate the question vector data with the dimensional correlation of the business dimension to generate a voice feedback result, and output the voice feedback result based on a pre-configured virtual playback object.

[0034] Furthermore, the device also includes:

[0035] The retrieval module is used to retrieve power grid business data collected from different sources, and associate the power grid business data with the corresponding business dimension according to the power grid scheduling dimension matched by the source.

[0036] The storage module is used to store the power grid business data according to the business dimension.

[0037] Furthermore, the storage module is specifically used to convert the power grid business data into business vector data and store it as business vector data. The business vector data tv is determined based on the functional relationship between the number of times words appear in the power grid business data, the total number of words, the total number of data, and the quantization value of each power grid business data under different business dimensions.

[0038] Furthermore, the device also includes:

[0039] The module is used to establish the correlation between multiple business dimensions, wherein the correlation is represented as follows:

[0040] D=(∑(if(|tv i -tv j |≤α,1,0))) / S;

[0041] Among them, TV i TV j For each business vector data under two different business dimensions i and j, α is a preset vector threshold parameter, and S is the total number of business vectors in the business vector data dimension of the two different business dimensions i and j; if D≥β, then there is a correlation between the two different business dimensions i and j, otherwise there is no correlation; where β is a preset correlation threshold parameter.

[0042] Furthermore, the determining module is specifically used to perform word segmentation on the speech text information and count the total number of keywords obtained after segmentation; based on the total number of keywords, a vector transformation is performed to obtain the question vector data;

[0043] Wherein, the problem vector data qv is represented as:

[0044] N k The frequency of each word k in the spoken text information, tv f,n For each business dimension n, there is business vector data; X is the total number of business vector data under each business dimension; B is the total number of business dimensions; and M is the total number of keywords in the voice text information.

[0045] Furthermore, the output module is specifically used to compare the problem vector data with the business vector data corresponding to the target business dimension in the business dimension according to constraints, wherein the constraints are: TV nFor each business dimension n, each business vector data is used, u is a preset vector threshold parameter, and X is the total number of business vector data under each business dimension. When the constraint condition is matched, the voice feedback result is obtained based on the words corresponding to the business vector data. When the constraint condition is not matched, the question vector data is re-compared with the business vector data corresponding to other business dimensions in the business dimension to determine the business vector data that matches the constraint condition.

[0046] Furthermore, the output module is specifically used to retrieve at least one configured virtual playback object, and play the voice feedback result according to at least one of the playback voice mode and playback scene mode corresponding to the target virtual playback object selected by the user.

[0047] According to another aspect of the present invention, a storage medium is provided, wherein at least one executable instruction is stored therein, the executable instruction causing a processor to perform an operation corresponding to the above-described voice question-and-answer processing method.

[0048] According to another aspect of the present invention, a terminal is provided, comprising: a processor, a memory, a communication interface, and a communication bus, wherein the processor, the memory, and the communication interface communicate with each other through the communication bus;

[0049] The memory is used to store at least one executable instruction, which causes the processor to perform the operation corresponding to the above-described voice question-and-answer processing method.

[0050] By employing the above-described technical solutions, the technical solutions provided by the embodiments of the present invention have at least the following advantages:

[0051] This invention provides a voice question-and-answer processing method and apparatus. Compared with the prior art, the embodiments of this invention acquire voice data collected for power grid dispatching services, convert the voice data to obtain voice-text data, perform vector conversion on the voice-text data to obtain question vector data, and determine at least one business dimension of the question vector data. The business dimension is obtained by parsing the collected power grid service data based on the power grid dispatching dimension. Based on the dimensional correlation between the question vector data and the business dimension, a voice feedback result is generated, and the voice feedback result is output based on a pre-configured virtual playback object. This achieves more accurate power grid dispatching question-and-answer results, increases the speed of question-and-answer matching for power grid dispatching related services, shortens the work cycle time, and reduces the time spent processing voice data during power grid dispatching, thereby meeting the processing needs of different power grid dispatching based on voice question-and-answer.

[0052] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, and in order to make the above and other objects, features and advantages of the present invention more apparent and understandable, specific embodiments of the present invention are described below. Attached Figure Description

[0053] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings:

[0054] Figure 1 A flowchart of a voice question-and-answer processing method provided by an embodiment of the present invention is shown;

[0055] Figure 2 This diagram illustrates a power grid dispatching question-and-answer system structure according to an embodiment of the present invention.

[0056] Figure 3 A block diagram of a voice question-and-answer processing device provided in an embodiment of the present invention is shown;

[0057] Figure 4 A schematic diagram of the structure of a terminal provided in an embodiment of the present invention is shown. Detailed Implementation

[0058] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0059] This invention provides a method for processing voice question and answer, such as... Figure 1 As shown, the method includes:

[0060] 101. Acquire voice data collected for power grid dispatching services, and convert the voice data to obtain voice-text data.

[0061] In this embodiment of the invention, the current execution terminal acts as a server for performing voice question-and-answer processing in power grid dispatching services. It can be a cloud server or a terminal server. Power grid dispatching services refer to the dispatching of electricity across different power transmission networks within a power system. Correspondingly, when a user inquires about power grid dispatching services, they can access the current execution terminal through a communication network, such as a telephone network or the internet. In this case, the current execution terminal acts as a server for an intelligent voice question-and-answer robot, processing the recorded voice data. This embodiment of the invention does not impose specific limitations. The voice data refers to one or more voice segments recorded by the user when inquiring about power grid dispatching services. Furthermore, the current execution terminal converts the voice data into voice-to-text data based on voice conversion technology, thereby processing the voice-to-text data. This embodiment of the invention does not impose specific limitations.

[0062] 102. Perform vector transformation on the voice text data to obtain question vector data, and determine at least one business dimension of the question vector data.

[0063] In this embodiment of the invention, after the current execution terminal obtains the voice text data, it performs vector conversion on the text data based on natural language processing technology to obtain question vector data. At this time, the question vector data is used to represent the numerical vector of the question generated by the user's inquiry about power grid dispatching services, and then the corresponding business dimension is matched based on the question vector data. Among them, the business dimension is obtained by parsing the collected power grid business data based on the power grid dispatching dimension. That is, the power grid dispatching dimension is used to represent the dimension that can classify the power grid business, including but not limited to the equipment type dimension, the jurisdiction dimension, the task urgency dimension, etc. Then, the current execution terminal pre-classifies the power grid business data according to the power grid dispatching dimension, and obtains the power grid business data associated with different classification dimensions, including but not limited to electricity fees, electricity usage duration, power outage distribution, etc. This embodiment of the invention does not make specific limitations.

[0064] 103. Based on the dimensional correlation between the question vector data and the business dimension, a voice feedback result is generated, and the voice feedback result is output based on a pre-configured virtual playback object.

[0065] In this embodiment of the invention, after determining the business dimension, the current execution end integrates the dimensional correlation between the question vector data and the business dimension to generate a voice feedback result. Since the question vector data can match at least one business dimension, in order to prepare an effective response, the current execution end calculates the dimensional correlation between the question vector data and multiple business dimensions to integrate and obtain the voice feedback result. Furthermore, to display diverse output effects, the current execution end schedules a pre-configured virtual playback object and outputs the voice feedback result according to the playback mode of this virtual playback object. At this time, the virtual playback object includes, but is not limited to, virtual voice playback objects, virtual visual playback objects, etc. Different virtual playback objects can be configured with different playback modes, such as different voices and tones; this embodiment of the invention does not impose specific limitations.

[0066] In another embodiment of the invention, to further define and illustrate, before the step of acquiring voice data collected for power grid dispatching services, the method further includes:

[0067] Retrieve power grid business data collected from different sources, and associate the power grid business data with the corresponding business dimension according to the power grid scheduling dimension matched by the collection source;

[0068] The power grid business data is stored according to the business dimensions.

[0069] To meet the requirement of multi-dimensional segmentation of voice data for accurate matching of responses, the current execution terminal first retrieves power grid business data from different collection sources. These sources include, but are not limited to, processing systems and transmission channels; this embodiment of the invention does not impose specific limitations. The power grid business data includes, but is not limited to, all data generated during power grid dispatching operations, such as electricity consumption and power supply duration. Furthermore, power grid dispatching dimensions are matched according to the collection source; different collection sources can be pre-configured with different power grid dispatching dimensions, such as equipment type, jurisdiction, or task urgency. This embodiment of the invention does not impose specific limitations. Additionally, business dimensions include, but are not limited to, electricity costs, power consumption duration, and power outage distribution. These can be matched based on a pre-configured mapping relationship between power grid dispatching dimensions and business dimensions to determine the association of different power grid dispatching dimensions with the power grid business data, thereby determining the business dimensions for storing the power grid business data.

[0070] In another embodiment of the invention, for further definition and explanation, the step of storing the power grid business data according to the business dimension includes:

[0071] The power grid business data is converted into business vector data and stored as business vector data.

[0072] To improve storage efficiency, when storing power grid business data according to business dimensions, it is specifically necessary to convert the power grid business data into business vector data. In this case, the business vector data `tv` is determined based on the functional relationship between the frequency of word occurrences, the total number of words, the total number of data points, and the quantified value of each power grid business data point under different business dimensions. In a specific implementation scenario, the business vector data `tv` is represented as:

[0073]

[0074] Among them, C w Let tx be the number of times each word w appears in each power grid dispatch data under each business dimension, tx be each power grid dispatch data under all business dimensions, A be the total number of power grid dispatch data under all business dimensions, Z be the total number of words in each power grid dispatch data under each business dimension, and if(x,y,z) means that if condition x is met, return result x, otherwise return result z.

[0075] In another embodiment of the invention, for further definition and explanation, the steps further include:

[0076] Establish the correlation between the various business dimensions.

[0077] To demonstrate the correlation between different power grid dispatch data across business dimensions, and to retrieve responses matching voice data based on this correlation, the current execution end, after determining multiple business dimensions, establishes the correlation between these dimensions. The correlation is expressed as follows:

[0078] D=(∑(if(|tv i -tv j |≤α,1,0))) / S;

[0079] Among them, TV i TV j For each business vector data under two different business dimensions i and j, α is a preset vector threshold parameter, and S is the total number of business vectors in the business vector data dimension of the two different business dimensions i and j; if D≥β, then there is a correlation between the two different business dimensions i and j, otherwise there is no correlation; where β is a preset correlation threshold parameter.

[0080] In another embodiment of the invention, for further definition and explanation, the step of performing vector conversion on the speech-text data to obtain question vector data includes:

[0081] The speech text information is segmented into words, and the total number of keywords obtained after segmentation is counted.

[0082] Based on the total number of iterations, a vector transformation is performed to obtain the problem vector data.

[0083] In order to accurately extract useful questions from the speech content during application, the current execution end performs vector conversion on the speech-text data to obtain question vector data. Specifically, it first performs word segmentation on the speech-text information. At this time, it can be based on word meaning in natural language processing technology to obtain words such as those containing modal particles, conjunctions, and names. This embodiment of the invention does not make specific limitations. At the same time, in order to avoid useless recognition, the current execution end can also delete meaningless words such as modal particles and conjunctions, retaining keywords. Then, it counts the total number of keywords after segmentation, and performs vector conversion according to the total number of times. Specifically, the question vector data qv is represented as:

[0084] N k The frequency of each word k in the spoken text information, tv f,n For each business dimension n, there is business vector data; X is the total number of business vector data under each business dimension; B is the total number of business dimensions; and M is the total number of keywords in the voice text information.

[0085] In another embodiment of the invention, for further definition and explanation, the step of generating voice feedback results based on the dimensional correlation between the question vector data and the business dimension includes:

[0086] The problem vector data is compared with the business vector data corresponding to the target business dimension in the business dimension according to the constraints.

[0087] When the constraints are matched, the data is integrated based on the words corresponding to the business vector data to obtain the voice feedback result;

[0088] If the constraint is not matched, the problem vector data is re-compared with the business vector data corresponding to other business dimensions in the business dimension to determine the business vector data that matches the constraint.

[0089] To improve the matching between question-and-answer vector data and business dimensions, thereby enhancing the accuracy and effectiveness of matching response content to voice data, the current execution end, when generating voice feedback results through dimension correlation integration, specifically compares the question vector data with the business vector data corresponding to the target business dimension according to constraints. The constraints are as follows:

[0090] TV nFor each business dimension n, each business vector data point is defined, where u is a pre-set vector threshold parameter, and X is the total number of business vector data points under each business dimension. In a specific implementation scenario, when the constraint condition is matched, the words corresponding to the business vector data are integrated to obtain a voice feedback result. When the constraint condition is not matched, the question vector data is re-compared with the business vector data corresponding to other business dimensions in the business dimension to determine the business vector data that matches the constraint condition. At this time, the other business dimensions are the uncompared business dimensions relative to the already compared business dimensions. This process continues until all business dimensions are compared or the constraint condition is matched. Then, the words corresponding to the matched business vector data are integrated to obtain a voice feedback result. During integration, the words can be combined and concatenated according to the logical order of language, such as subject, predicate, and object, to obtain a complete statement for feedback. This embodiment of the invention does not impose specific limitations on this.

[0091] In another embodiment of the invention, for further definition and explanation, the step of outputting the voice feedback result based on a pre-configured virtual playback object includes:

[0092] Retrieve at least one configured virtual playback object, and play the voice feedback result according to at least one of the playback voice mode and playback scene mode corresponding to the target virtual playback object selected by the user.

[0093] To meet the diverse display efficiency requirements of voice Q&A and enhance its playback effect, when the current execution terminal outputs voice feedback results, it first retrieves at least one configured virtual playback object. This virtual playback object can be a virtual sound constructed using 3D virtual technology, or it can be a virtual sound with different intonations configured based on intelligent voice technology, thus creating a realistic and flexible playback sound effect. After retrieving the virtual playback object, the user can select it as the target virtual playback object for playback. Different virtual playback objects are pre-configured with playback voice mode, playback image mode, and playback scene mode. After determining the target virtual playback object, the voice feedback result is played according to at least one of the corresponding playback voice mode and playback scene mode. The playback voice mode characterizes the speech rate, intonation, and other aspects of the voice playback effect, including but not limited to a child's voice playback mode or a woman's voice playback mode. The playback scene mode characterizes the scene of the voice playback, including but not limited to positive scenes, joyful scenes, and sad scenes. This embodiment of the invention does not impose specific limitations on these aspects. In addition, if the voice feedback result is output through an image in the embodiment of the present invention, the virtual playback object can also be a virtual human figure, such as a virtual anchor outputting the corresponding voice feedback result. The embodiment of the present invention does not make specific limitations.

[0094] In another embodiment of the present invention, the voice question-and-answer processing method of the present invention, when embedded into a computer system, may include, for example: Figure 2 The system architecture diagram shown illustrates various virtual units that execute corresponding methods. For example, the system includes: a power grid dispatch data acquisition unit, a database unit, a data preprocessing unit, a voice acquisition and recognition unit, a voice data processing unit, a comprehensive analysis and processing unit, a voice synthesis unit, and a virtual avatar construction unit; wherein:

[0095] Power grid dispatching data acquisition unit: used to collect business information data related to power grid dispatching;

[0096] Database unit: Used to associate and store business information data related to power grid dispatching collected by the power grid dispatching data acquisition unit according to preset dimensions;

[0097] Data preprocessing unit: Used for preprocessing operations such as transformation and mapping of business information data stored in the database unit;

[0098] Voice acquisition and recognition unit: used to acquire the user's voice and convert the acquired user voice into text information data;

[0099] The speech data processing unit is used to process the speech text information data converted by the speech acquisition and recognition unit, thereby converting it into question vector data.

[0100] The comprehensive analysis and processing unit is used to match the question vector data obtained by the voice data processing unit with dimensions that have high relevance, and integrate several dimensions with high relevance to form a feedback result.

[0101] Speech synthesis unit: used to synthesize speech based on the feedback results generated by the integrated analysis and processing unit;

[0102] Virtual avatar building unit: used to build virtual avatars that meet the requirements as needed, and to broadcast the speech synthesized by the speech synthesis unit through the virtual avatars.

[0103] Furthermore, the database unit includes: a dimension pre-setting module, a corresponding association module, and a data storage module;

[0104] Dimension Presetting Module: Used to set different power grid dispatch dimensions based on the source of power grid dispatch-related business data, including but not limited to the equipment type targeted by the power grid dispatch-related business data, the competent department of the power grid dispatch-related business, and the urgency of the dispatch tasks of the power grid dispatch-related business;

[0105] Correspondence module: used to associate the collected power grid dispatch data with different business dimensions;

[0106] Data storage module: Used to store power grid dispatch data from different business dimensions and the associated data under different business dimensions.

[0107] The data preprocessing unit consists of a data retrieval module, a data transformation module, and a return mapping module;

[0108] Data retrieval module: used to retrieve power grid dispatch data associated with different business dimensions from database units;

[0109] Data conversion module: used to convert the power grid dispatch data associated with different business dimensions retrieved by the data retrieval module into business vector data;

[0110] Feedback mapping module: used to map the business vector data converted by the data conversion module to the power grid dispatch data in the corresponding dimension.

[0111] When the data conversion module converts power grid dispatch data into business vector data, it converts every piece of power grid dispatch data in each dimension retrieved by the data retrieval module into business vector data tv.

[0112] After the data transformation module converts each power grid dispatch data under each dimension into business vector data, it establishes a correlation degree D between several business dimensions.

[0113] When the speech data processing unit converts speech text information data into question vector data, it first divides the speech text information data into several words, removes meaningless words such as interjections and conjunctions, obtains the number of occurrences of each remaining word, and then converts the speech text information data into question vector data qv.

[0114] The comprehensive analysis and processing unit includes: a data acquisition and matching module, a feedback result generation module, and a feedback result forwarding module;

[0115] Data acquisition and matching module: used to match dimensions from database cells that have a high degree of consistency with the question vector data;

[0116] Feedback Result Generation Module: This module integrates several dimensions that match the problem vector data highly with the data acquisition and matching module to form feedback results.

[0117] Feedback Result Forwarding Module: Used to forward the feedback results generated by the feedback result generation module to the speech synthesis unit.

[0118] During the process of the data acquisition and matching module matching dimensions with high consistency with the problem vector data from the database unit, the problem vector data is compared with each business vector data under each dimension. If the current dimension is determined to have a high degree of matching with the problem vector data, the comparison with business vector data under other dimensions is stopped. Based on the correlation between dimensions, other dimensions that are related to the current dimension are obtained.

[0119] In a specific scenario, the speech synthesis unit selects a voice suitable for the style and persona of the power grid dispatch virtual anchor to synthesize speech, thereby generating the virtual anchor's answer audio. The virtual avatar construction unit uses the audio results synthesized by the speech synthesis unit, along with lip prediction technology, facial recognition technology, and virtual anchor generation technology, to drive the virtual anchor to speak and generate image and audio frames, thus completing the construction of the virtual avatar.

[0120] This invention provides a voice question-and-answer processing method. Compared with the prior art, this invention acquires voice data collected for power grid dispatching services and converts the voice data to obtain voice-text data; it then performs vector conversion on the voice-text data to obtain question vector data and determines at least one business dimension of the question vector data. The business dimension is obtained by parsing the collected power grid service data based on the power grid dispatching dimension; a voice feedback result is generated based on the dimensional correlation between the question vector data and the business dimension, and the voice feedback result is output based on a pre-configured virtual playback object. This achieves more accurate power grid dispatching question-and-answer results, increases the speed of question-and-answer matching for power grid dispatching related services, shortens the work cycle time, and reduces the time spent processing voice data during power grid dispatching, thereby meeting the processing needs of different power grid dispatching based on voice question-and-answer.

[0121] Furthermore, as a response to the above Figure 1 The implementation of the method shown in this embodiment of the invention provides a voice question-and-answer processing device, such as... Figure 3 As shown, the device includes:

[0122] The acquisition module 21 is used to acquire voice data collected for power grid dispatching services and convert the voice data to obtain voice text data.

[0123] The determining module 22 is used to perform vector conversion on the voice text data to obtain question vector data, and to determine at least one business dimension of the question vector data, wherein the business dimension is obtained by parsing the collected power grid business data based on the power grid scheduling dimension;

[0124] Output module 23 is used to generate voice feedback results by integrating the problem vector data with the dimensional correlation of the business dimension, and output the voice feedback results based on a pre-configured virtual playback object.

[0125] Furthermore, the device also includes:

[0126] The retrieval module is used to retrieve power grid business data collected from different sources, and associate the power grid business data with the corresponding business dimension according to the power grid scheduling dimension matched by the source.

[0127] The storage module is used to store the power grid business data according to the business dimension.

[0128] Furthermore, the storage module is specifically used to convert the power grid business data into business vector data and store it as business vector data. The business vector data tv is determined based on the functional relationship between the number of times words appear in the power grid business data, the total number of words, the total number of data, and the quantization value of each power grid business data under different business dimensions.

[0129] Furthermore, the device also includes:

[0130] The module is used to establish the correlation between multiple business dimensions, wherein the correlation is represented as follows:

[0131] D=(∑(if(|tv i -tv j |≤α,1,0))) / S;

[0132] Among them, TV i TV j For each business vector data under two different business dimensions i and j, α is a preset vector threshold parameter, and S is the total number of business vectors in the business vector data dimension of the two different business dimensions i and j; if D≥β, then there is a correlation between the two different business dimensions i and j, otherwise there is no correlation; where β is a preset correlation threshold parameter.

[0133] Furthermore, the determining module is specifically used to perform word segmentation on the speech text information and count the total number of keywords obtained after segmentation; based on the total number of keywords, a vector transformation is performed to obtain the question vector data;

[0134] Wherein, the problem vector data qv is represented as:

[0135] N k The frequency of each word k in the spoken text information, tv f,nFor each business dimension n, there is business vector data; X is the total number of business vector data under each business dimension; B is the total number of business dimensions; and M is the total number of keywords in the voice text information.

[0136] Furthermore, the output module is specifically used to compare the problem vector data with the business vector data corresponding to the target business dimension in the business dimension according to constraints, wherein the constraints are: TV n For each business dimension n, each business vector data is used, u is a preset vector threshold parameter, and X is the total number of business vector data under each business dimension. When the constraint condition is matched, the voice feedback result is obtained based on the words corresponding to the business vector data. When the constraint condition is not matched, the question vector data is re-compared with the business vector data corresponding to other business dimensions in the business dimension to determine the business vector data that matches the constraint condition.

[0137] Furthermore, the output module is specifically used to retrieve at least one configured virtual playback object, and play the voice feedback result according to at least one of the playback voice mode and playback scene mode corresponding to the target virtual playback object selected by the user.

[0138] This invention provides a voice question-and-answer processing device. Compared with the prior art, this invention acquires voice data collected for power grid dispatching services, converts the voice data to obtain voice-text data, performs vector conversion on the voice-text data to obtain question vector data, and determines at least one business dimension of the question vector data. The business dimension is obtained by parsing the collected power grid service data based on the power grid dispatching dimension. Voice feedback results are generated based on the dimensional correlation between the question vector data and the business dimension, and the voice feedback results are output based on a pre-configured virtual playback object. This achieves more accurate power grid dispatching question-and-answer results, increases the speed of question-and-answer matching for power grid dispatching related services, shortens the work cycle time, and reduces the time spent processing voice data during power grid dispatching, thereby meeting the processing needs of different power grid dispatching based on voice question-and-answer.

[0139] According to one embodiment of the present invention, a storage medium is provided, the storage medium storing at least one executable instruction, the computer-executable instruction being capable of executing the voice question-and-answer processing method in any of the above method embodiments.

[0140] Figure 4 The diagram shows a structural schematic of a terminal according to an embodiment of the present invention. The specific implementation of the terminal is not limited by the specific embodiments of the present invention.

[0141] like Figure 4 As shown, the terminal may include: a processor 302, a communications interface 304, a memory 306, and a communications bus 308.

[0142] The processor 302, communication interface 304, and memory 306 communicate with each other via communication bus 308.

[0143] Communication interface 304 is used to communicate with other network elements such as clients or other servers.

[0144] The processor 302 is used to execute program 310, specifically to execute the relevant steps in the above-described voice question and answer processing method embodiment.

[0145] Specifically, program 310 may include program code that includes computer operation instructions.

[0146] Processor 302 may be a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement embodiments of the present invention. The terminal may include one or more processors of the same type, such as one or more CPUs; or it may include processors of different types, such as one or more CPUs and one or more ASICs.

[0147] Memory 306 is used to store program 310. Memory 306 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage device.

[0148] Specifically, program 310 can be used to cause processor 302 to perform the following operations:

[0149] Acquire voice data collected for power grid dispatching services, and convert the voice data to obtain voice-text data;

[0150] The voice and text data are vectorized to obtain question vector data, and at least one business dimension of the question vector data is determined. The business dimension is obtained by parsing the collected power grid business data based on the power grid scheduling dimension.

[0151] The voice feedback result is generated by integrating the question vector data with the dimensional correlation of the business dimension, and the voice feedback result is output based on the pre-configured virtual playback object.

[0152] It is obvious to those skilled in the art that the modules or steps of the present invention described above can be implemented using general-purpose computing devices. They can be centralized on a single computing device or distributed across a network of multiple computing devices. Optionally, they can be implemented using computer-executable program code, thereby storing them in a storage device for execution by a computing device. In some cases, the steps shown or described can be performed in a different order than those presented herein, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, the present invention is not limited to any particular combination of hardware and software.

[0153] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for processing voice question and answer, characterized in that, include: Acquire voice data collected for power grid dispatching services, and convert the voice data to obtain voice-text data; The voice and text data are vectorized to obtain question vector data, and at least one business dimension of the question vector data is determined. The business dimension is obtained by parsing the collected power grid business data based on the power grid scheduling dimension. Based on the dimensional correlation between the question vector data and the business dimension, a voice feedback result is generated, and the voice feedback result is output based on a pre-configured virtual playback object; The method further includes: Establish correlations between multiple business dimensions, wherein the correlation is expressed as: ； in, , For each business vector data under two different business dimensions i and j, the For a pre-defined vector threshold parameter, S represents the total number of business vectors in the business vector data dimensions of two different business dimensions i and j; if If i and j are related, then there is a correlation between the two different business dimensions i and j; otherwise, there is no correlation. Wherein, the... This is a preset correlation threshold parameter; The process of generating voice feedback results based on the dimensional correlation between the question vector data and the business dimension includes: The problem vector data is compared with the business vector data corresponding to the target business dimension in the business dimension according to the following constraints: ; wherein, the For each business vector data under each business dimension n, the... X is a preset vector threshold parameter, where X is the total number of business vector data under each business dimension, and qv is the problem vector data. When the constraints are matched, the data is integrated based on the words corresponding to the business vector data to obtain the voice feedback result; If the constraint is not matched, the problem vector data is re-compared with the business vector data corresponding to other business dimensions in the business dimension to determine the business vector data that matches the constraint.

2. The method according to claim 1, characterized in that, Before acquiring the voice data collected for power grid dispatching services, the method further includes: Retrieve power grid business data collected from different sources, and associate the power grid business data with the corresponding business dimension according to the power grid scheduling dimension matched by the collection source; The power grid business data is stored according to the business dimensions.

3. The method according to claim 2, characterized in that, The storage of the power grid business data according to the business dimension includes: The power grid business data is converted into business vector data and stored as business vector data. The business vector data tv is determined based on the functional relationship between the number of times words appear in the power grid business data, the total number of words, the total number of data, and the quantification value of each power grid business data under different business dimensions.

4. The method according to claim 1, characterized in that, The vector conversion of the speech-text data to obtain the question vector data includes: The speech text information is segmented into words, and the total number of keywords obtained after segmentation is counted. Based on the total number of iterations, a vector transformation is performed to obtain the problem vector data; Wherein, the problem vector data qv is represented as: , The frequency of each word k in the spoken text information. For each business dimension n, there is business vector data; X is the total number of business vector data under each business dimension; B is the total number of business dimensions; and M is the total number of keywords in the voice text information.

5. The method according to any one of claims 1-4, characterized in that, The output of the voice feedback result based on the pre-configured virtual playback object includes: Retrieve at least one configured virtual playback object, and play the voice feedback result according to at least one of the playback voice mode and playback scene mode corresponding to the target virtual playback object selected by the user.

6. A voice question-and-answer processing device, characterized in that, include: The acquisition module is used to acquire voice data collected for power grid dispatching services and convert the voice data to obtain voice-text data. The determination module is used to perform vector conversion on the voice text data to obtain question vector data, and to determine at least one business dimension of the question vector data, wherein the business dimension is obtained by parsing the collected power grid business data based on the power grid scheduling dimension; The output module is used to integrate the question vector data with the dimensional correlation of the business dimension to generate a voice feedback result, and output the voice feedback result based on a pre-configured virtual playback object; The device further includes: The module is used to establish the correlation between multiple business dimensions, wherein the correlation is represented as follows: ； in, , For each business vector data under two different business dimensions i and j, the For a pre-defined vector threshold parameter, S represents the total number of business vectors in the business vector data dimensions of two different business dimensions i and j; if If i and j are related, then there is a correlation between the two different business dimensions i and j; otherwise, there is no correlation. Wherein, the... This is a preset correlation threshold parameter; The output module is specifically used to compare the problem vector data with the business vector data corresponding to the target business dimension in the business dimension according to constraints, wherein the constraints are: ; For each business vector data under each business dimension n, the... X is a preset vector threshold parameter, where X is the total number of business vector data under each business dimension, and qv is the question vector data. When the constraint condition is matched, the voice feedback result is obtained by integrating the words corresponding to the business vector data. When the constraint condition is not matched, the question vector data is re-compared with the business vector data corresponding to other business dimensions in the business dimension to determine the business vector data that matches the constraint condition.

7. A storage medium storing at least one executable instruction that causes a processor to perform an operation corresponding to the voice question-and-answer processing method as described in any one of claims 1-5.

8. A terminal, comprising: The processor, memory, communication interface, and communication bus are provided, wherein the processor, memory, and communication interface communicate with each other via the communication bus. The memory is used to store at least one executable instruction, which causes the processor to perform the operation corresponding to the voice question-and-answer processing method as described in any one of claims 1-5.

Images