AI-based Intelligent Voice Customer Service Response Method and System for Water Services

Abstract

The present invention relates to the technical field of intelligent voice customer service response, and a water service intelligent voice customer service response method and system based on AI, including: obtaining a plurality of users, and performing the following operations on each user among the plurality of users: obtaining the user's voice, performing voice signal processing on the user's voice to obtain user voice features, retrieving matching voice features, detecting the feature similarity between the user voice features and the matching voice features, constructing a water service knowledge graph, obtaining inquiry information voice and encrypted user data, obtaining updated voice, using the updated voice as the user's voice, returning to the step of performing voice signal processing on the user's voice, obtaining the number of inquiries of the inquiry information voice, updating the water service voice database to obtain an optimized water service voice database, and completing the water service intelligent voice customer service response based on AI based on the encrypted user data set and the optimized water service voice database. The present invention can improve the efficiency and quality of water service customer service and enhance the user experience.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent voice customer service response, and particularly to an AI-based intelligent voice customer service response method and system for water utilities. Background Art

[0002] Water utilities refer to industries related to the development, utilization, management, and service of water resources. The intelligent voice customer service response method refers to a method that uses artificial intelligence technology, especially natural language processing and speech recognition technology, to implement automated voice interaction services.

[0003] With the acceleration of urbanization and the improvement of people's requirements for the quality of life, the demand for water utility services is also increasing continuously. However, traditional water utility customer service mainly relies on manual telephone services. Due to the different levels and experiences of manual customer service, the service quality is inconsistent, and misunderstandings and incorrect answers are likely to occur. Secondly, it is difficult for manual customer service to process a large amount of data and provide the latest water utility information in real time. Therefore, how to improve the efficiency and quality of water utility customer service and enhance the user experience is a technical problem that urgently needs to be solved. Summary of the Invention

[0004] The present invention provides an AI-based intelligent voice customer service response method for water utilities and a computer-readable storage medium, and its main purpose is to improve the efficiency and quality of water utility customer service and enhance the user experience.

[0005] To achieve the above object, an AI-based intelligent voice customer service response method for water utilities provided by the present invention includes:

[0006] Confirm an intelligent voice system for water utilities, where the intelligent voice system for water utilities includes: a semantic understanding enhancement unit, a water utility voice database, and an intelligent voice customer service;

[0007] Obtain multiple users, and perform the following operations on each user among the multiple users:

[0008] In the intelligent voice system for water utilities, obtain the user voice of the user, perform voice signal processing on the user voice to obtain user voice features, and retrieve matching voice features from the water utility voice database according to the user voice features;

[0009] Detect the feature similarity between the user voice features and the matching voice features, construct a water utility knowledge graph, and obtain inquiry information voice and encrypted user data according to the feature similarity and the water utility knowledge graph;

[0010] According to the inquiry information voice and the user, obtain an updated voice, use the updated voice as the user voice, and return to the step of performing voice signal processing on the user voice to obtain the inquiry times of the inquiry information voice;

[0011] Compare the inquiry times with a preset maximum number of inquiries;

[0012] If it is confirmed that the number of inquiries is greater than the preset maximum number of inquiries, the semantic understanding enhancement unit is used to deeply analyze the user voice to obtain the analyzed language data, and a solution is provided according to the analyzed language data;

[0013] The encrypted user data, the analyzed language data, and the solution are respectively summarized to obtain an encrypted user data set, an analyzed language data set, and a solution set, where the analyzed language data corresponds to the solution one by one;

[0014] The water service voice database is updated according to the analyzed language data set and the solution set to obtain an optimized water service voice database, and the AI-based water service intelligent voice customer service response is completed based on the encrypted user data set and the optimized water service voice database.

[0015] Optionally, the voice signal processing of the user voice to obtain user voice features includes:

[0016] Time series data is obtained according to the user voice, where the time series data includes multiple sampling points;

[0017] Sampling points are sequentially extracted from multiple sampling points, and the following operations are performed on each of the extracted sampling points:

[0018] Taking the sampling point as a precursor point, the next sampling point adjacent to the precursor point is extracted from multiple sampling points to obtain an adjacent sampling point;

[0019] Using a preset pre-emphasis coefficient and the precursor point to calculate the pre-emphasized sampling point of the adjacent sampling point, taking the adjacent sampling point as the precursor point, and returning to the step of extracting the next sampling point adjacent to the precursor point from multiple sampling points until all the multiple sampling points are extracted;

[0020] The pre-emphasized sampling points are summarized to obtain a pre-emphasized sampling point set, frame division parameters are set, and the pre-emphasized sampling point set is frame-divided using the frame division parameters to obtain a frame-divided sampling point segment set, where the frame-divided sampling point segment set includes: multiple frame-divided sampling point segments, and the number of sampling points in each frame-divided sampling point segment is the same;

[0021] Frame-divided sampling point segments are sequentially extracted from the frame-divided sampling point segment set, and the window function value of each sampling point in the frame-divided sampling point segment is calculated to obtain a window function value group;

[0022] A windowed signal value group is obtained according to the window function value group and each sampling point in the frame-divided sampling point segment;

[0023] A fast Fourier transform operation is performed on each windowed signal value in the windowed signal value group to obtain a frequency spectrum coefficient group, and the frequency domain energy spectrum of each frequency spectrum coefficient in the frequency spectrum coefficient group is calculated to obtain a frequency domain energy spectrum group;

[0024] Summarize the frequency-domain energy spectrum groups to obtain a set of frequency-domain energy spectrum groups, and obtain user voice features based on the set of frequency-domain energy spectrum groups.

[0025] Optionally, perform a fast Fourier transform operation on each windowed signal value in the windowed signal value group to obtain a group of spectrum coefficients, including:

[0026] Extract windowed signal values from the windowed signal value group in sequence, and perform the following operations on the extracted windowed signal values:

[0027] Calculate the spectrum coefficients of the windowed signal values using a pre-constructed fast Fourier transform formula, where the fast Fourier transform formula is as follows:

[0028] ;

[0029] Where represents the spectrum coefficient of the th windowed signal value in the windowed signal value group, represents the windowed signal value, represents the natural constant, represents the index of the windowed signal value, represents the preset imaginary unit, represents the number of windowed signal values in the windowed signal value group, represents pi, represents the preset time index;

[0030] Summarize the spectrum coefficients to obtain the group of spectrum coefficients corresponding to the windowed signal value group.

[0031] Optionally, the obtaining user voice features based on the set of frequency-domain energy spectrum groups includes:

[0032] Obtain a frequency range, and divide a triangular filter bank according to the frequency range, where the triangular filter bank includes multiple triangular filters and the frequency range of each triangular filter is different;

[0033] Perform the following operations on each frequency-domain energy spectrum group in the set of frequency-domain energy spectrum groups:

[0034] Filter each frequency-domain energy spectrum in the frequency-domain energy spectrum group using the triangular filter bank to obtain an energy spectrum weight group, and perform weighted summation on the frequency-domain energy spectrum group using the energy spectrum weight group to obtain filter energy;

[0035] Perform a logarithmic operation on the filter energy to obtain logarithmic energy coefficients, and perform a discrete cosine transform on the logarithmic energy coefficients to obtain cepstral coefficients;

[0036] Calculate the first-order difference of the cepstral coefficients, summarize the first-order differences to obtain a first-order difference group, and obtain the user speech features based on the first-order difference group.

[0037] Optionally, the filtering of each frequency-domain energy spectrum in the frequency-domain energy spectrum group by using the triangular filter bank to obtain an energy spectrum weight group includes:

[0038] Extract the frequency-domain energy spectra from the frequency-domain energy spectrum group in sequence, and perform the following operations on each of the extracted frequency-domain energy spectra:

[0039] Based on the frequency-domain energy spectrum, identify the target triangular filter from the triangular filter bank and calculate the center frequency of the target triangular filter;

[0040] Calculate the energy spectrum weight according to the center frequency and the frequency-domain energy spectrum, where the calculation formula of the energy spectrum weight is as follows:

[0041] ;

[0042] Where represents the energy spectrum weight of the rd target triangular filter in the triangular filter bank in the frequency-domain energy spectrum, represents the index of the triangular filter in the triangular filter bank, represents the th center frequency of the target triangular filter, represents the frequency-domain energy spectrum index, represents the th center frequency of the target triangular filter, represents the th center frequency of the target triangular filter;

[0043] Summarize the energy spectrum weights to obtain an energy spectrum weight group.

[0044] Optionally, the calculation of the first-order difference of the cepstral coefficients includes:

[0045] Obtain the order of the cepstral coefficients, and calculate the first-order difference of the cepstral coefficients by using the order of the cepstral coefficients. The calculation formula of the first-order difference of the cepstral coefficients is as follows:

[0046] ;

[0047] Where represents the first-order difference of the th cepstral coefficient, represents the th cepstral coefficient, represents the th cepstral coefficient, represents the order of the cepstral coefficients, represents a preset first derivative time difference represents an index represents the index of the cepstral coefficient

[0048] Optionally, the feature similarity between the detected user voice feature and the matching voice feature includes:

[0049] Obtain a language frequency word group according to the user voice feature, obtain the occurrence times of each language frequency word in the language frequency word group to obtain an occurrence times group, and obtain the high-frequency occurrence times according to the occurrence times group;

[0050] Extract the occurrence times from the occurrence times group in sequence, and calculate the user word frequency vector according to the high-frequency occurrence times and the occurrence times. The calculation formula of the user word frequency vector is as follows:

[0051] ;

[0052] where represents the user word frequency vector represents the th occurrence times of the language frequency word represents the high-frequency occurrence times represents a preset water service field corpus represents the th document number of the language frequency word represents the index of the language frequency word represents the logarithmic function with base 10;

[0053] Summarize the user word frequency vectors to obtain a user word frequency vector group, obtain the water service field word frequency vector group of the water service voice database, and calculate the feature similarity according to the user word frequency vector group and the water service field word frequency vector group.

[0054] Optionally, the obtaining the inquiry information voice and the encrypted user data according to the feature similarity and the water service knowledge graph includes:

[0055] Compare the feature matching degree with a preset feature matching degree threshold;

[0056] If the feature matching degree is greater than or equal to the preset feature matching degree threshold, obtain the water service associated information from the water service knowledge graph according to the user voice corresponding to the feature matching degree, use the intelligent voice customer service to perform voice output on the water service associated information, and perform real-time text output to obtain a smooth voice and a language text, and encrypt the smooth voice and the language text to obtain the encrypted user data;

[0057] If the feature matching degree is less than a preset feature matching degree threshold, the user voice corresponding to the feature matching degree is regarded as an unmatched voice, and an instruction to ask the user is sent according to the unmatched voice. According to the instruction to ask the user, the intelligent voice customer service sends an inquiry message voice to the user.

[0058] Optionally, calculating the feature similarity according to the user word frequency vector group and the water service field word frequency vector group includes:

[0059] Calculating the feature similarity according to the user word frequency vector group and the water service field word frequency vector group, where the calculation formula of the feature similarity is as follows:

[0060] ;

[0061] Wherein, represents the feature similarity, represents the weight value of the th language frequency word in the water service field word frequency vector group, represents the index of the language frequency word, represents the total number of language frequency words in the language frequency word group, represents the weight value of the th language frequency word in the user word frequency vector group.

[0062] To achieve the above object, the present invention also provides an AI-based water service intelligent voice customer service response system, including:

[0063] A user voice acquisition module for confirming a water service intelligent voice system, wherein the water service intelligent voice system includes: a semantic understanding enhancement unit, a water service voice database, and an intelligent voice customer service. Multiple users are acquired, and the following operations are performed on each user among the multiple users: In the water service intelligent voice system, the user voice of the user is acquired;

[0064] A voice signal processing module for performing voice signal processing on the user voice to obtain user voice features. According to the user voice features, a matching voice feature is retrieved from the water service voice database to detect the feature similarity between the user voice features and the matching voice features, construct a water service knowledge graph, and obtain an inquiry message voice and encrypted user data according to the feature similarity and the water service knowledge graph;

[0065] A user interaction module for obtaining an updated voice according to the inquiry message voice and the user, using the updated voice as the user voice, returning to the step of performing voice signal processing on the user voice, obtaining the number of inquiry times of the inquiry message voice, comparing the number of inquiry times with a preset maximum number of inquiry times. If it is confirmed that the number of inquiry times is greater than the preset maximum number of inquiry times, the semantic understanding enhancement unit is used to perform in-depth analysis on the user voice to obtain analyzed language data, and a solution is provided according to the analyzed language data;

[0066] A system optimization module is used to respectively summarize the encrypted user data, the parsed language data, and the solutions to obtain an encrypted user data set, a parsed language data set, and a solution set. Among them, the parsed language data corresponds to the solutions one by one. The water service voice database is updated according to the parsed language data set and the solution set to obtain an optimized water service voice database. Based on the encrypted user data set and the optimized water service voice database, the AI-based water service intelligent voice customer service response is completed.

[0067] To solve the above problems, the present invention also provides an electronic device, which includes:

[0068] A memory that stores at least one instruction;

[0069] A processor that executes the instructions stored in the memory to implement the above-mentioned AI-based water service intelligent voice customer service response method.

[0070] To solve the above problems, the present invention also provides a computer-readable storage medium, in which at least one instruction is stored, and the at least one instruction is executed by a processor in an electronic device to implement the above-mentioned AI-based water service intelligent voice customer service response method.

[0071] To solve the problems described in the background art, the present invention identifies a water service intelligent voice system. The water service intelligent voice system includes: a semantic understanding enhancement unit, a water service voice database, and an intelligent voice customer service. The present invention ensures that the system has complete functional modules, can process user voice input, understand user intentions, and provide accurate responses. Multiple users are obtained, and the following operations are performed for each user among the multiple users: The present invention provides personalized services for each user to improve the user experience. In the water service intelligent voice system, user voice is obtained, voice signal processing is performed on the user voice to obtain user voice features, and according to the user voice features, matching voice features are retrieved from the water service voice database. The present invention accurately extracts voice features through voice signal processing, improves the accuracy of voice recognition, and the extracted voice features are used for subsequent matching and analysis to enhance the robustness of the system. The feature similarity between the user voice features and the matching voice features is detected, a water service knowledge graph is constructed, and according to the feature similarity and the water service knowledge graph, inquiry information voice and encrypted user data are obtained. The present invention more intelligently matches user needs through similarity detection, improves the intelligence of the system, constructs a knowledge graph, and enhances the semantic understanding and knowledge association capabilities of the system. According to the inquiry information voice and the user, updated voice is obtained, and the updated voice is used as the user voice, and the step of performing voice signal processing on the user voice is returned to obtain the number of inquiries of the inquiry information voice. The present invention supports multi-round conversations, gradually refines user needs, and improves the accuracy of interactions. The number of inquiries is counted to provide a basis for subsequent in-depth analysis. The number of inquiries is compared with a preset maximum number of inquiries. The present invention limits the number of inquiries, avoids endless interactions, improves the efficiency of the system, reasonably allocates system resources, and ensures the stable operation of the system under high load. If it is confirmed that the number of inquiries is greater than the preset maximum number of inquiries, then the semantic understanding enhancement unit is used to perform in-depth analysis on the user voice to obtain post-analysis language data, and a solution is provided according to the post-analysis language data. The present invention more accurately understands user intentions through in-depth analysis, provides more accurate solutions, can handle complex user needs, and improves the intelligence and adaptability of the system. The encrypted user data, post-analysis language data, and solutions are respectively summarized to obtain an encrypted user data set, a post-analysis language data set, and a solution set. Among them, the post-analysis language data corresponds to the solutions one by one. The present invention updates the database to make it richer and more accurate, improves the performance of the system, provides AI-based intelligent responses, improves the user experience and system efficiency, updates the water service voice database according to the post-analysis language data set and the solution set to obtain an optimized water service voice database, and completes the AI-based water service intelligent voice customer service response based on the encrypted user data set and the optimized water service voice database. Therefore, the present invention can improve the efficiency and quality of water service customer service and enhance the user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] Figure 1Schematic flowchart of the AI-based intelligent voice customer service response method for water utilities provided by an embodiment of the present invention;

[0073] Figure 2 Functional module diagram of the AI-based intelligent voice customer service response system for water utilities provided by an embodiment of the present invention;

[0074] Figure 3 Schematic structural diagram of an electronic device for implementing the AI-based intelligent voice customer service response method provided by an embodiment of the present invention.

[0075] Description of reference numerals:

[0076] 1. Electronic device; 10. Processor; 11. Memory; 12. Bus.

[0077] The realization, functional features and advantages of the object of the present invention will be further described with reference to the embodiments and the accompanying drawings. Detailed implementation manners

[0078] It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0079] An embodiment of the present application provides an AI-based intelligent voice customer service response method for water utilities. The execution subject of the AI-based intelligent voice customer service response method includes but is not limited to at least one of electronic devices such as a server, a terminal, etc. that can be configured to execute the method provided by the embodiment of the present application. In other words, the AI-based intelligent voice customer service response method can be executed by software or hardware installed on a terminal device or a server device, and the software can be a blockchain platform. The server includes but is not limited to: a single server, a server cluster, a cloud server or a cloud server cluster, etc.

[0080] Refer to Figure 1 As shown, it is a schematic flowchart of the AI-based intelligent voice customer service response method provided by an embodiment of the present invention. In this embodiment, the AI-based intelligent voice customer service response method includes:

[0081] S1. Confirm the intelligent voice system for water utilities, where the intelligent voice system for water utilities includes: a semantic understanding enhancement unit, a water utility voice database, and an intelligent voice customer service.

[0082] It should be noted that the water service intelligent voice system refers to a system that integrates multiple functional modules and is used to process water service-related voice interaction services. It aims to utilize advanced voice technologies and intelligent algorithms to achieve more efficient, convenient, and intelligent communication in water services. The system interacts with users through voice, can understand users' voice needs and make corresponding responses, and provides users with various information and services related to water services. An intelligent voice customer service refers to a customer service system based on artificial intelligence technology. It realizes voice interaction with users through technologies such as speech recognition, natural language processing, and speech synthesis, and provides users with automated customer service and question answers.

[0083] S2. Obtain multiple users, and perform the following operations for each user among the multiple users: In the water service intelligent voice system, obtain the user's voice, perform voice signal processing on the user's voice to obtain user voice features, and retrieve matching voice features from the water service voice database according to the user voice features.

[0084] Exemplarily, Zhang said through the water service intelligent voice system: "I want to query this month's water fee", Zhao said through the water service intelligent voice system: "My water pipe is leaking and needs to be repaired", and Wang said through the water service intelligent voice system: "I want to know about water conservation policies", that is, Zhang, Zhao, and Wang are multiple users.

[0085] Specifically, the performing voice signal processing on the user's voice to obtain user voice features includes:

[0086] Obtain time series data according to the user's voice, where the time series data includes multiple sampling points;

[0087] Successively extract sampling points from the multiple sampling points, and perform the following operations on each of the extracted sampling points:

[0088] Take the sampling point as a precursor point, extract the next sampling point adjacent to the precursor point from the multiple sampling points to obtain an adjacent sampling point;

[0089] Calculate the pre-emphasized sampling point of the adjacent sampling point by using a preset pre-emphasis coefficient and the precursor point, take the adjacent sampling point as the precursor point, and return to the step of extracting the next sampling point adjacent to the precursor point from the multiple sampling points until all the multiple sampling points are extracted;

[0090] Summarize the pre-emphasized sampling points to obtain a set of pre-emphasized sampling points, set frame division parameters, and perform frame division on the set of pre-emphasized sampling points by using the frame division parameters to obtain a set of framed sampling point segments, where the set of framed sampling point segments includes: multiple framed sampling point segments, and the number of sampling points in each framed sampling point segment is the same;

[0091] Successively extract framed sampling point segments from the framed sampling point segment set, calculate the window function values of each sampling point in the framed sampling point segments to obtain a window function value group;

[0092] Obtain a windowed signal value group according to the window function value group and each sampling point in the framed sampling point segments;

[0093] Perform a fast Fourier transform operation on each windowed signal value in the windowed signal value group to obtain a spectrum coefficient group, and calculate the frequency-domain energy spectrum of each spectrum coefficient in the spectrum coefficient group to obtain a frequency-domain energy spectrum group;

[0094] Summarize the frequency-domain energy spectrum groups to obtain a set of frequency-domain energy spectrum groups, and obtain user voice features based on the set of frequency-domain energy spectrum groups.

[0095] It should be explained that time series data refers to a series of data arranged in chronological order obtained according to the user's voice when performing voice signal processing on the user's voice. The precursor point refers to the first sampling point among multiple sampling points. The adjacent sampling point is the next sampling point adjacent to the precursor point in the time series. The pre-emphasis coefficient refers to a preset value, which plays a role in emphasizing the high-frequency part of the voice signal in voice signal processing. Since the energy of the voice signal is usually concentrated in the low-frequency part and the high-frequency part is relatively weak, by using the pre-emphasis coefficient and the pre-emphasis formula to process the adjacent sampling points, the amplitude of the high-frequency signal can be increased, so as to more clearly extract the voice features. The pre-emphasized sampling point set refers to the set of all pre-emphasized sampling points obtained by summarizing after pre-emphasizing each sampling point in the time series data. The setting of the framing parameters refers to the operation of artificially setting parameters. The framing parameters refer to the parameters used for framing the pre-emphasized sampling point set. Among them, the framing parameters include the frame length and the frame shift. The frame length refers to the number of sampling points included in each framed sampling point segment, and the frame shift refers to the interval distance between two adjacent framed sampling point segments. The calculation formula in the step of calculating the pre-emphasized sampling point of the adjacent sampling point by using the preset pre-emphasis coefficient and the precursor point is as follows:

[0096] ;

[0097] Among them, represents the pre-emphasized sampling point, represents the adjacent sampling point, represents the pre-emphasis coefficient, represents the precursor point, represents the sampling point index;

[0098] It should be explained that the windowed signal value is the value obtained by multiplying the window function value by the time series data of the corresponding sampling point. The step of framing the pre-emphasized sampling point set using the framing parameters to obtain the framed sampling point segment set means that starting from the starting sampling point of the pre-emphasized sampling point set, a segment of sampling points is intercepted according to the frame length as a framed sampling point segment, and then according to the frame shift, a second sampling point is obtained, and the second sampling point is used as the starting sampling point, and so on, until the entire pre-emphasized sampling point set is framed, and the framed sampling point segment set is obtained.

[0099] Exemplarily, Xiao Zhang sets the frame length to: 4, the frame shift to: 2, and the pre-emphasized sampling point set to: {1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12}, where the starting sampling point is: 1, and the index of the starting sampling point is 0. According to a segment of sampling points with a frame length of 4, the framed sampling point segment obtained is: {1, 2, 3, 4}. Then, according to the frame shift of 2, that is, moving from the index 0 of the starting sampling point to the sampling point with index 2, a second sampling point is obtained, and the second sampling point is used as the starting sampling point, and so on until the entire pre-emphasized sampling point set is framed, and the framed sampling point segment set is obtained.

[0100] Importantly, the framed sampling point segment set refers to a set composed of multiple framed sampling point segments. The calculation formula in the step of calculating the window function value of each sampling point in the framed sampling point segment is as follows:

[0101] ;

[0102] where, represents the window function value of the th sampling point, represents a preset constant, represents the index of the sampling point, represents the total number of sampling points, represents the cosine function. The constant is a value used to control the shape of the window function. For example, the constant is 0.46.

[0103] Specifically, performing a fast Fourier transform operation on each windowed signal value in the windowed signal value group to obtain a spectrum coefficient group includes:

[0104] Successively extract the windowed signal values from the windowed signal value group, and perform the following operations on the extracted windowed signal values:

[0105] Calculate the spectrum coefficient of the windowed signal value using a pre-constructed fast Fourier transform formula, where the fast Fourier transform formula is as follows:

[0106] ;

[0107] where, Represents the spectral coefficient of the th windowed signal value in the set of windowed signal values, represents the windowed signal value, represents the natural constant, represents the index of the windowed signal value, represents the preset imaginary unit, represents the number of windowed signal values in the set of windowed signal values, represents pi, represents the preset time index;

[0108] Summarize the spectral coefficients to obtain the spectral coefficient group corresponding to the set of windowed signal values.

[0109] It should be explained that the spectral coefficient refers to the spectral value of the windowed signal value at the frequency index, which is used to describe the energy distribution of the windowed signal value in different frequency components. The imaginary unit is the unit that enables the fast Fourier transform to process complex signals, so as to capture both the amplitude and phase information of the signal at the same time. The time index refers to the sampling point position in the windowed signal value. The spectral coefficient group refers to the set composed of all spectral coefficients.

[0110] It can be understood that the frequency-domain energy calculation formula of the spectral coefficient in the step of calculating the frequency-domain energy spectrum of each spectral coefficient in the spectral coefficient group is as follows:

[0111] ;

[0112] where represents the frequency-domain energy. The frequency-domain energy spectrum group refers to the set composed of all frequency-domain energy spectra. Obtaining the user voice feature based on the frequency-domain energy spectrum group set refers to the statistical features extracted from the frequency-domain energy spectrum group set, such as mean, variance, peak value, etc.

[0113] Specifically, obtaining the user voice feature based on the frequency-domain energy spectrum group set includes:

[0114] Obtain the frequency range, and divide the triangular filter bank according to the frequency range. Among them, the triangular filter bank includes multiple triangular filters, and the frequency range of each triangular filter is different;

[0115] Perform the following operations on each frequency-domain energy spectrum group in the frequency-domain energy spectrum group set:

[0116] Filter each frequency-domain energy spectrum in the frequency-domain energy spectrum group by using the triangular filter bank to obtain the energy spectrum weight group, and perform weighted summation on the frequency-domain energy spectrum group by using the energy spectrum weight group to obtain the filter energy;

[0117] Perform a logarithmic operation on the filter energy to obtain logarithmic energy coefficients, and perform a discrete cosine transform on the logarithmic energy coefficients to obtain cepstral coefficients;

[0118] Calculate the first-order difference of the cepstral coefficients, aggregate the first-order differences to obtain a first-order difference group, and obtain the user voice features based on the first-order difference group.

[0119] It should be explained that the frequency range refers to the frequency interval of the frequency-domain energy spectrum set. The steps of dividing the triangular filter bank according to the frequency range are as follows: obtain the lowest frequency and the highest frequency according to the frequency range, and calculate the highest Mel frequency according to the highest frequency. The calculation formula for the highest Mel frequency is as follows:

[0120] ;

[0121] Among them, represents the highest Mel frequency, represents the logarithmic function, represents the highest frequency;

[0122] Take the lowest frequency as the lowest Mel frequency, obtain the Mel frequency range according to the lowest Mel frequency and the highest Mel frequency, and divide the Mel frequency range by the preset number of filters to obtain a triangular filter bank. The triangular filter bank is a group of filters evenly distributed in the Mel frequency range, used to distribute the spectral energy to different frequency intervals. A triangular filter refers to a simple filter with a triangular shape. The highest Mel frequency refers to the frequency obtained by converting the highest frequency to the Mel frequency scale. The Mel frequency refers to a non-linear frequency scale that is closer to the human ear's perception of frequency. The Mel frequency range refers to the interval from the lowest Mel frequency to the highest Mel frequency. The filter energy refers to the sum of the frequency-domain energy spectra weighted by the filter.

[0123] It should be explained that the step of performing a logarithmic operation on the filter energy to obtain logarithmic energy coefficients is as follows: calculate the logarithmic energy coefficients using the following formula. The calculation formula for the logarithmic energy coefficients is as follows:

[0124] ;

[0125] Among them, represents the logarithmic energy coefficient, represents the filter energy, represents the natural logarithm. The step of performing a discrete cosine transform on the logarithmic energy coefficients to obtain cepstral coefficients is as follows: calculate the cepstral coefficients using the following formula. The calculation formula is as follows:

[0126] ;

[0127] Among them, represents the cepstral coefficient, represents the number of triangular filters in the triangular filter bank, represents the index of the triangular filter, represents the index of the cepstral coefficient, represents the preset order of the cepstral coefficient. The first-order difference group refers to the set of first-order differences of all cepstral coefficients, which is used to represent the dynamic characteristics of the user's speech signal.

[0128] Specifically, filtering each frequency-domain energy spectrum in the frequency-domain energy spectrum group by using the triangular filter bank to obtain an energy spectrum weight group includes:

[0129] Sequentially extract the frequency-domain energy spectra from the frequency-domain energy spectrum group, and perform the following operations on the extracted frequency-domain energy spectra:

[0130] According to the frequency-domain energy spectrum, confirm the target triangular filter from the triangular filter bank, and calculate the center frequency of the target triangular filter;

[0131] Calculate the energy spectrum weight according to the center frequency and the frequency-domain energy spectrum. Among them, the calculation formula of the energy spectrum weight is as follows:

[0132] ;

[0133] Among them, represents the energy spectrum weight of the th target triangular filter in the frequency-domain energy spectrum, represents the index of the triangular filter in the triangular filter bank, represents the th center frequency of the target triangular filter, represents the frequency-domain energy spectrum index, represents the th center frequency of the target triangular filter, represents the th center frequency of the target triangular filter;

[0134] Summarize the energy spectrum weights to obtain an energy spectrum weight group.

[0135] It should be explained that the calculation formula in the step of calculating the center frequency of the target triangular filter is as follows:

[0136] ;

[0137] Among them, represents the th center frequency of the target triangular filter, Indicates the index of the target triangular filter.

[0138] Specifically, the calculation of the first-order difference of the cepstral coefficients includes:

[0139] Obtain the order of the cepstral coefficients, and calculate the first-order difference of the cepstral coefficients using the order of the cepstral coefficients. The calculation formula for the first-order difference of the cepstral coefficients is as follows:

[0140] ;

[0141] Where represents the first-order difference of the th cepstral coefficient, represents the th cepstral coefficient, represents the th cepstral coefficient, represents the order of the cepstral coefficients, represents the preset first-order derivative time difference, represents the index, represents the index of the cepstral coefficients.

[0142] It should be explained that the order of the cepstral coefficients refers to the dimension of the cepstral coefficient vector. For example, the order of the cepstral coefficients is 13. The first-order derivative time difference is used to determine the number of cepstral coefficients when calculating the weighted difference. The larger the first-order derivative time difference, the smoother the difference result; the smaller the first-order derivative time difference, the rougher the difference result. For example, the first-order derivative time difference is 2.

[0143] S3. Detect the feature similarity between the user speech features and the matching speech features, construct a water service knowledge graph, and obtain the query information speech and encrypted user data according to the feature similarity and the water service knowledge graph.

[0144] Specifically, the detection of the feature similarity between the user speech features and the matching speech features includes:

[0145] Obtain the language frequency word group according to the user speech features, obtain the occurrence times of each language frequency word in the language frequency word group to obtain the occurrence times group, and obtain the high-frequency occurrence times according to the occurrence times group;

[0146] Extract the occurrence times from the occurrence times group in sequence, and calculate the user word frequency vector according to the high-frequency occurrence times and the occurrence times. The calculation formula for the user word frequency vector is as follows:

[0147] ;

[0148] Where represents the user word frequency vector, represents the th occurrence time of the language frequency word, represents the high-frequency occurrence times, represents a preset corpus in the water service field, represents the number of documents of the nth language frequency word, represents the index of the language frequency word, represents the logarithmic function with base 10;

[0149] Summarize the user word frequency vectors to obtain a set of user word frequency vectors, obtain the word frequency vector set of the water service field in the water service speech database, and calculate the feature similarity according to the user word frequency vector set and the word frequency vector set of the water service field.

[0150] It should be explained that obtaining the language frequency word group according to the user speech feature means extracting the user speech text from the user speech feature, using natural language processing tools to segment the user speech text to obtain the language frequency word group. The language frequency word group is a set composed of all language frequency words. The high-frequency occurrence times refer to the language frequency word with the most occurrences in the language frequency word group. The user word frequency vector set is a set composed of all user word frequency vectors. The water service field corpus is a pre-set database used to contain words or idioms in the water service field. The number of documents of the language frequency word refers to the number of documents in which a certain language frequency word appears in the water service field corpus.

[0151] Specifically, obtaining the inquiry information speech and encrypted user data according to the feature similarity and the water service knowledge graph includes:

[0152] Compare the feature matching degree with a preset feature matching degree threshold;

[0153] If the feature matching degree is greater than or equal to the preset feature matching degree threshold, obtain the water service association information from the water service knowledge graph according to the user speech corresponding to the feature matching degree, use the intelligent voice customer service to output the water service association information in voice and perform real-time text output to obtain smooth speech and language text, and encrypt the smooth speech and language text to obtain encrypted user data;

[0154] If the feature matching degree is less than the preset feature matching degree threshold, use the user speech corresponding to the feature matching degree as the unmatched speech, send an inquiry user instruction according to the unmatched speech, and the intelligent voice customer service sends an inquiry information speech to the user according to the inquiry user instruction.

[0155] It should be noted that the feature matching degree threshold is a preset value used to determine whether the user word frequency vector group in the user's speech matches the water service field word frequency vector group. The water service knowledge graph refers to a database storing knowledge related to the water service field. The water service associated information refers to the information related to the user's speech extracted from the knowledge graph. The fluent speech refers to the natural speech answer generated by the intelligent voice customer service. The language text refers to the text content corresponding to the fluent speech. The encrypted user data refers to the data obtained by encrypting the fluent speech and the language text. The unmatched speech refers to the user's speech with a feature matching degree lower than the feature matching degree threshold. The query user instruction refers to the instruction issued by the intelligent voice customer service to query the user. The query information speech refers to the speech generated according to the query user instruction, used to query more information from the user. The encryption of the fluent speech and the language text means encrypting the fluent speech and the language text using an encryption algorithm. For example, the encryption algorithm is a symmetric encryption algorithm.

[0156] Specifically, the calculation of the feature similarity according to the user word frequency vector group and the water service field word frequency vector group includes:

[0157] Calculate the feature similarity according to the user word frequency vector group and the water service field word frequency vector group, where the calculation formula of the feature similarity is as follows:

[0158] ;

[0159] Where, represents the feature similarity, represents the weight value of the th language frequency word in the water service field word frequency vector group, represents the index of the language frequency word, represents the total number of language frequency words in the language frequency word group, represents the weight value of the th language frequency word in the user word frequency vector group.

[0160] It should be noted that the feature similarity refers to the similarity between the user word frequency vector group and the water service field word frequency vector group. The greater the feature similarity, the more similar the two vectors are. The weight value of the th language frequency word in the water service field word frequency vector group refers to the frequency of occurrence of this language frequency word in the water service field word frequency vector group. The weight value of the th language frequency word in the user word frequency vector group refers to the frequency of occurrence of this language frequency word in the user word frequency vector group.

[0161] S4. According to the query information speech and the user's obtained updated speech, take the updated speech as the user's speech, and return to the step of performing speech signal processing on the user's speech to obtain the number of queries of the query information speech.

[0162] It should be noted that the updated voice refers to the voice sent by the user in response to the information queried by the intelligent voice customer service. The number of inquiries refers to the number of times the voice customer service sends inquiry information voices to the user, which is used to evaluate the complexity and efficiency of the interaction.

[0163] S5. Compare the number of inquiries with the preset maximum number of inquiries. If it is confirmed that the number of inquiries is greater than the preset maximum number of inquiries, then use the semantic understanding enhancement unit to deeply analyze the user voice to obtain the analyzed language data, and provide a solution according to the analyzed language data.

[0164] It should be noted that the maximum number of inquiries is a preset value used to limit the number of interactions between the intelligent voice customer service and the user to ensure the efficiency of the interaction and the user experience. The use of the semantic understanding enhancement unit to deeply analyze the user voice means converting the user voice into a voice text, preprocessing the voice text to obtain a preprocessed text, where the preprocessing includes: word segmentation, part-of-speech tagging, and syntactic analysis, and using the semantic understanding enhancement unit to deeply analyze the preprocessed text, where the semantic understanding enhancement unit includes a deep learning model. For example, the deep learning model is BERT, GPT, etc.

[0165] It can be understood that the analyzed language data refers to the user language data deeply analyzed by the semantic understanding enhancement unit, which provides more accurate user intention and requirement information for generating a solution. The provision of a solution according to the analyzed language data means matching a solution from the water service knowledge graph according to the analyzed language data and generating responses in the form of voice and text.

[0166] S6. Aggregate the encrypted user data, the analyzed language data, and the solution respectively to obtain an encrypted user data set, an analyzed language data set, and a solution set, where the analyzed language data corresponds to the solution one by one.

[0167] It should be noted that the encrypted user data set refers to a set composed of all encrypted user data. The analyzed language data set refers to a set composed of all analyzed language data. The solution set refers to a set composed of all solutions.

[0168] S7. Update the water service voice database according to the analyzed language data set and the solution set to obtain an optimized water service voice database, and complete the AI-based water service intelligent voice customer service response based on the encrypted user data set and the optimized water service voice database.

[0169] It should be noted that the optimized water service voice database refers to a database that reflects user needs and system responses more accurately and comprehensively by updating and improving the data in the water service voice database. The encrypted user data set and the optimized water service voice database ensure the security and privacy of user data and improve the response quality and user experience.

[0170] To solve the problems described in the background art, the present invention identifies a water service intelligent voice system. The water service intelligent voice system includes: a semantic understanding enhancement unit, a water service voice database, and an intelligent voice customer service. The present invention ensures that the system has complete functional modules, can process user voice input, understand user intentions, and provide accurate responses. For multiple users, the following operations are performed for each user: The present invention provides personalized services for each user to improve the user experience. In the water service intelligent voice system, user voice is obtained, and voice signal processing is performed on the user voice to obtain user voice features. According to the user voice features, matching voice features are retrieved from the water service voice database. The present invention accurately extracts voice features through voice signal processing, improves the accuracy of voice recognition, and the extracted voice features are used for subsequent matching and analysis to enhance the robustness of the system. The feature similarity between the user voice features and the matching voice features is detected, a water service knowledge graph is constructed, and query information voice and encrypted user data are obtained based on the feature similarity and the water service knowledge graph. The present invention more intelligently matches user needs through similarity detection, improves the intelligence of the system, constructs a knowledge graph, and enhances the semantic understanding and knowledge association capabilities of the system. According to the query information voice and the user, updated voice is obtained, and the updated voice is used as the user voice, and the step of performing voice signal processing on the user voice is returned, and the query times of the query information voice are obtained. The present invention supports multi-round conversations, gradually refines user needs, and improves the accuracy of interactions. The query times are counted to provide a basis for subsequent in-depth analysis. The query times are compared with a preset maximum query times. The present invention limits the query times, avoids endless interactions, improves the efficiency of the system, reasonably allocates system resources, and ensures the stable operation of the system under high load. If it is confirmed that the query times are greater than the preset maximum query times, the semantic understanding enhancement unit is used to perform in-depth analysis on the user voice to obtain post-analysis language data, and a solution is provided according to the post-analysis language data. The present invention more accurately understands user intentions through in-depth analysis, provides more accurate solutions, can handle complex user needs, and improves the intelligence and adaptability of the system. The encrypted user data, the post-analysis language data, and the solutions are respectively summarized to obtain an encrypted user data set, a post-analysis language data set, and a solution set. Among them, the post-analysis language data and the solutions are in one-to-one correspondence. The present invention updates the database to make it richer and more accurate, improves the performance of the system, provides AI-based intelligent responses, improves the user experience and system efficiency. The water service voice database is updated according to the post-analysis language data set and the solution set to obtain an optimized water service voice database. Based on the encrypted user data set and the optimized water service voice database, an AI-based water service intelligent voice customer service response is completed. Therefore, the present invention can improve the efficiency and quality of water service customer service and enhance the user experience.

[0171] Such as Figure 2As shown in the figure, it is a functional module diagram of an AI-based intelligent voice customer service response system for water utilities provided by an embodiment of the present invention.

[0172] The AI-based intelligent voice customer service response system 100 for water utilities of the present invention can be installed in an electronic device. According to the functions achieved, the AI-based intelligent voice customer service response system 100 can include a user voice acquisition module 101, a voice signal processing module 102, a user interaction module 103, and a system optimization module 104. The modules of the present invention can also be referred to as units, which refer to a series of computer program segments that can be executed by a processor of an electronic device and can complete fixed functions, and are stored in the memory of the electronic device;

[0173] The user voice acquisition module 101 is used to confirm the intelligent voice system for water utilities. Among them, the intelligent voice system for water utilities includes: a semantic understanding enhancement unit, a water utility voice database, and an intelligent voice customer service. For multiple users, the following operations are performed on each user: In the intelligent voice system for water utilities, the user's voice is acquired;

[0174] The voice signal processing module 102 is used to perform voice signal processing on the user voice to obtain user voice features. According to the user voice features, matching voice features are retrieved from the water utility voice database, the feature similarity between the user voice features and the matching voice features is detected, a water utility knowledge graph is constructed, and query information voice and encrypted user data are obtained according to the feature similarity and the water utility knowledge graph;

[0175] The user interaction module 103 is used to update the voice according to the query information voice and the user acquisition, use the updated voice as the user voice, return to the step of performing voice signal processing on the user voice, obtain the query times of the query information voice, compare the query times with the preset maximum query times. If it is confirmed that the query times are greater than the preset maximum query times, the semantic understanding enhancement unit is used to deeply analyze the user voice to obtain the analyzed language data, and a solution is provided according to the analyzed language data;

[0176] The system optimization module 104 is used to summarize the encrypted user data, the analyzed language data, and the solutions respectively to obtain an encrypted user data set, an analyzed language data set, and a solution set. Among them, the analyzed language data and the solutions are in one-to-one correspondence. The water utility voice database is updated according to the analyzed language data set and the solution set to obtain an optimized water utility voice database, and the AI-based intelligent voice customer service response for water utilities is completed based on the encrypted user data set and the optimized water utility voice database.

[0177] Specifically, each module in the AI-based intelligent voice customer service response system 100 in the embodiment of the present invention is used in the same way as the above Figure 1The technical means are the same as those of the AI-based intelligent voice customer service response method for water utilities described in [reference], and can produce the same technical effects, which will not be elaborated here.

[0178] As Figure 3 shown, it is a schematic structural diagram of an electronic device for implementing the AI-based intelligent voice customer service response method for water utilities provided by an embodiment of the present invention.

[0179] The electronic device 1 may include a processor 10, a memory 11, and a bus 12, and may also include a computer program stored in the memory 11 and executable on the processor 10, such as the AI-based intelligent voice customer service response method program.

[0180] Among them, the memory 11 includes at least one type of readable storage medium, and the readable storage medium includes flash memory, mobile hard disk, multimedia card, card-type memory (such as SD or DX memory, etc.), magnetic memory, magnetic disk, optical disk, etc. The memory 11 may be an internal storage unit of the electronic device 1 in some embodiments, such as the mobile hard disk of the electronic device 1. The memory 11 may also be an external storage device of the electronic device 1 in other embodiments, such as a plug-in mobile hard disk, a Smart Media Card (SMC), a Secure Digital (SD) card, a Flash Card, etc. equipped on the electronic device 1. Further, the memory 11 also includes the internal storage unit of the electronic device 1 and the external storage device. The memory 11 can be used not only to store application software installed in the electronic device 1 and various types of data, such as the code of the AI-based intelligent voice customer service response method program, etc., but also to temporarily store data that has been output or will be output.

[0181] The processor 10 may be composed of integrated circuits in some embodiments. For example, it may be composed of a single packaged integrated circuit, or may be composed of multiple integrated circuits with the same or different functions, including a combination of one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and various control chips. The processor 10 is the control core (Control Unit) of the electronic device, connecting various components of the entire electronic device through various interfaces and circuits, and by running or executing programs or modules stored in the memory 11 (such as the AI-based intelligent voice customer service response method program, etc.), and calling data stored in the memory 11, to perform various functions of the electronic device 1 and process data.

[0182] The bus 12 may be a Peripheral Component Interconnect (PCI) bus, an Extended Industry Standard Architecture (EISA) bus, or the like. The bus 12 may be divided into an address bus, a data bus, a control bus, etc. The bus 12 is configured to enable connection and communication between the memory 11 and at least one processor 10, etc.

[0183] Figure 3 Only an electronic device with components is shown, and those skilled in the art can understand that Figure 3 the shown structure does not constitute a limitation on the electronic device 1, and it may include fewer or more components than shown, or combine certain components, or have a different component arrangement.

[0184] For example, although not shown, the electronic device 1 may further include a power source (such as a battery) for supplying power to each component. Preferably, the power source may be logically connected to the at least one processor 10 through a power management device, so as to implement functions such as charge management, discharge management, and power consumption management through the power management device. The power source may also include any components such as one or more DC or AC power sources, a recharge device, a power failure detection circuit, a power converter or inverter, and a power status indicator. The electronic device 1 may also include various sensors, a Bluetooth module, a Wi-Fi module, etc., which will not be elaborated here.

[0185] Further, the electronic device 1 may also include a network interface. Optionally, the network interface may include a wired interface and / or a wireless interface (such as a WI-FI interface, a Bluetooth interface, etc.), which is generally used to establish a communication connection between the electronic device 1 and other electronic devices.

[0186] Optionally, the electronic device 1 may further include a user interface. The user interface may be a display, an input unit (such as a keyboard), and optionally, the user interface may also be a standard wired interface or a wireless interface. Optionally, in some embodiments, the display may be an LED display, a liquid crystal display, a touch liquid crystal display, and an OLED (Organic Light-Emitting Diode) toucher, etc. Among them, the display may also be appropriately referred to as a display screen or a display unit, which is used to display the information processed in the electronic device 1 and to display a visual user interface.

[0187] The program of the AI-based intelligent voice customer service response method for water services stored in the memory 11 of the electronic device 1 is a combination of multiple instructions. When running in the processor 10, it can achieve the following:

[0188] Confirm the intelligent voice system for water services, where the intelligent voice system for water services includes: a semantic understanding enhancement unit, a water service voice database, and an intelligent voice customer service;

[0189] Obtain multiple users, and perform the following operations for each user among the multiple users:

[0190] In the intelligent voice system for water services, obtain the user's voice, perform voice signal processing on the user's voice to obtain user voice features, and retrieve matching voice features from the water service voice database according to the user voice features;

[0191] Detect the feature similarity between the user voice features and the matching voice features, construct a water service knowledge graph, and obtain the inquiry information voice and encrypted user data according to the feature similarity and the water service knowledge graph;

[0192] According to the inquiry information voice and the user, obtain an updated voice, use the updated voice as the user voice, and return to the step of performing voice signal processing on the user voice to obtain the number of inquiries of the inquiry information voice;

[0193] Compare the number of inquiries with a preset maximum number of inquiries;

[0194] If it is confirmed that the number of inquiries is greater than the preset maximum number of inquiries, then use the semantic understanding enhancement unit to deeply analyze the user voice to obtain the analyzed language data, and provide a solution according to the analyzed language data;

[0195] Aggregate the encrypted user data, the analyzed language data, and the solutions respectively to obtain an encrypted user data set, an analyzed language data set, and a solution set, where the analyzed language data and the solutions are in one-to-one correspondence;

[0196] Update the water service voice database according to the analyzed language data set and the solution set to obtain an optimized water service voice database, and complete the AI-based intelligent voice customer service response for water services based on the encrypted user data set and the optimized water service voice database.

[0197] Specifically, the specific implementation method of the processor 10 for the above instructions can refer to Figures 1 to 3 the description of the relevant steps in the corresponding embodiment, which will not be elaborated here.

[0198] Furthermore, if the modules / units integrated in the electronic device 1 are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. The computer-readable storage medium can be volatile or non-volatile. For example, the computer-readable medium can include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disc, computer memory, read-only memory (ROM, Read-Only Memory).

[0199] The present invention also provides a computer-readable storage medium storing a computer program, which when executed by a processor of an electronic device, can implement:

[0200] Identify a water service intelligent voice system, where the water service intelligent voice system includes: a semantic understanding enhancement unit, a water service voice database, and an intelligent voice customer service;

[0201] Obtain a plurality of users, and perform the following operations on each user among the plurality of users:

[0202] In the water service intelligent voice system, obtain the user's voice, perform voice signal processing on the user's voice to obtain user voice features, and retrieve matching voice features from the water service voice database according to the user voice features;

[0203] Detect the feature similarity between the user voice features and the matching voice features, construct a water service knowledge graph, and obtain inquiry information voice and encrypted user data according to the feature similarity and the water service knowledge graph;

[0204] According to the inquiry information voice and the user, obtain an updated voice, use the updated voice as the user voice, and return to the step of performing voice signal processing on the user voice to obtain the number of inquiries of the inquiry information voice;

[0205] Compare the number of inquiries with a preset maximum number of inquiries;

[0206] If it is confirmed that the number of inquiries is greater than the preset maximum number of inquiries, then use the semantic understanding enhancement unit to deeply analyze the user voice to obtain analyzed language data, and provide a solution according to the analyzed language data;

[0207] Aggregate the encrypted user data, the analyzed language data, and the solutions respectively to obtain an encrypted user data set, an analyzed language data set, and a solution set, where the analyzed language data and the solutions are in one-to-one correspondence;

[0208] Update the water service voice database according to the analyzed language dataset and solution set to obtain an optimized water service voice database, and complete the AI-based intelligent voice customer service response for water services based on the encrypted user dataset and the optimized water service voice database.

[0209] In several embodiments provided by the present invention, it should be understood that the disclosed devices, systems, and methods can be implemented in other ways. For example, the system embodiments described above are only illustrative, and there may be other partitioning methods in actual implementation.

[0210] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0211] In addition, the functional modules in each embodiment of the present invention can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. The above integrated unit can be implemented in the form of hardware or in the form of a hardware plus software functional module.

[0212] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above exemplary embodiments, and without departing from the spirit or basic characteristics of the present invention, the present invention can be implemented in other specific forms.

[0213] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. An AI-based intelligent voice customer service response method for water services, characterized in that, The method includes: Confirming a water service intelligent voice system, where the water service intelligent voice system includes: a semantic understanding enhancement unit, a water service voice database, and an intelligent voice customer service; Obtaining a plurality of users, and performing the following operations on each user among the plurality of users: In the water service intelligent voice system, obtaining the user's voice, performing voice signal processing on the user's voice to obtain user voice features, and retrieving matching voice features from the water service voice database according to the user voice features; Among them, the performing voice signal processing on the user's voice to obtain user voice features includes: Obtaining time series data according to the user's voice, where the time series data includes a plurality of sampling points; Successively extracting sampling points from the plurality of sampling points, and performing the following operations on each of the extracted sampling points: Taking the sampling point as a precursor point, extracting the next sampling point adjacent to the precursor point from the plurality of sampling points to obtain an adjacent sampling point; Calculating a pre-emphasized sampling point of the adjacent sampling point by using a preset pre-emphasis coefficient and the precursor point, taking the adjacent sampling point as the precursor point, and returning to the step of extracting the next sampling point adjacent to the precursor point from the plurality of sampling points until all the plurality of sampling points are extracted; Summarizing the pre-emphasized sampling points to obtain a set of pre-emphasized sampling points, setting framing parameters, and framing the set of pre-emphasized sampling points by using the framing parameters to obtain a set of framed sampling point segments, where the set of framed sampling point segments includes: a plurality of framed sampling point segments, and the number of sampling points in each framed sampling point segment is the same; Successively extracting framed sampling point segments from the set of framed sampling point segments, calculating the window function value of each sampling point in the framed sampling point segment to obtain a set of window function values; Obtaining a set of windowed signal values according to the set of window function values and each sampling point in the framed sampling point segment; Performing a fast Fourier transform operation on each windowed signal value in the set of windowed signal values to obtain a set of spectral coefficients, and calculating the frequency domain energy spectrum of each spectral coefficient in the set of spectral coefficients to obtain a set of frequency domain energy spectra; Summarizing the set of frequency domain energy spectra to obtain a set of frequency domain energy spectrum sets, and obtaining user voice features based on the set of frequency domain energy spectrum sets; Detecting the feature similarity between the user voice features and the matching voice features, constructing a water service knowledge graph, and obtaining an inquiry information voice and encrypted user data according to the feature similarity and the water service knowledge graph; According to the inquiry information voice and the user, obtaining an updated voice, taking the updated voice as the user voice, and returning to the step of performing voice signal processing on the user voice to obtain the number of inquiries of the inquiry information voice; Comparing the number of inquiries with a preset maximum number of inquiries; If it is confirmed that the number of inquiries is greater than the preset maximum number of inquiries, then using the semantic understanding enhancement unit to deeply analyze the user voice to obtain analyzed language data, and providing a solution according to the analyzed language data; Respectively summarizing the encrypted user data, the analyzed language data, and the solution to obtain an encrypted user data set, an analyzed language data set, and a solution set, where the analyzed language data corresponds to the solution one by one; Update the water service voice database according to the parsed language dataset and solution set to obtain an optimized water service voice database, and complete the AI-based intelligent voice customer service response for water services based on the encrypted user dataset and the optimized water service voice database.

2. The AI-based intelligent voice customer service response method for water affairs according to claim 1, wherein Performing a fast Fourier transform operation on each windowed signal value in the windowed signal value group to obtain a spectrum coefficient group, including: Successively extract windowed signal values from the windowed signal value group, and perform the following operations on the extracted windowed signal values: Calculate the spectrum coefficient of the windowed signal value using a pre-constructed fast Fourier transform formula, where the fast Fourier transform formula is as follows: ； Among them, represents the spectral coefficient of the th windowed signal value in the windowed signal value group, represents the windowed signal value, represents the natural constant, represents the index of the windowed signal value, represents the preset imaginary unit, represents the number of windowed signal values in the windowed signal value group, represents the pi, represents the preset time index; Summarize the spectrum coefficients to obtain the spectrum coefficient group corresponding to the windowed signal value group.

3. The AI-based intelligent voice customer service response method for water services according to claim 2, wherein, Obtaining user voice features based on the frequency domain energy spectrum group set, including: Obtain a frequency range, and divide a triangular filter bank according to the frequency range, where the triangular filter bank includes a plurality of triangular filters, and the frequency ranges of each triangular filter are different; Perform the following operations on each frequency domain energy spectrum group in the frequency domain energy spectrum group set: Filter each frequency domain energy spectrum in the frequency domain energy spectrum group using the triangular filter bank to obtain an energy spectrum weight group, and perform weighted summation on the frequency domain energy spectrum group using the energy spectrum weight group to obtain filter energy; Perform a logarithmic operation on the filter energy to obtain logarithmic energy coefficients, and perform a discrete cosine transform on the logarithmic energy coefficients to obtain cepstral coefficients; Calculate the first-order difference of the cepstral coefficients, summarize the first-order differences to obtain a first-order difference group, and obtain user voice features based on the first-order difference group.

4. The AI-based intelligent voice customer service response method for water services according to claim 3, wherein, Filtering each frequency domain energy spectrum in the frequency domain energy spectrum group using the triangular filter bank to obtain an energy spectrum weight group, including: Successively extract frequency domain energy spectra from the frequency domain energy spectrum group, and perform the following operations on the extracted frequency domain energy spectra: Identify a target triangular filter from the triangular filter bank according to the frequency domain energy spectrum, and calculate the center frequency of the target triangular filter; Calculate the energy spectrum weight according to the center frequency and the frequency domain energy spectrum, where the calculation formula of the energy spectrum weight is as follows: ； Among them, represents the energy spectrum weight of the th target triangular filter in the frequency domain energy spectrum, represents the index of the triangular filter in the triangular filter bank, represents the th center frequency of the target triangular filter, represents the frequency domain energy spectrum index, represents the th center frequency of the target triangular filter, represents the th center frequency of the target triangular filter; Summarize the energy spectrum weights to obtain an energy spectrum weight group.

5. The AI-based intelligent voice customer service response method for water utilities according to claim 4, wherein Calculating the first-order difference of the cepstral coefficients, including: Obtain the order of the cepstral coefficients, and calculate the first-order difference of the cepstral coefficients using the order of the cepstral coefficients, where the calculation formula of the first-order difference of the cepstral coefficients is as follows: ； Among them, represents the first-order difference of the th cepstral coefficient, represents the th cepstral coefficient, represents the th cepstral coefficient, represents the order of the cepstral coefficient, represents a preset first-order derivative time difference, represents an index, represents the index of the cepstral coefficient.

6. The AI-based intelligent voice customer service response method for water services according to claim 5, wherein, Detecting the feature similarity between the user voice features and the matching voice features, including: Obtain a language frequency word group according to the user voice features, obtain the occurrence times of each language frequency word in the language frequency word group to obtain an occurrence times group, and obtain the high-frequency occurrence times according to the occurrence times group; Successively extract the occurrence times from the occurrence times group, and calculate the user word frequency vector according to the high-frequency occurrence times and the occurrence times, where the calculation formula of the user word frequency vector is as follows: ； Among them, represents the user word frequency vector, represents the occurrence times of the th language frequent word, represents the high-frequency occurrence times, represents the preset water service field corpus, represents the number of documents of the th language frequent word, represents the logarithmic function with base 10; Summarize the user word frequency vectors to obtain a user word frequency vector group, obtain the water service domain word frequency vector group of the water service voice database, and calculate the feature similarity according to the user word frequency vector group and the water service domain word frequency vector group.

7. The AI-based intelligent voice customer service response method for water utilities according to claim 6, wherein Obtaining the inquiry information voice and the encrypted user data according to the feature similarity and the water service knowledge graph, including: Compare the feature matching degree with a preset feature matching degree threshold; If the feature matching degree is greater than or equal to the preset feature matching degree threshold, obtain water-related information from the water service knowledge graph according to the user voice corresponding to the feature matching degree, use the intelligent voice customer service to perform voice output of the water-related information, and perform real-time text output to obtain smooth voice and language text, and encrypt the smooth voice and language text to obtain encrypted user data; If the feature matching degree is less than the preset feature matching degree threshold, use the user voice corresponding to the feature matching degree as the unmatched voice, send an inquiry user instruction according to the unmatched voice, and according to the inquiry user instruction, the intelligent voice customer service sends an inquiry information voice to the user.

8. The AI-based intelligent voice customer service response method for water services according to claim 7, wherein The calculating the feature similarity according to the user word frequency vector group and the water service field word frequency vector group includes: Calculate the feature similarity according to the user word frequency vector group and the water service field word frequency vector group, wherein the calculation formula of the feature similarity is as follows: ； Among them, represents the feature similarity, represents the weight value of the th language frequency word in the water service field word frequency vector group, represents the index of the language frequency word, represents the total number of language frequency words in the language frequency word group, represents the weight value of the th language frequency word in the user word frequency vector group.

9. A system using the AI-based intelligent voice customer service response method for water services as described in claim 1, characterized in that, The system includes: A user voice acquisition module, used to confirm the water service intelligent voice system, wherein the water service intelligent voice system includes: a semantic understanding enhancement unit, a water service voice database and an intelligent voice customer service, obtain multiple users, and perform the following operations on each user among the multiple users: in the water service intelligent voice system, obtain the user voice of the user; A voice signal processing module, used to perform voice signal processing on the user voice to obtain user voice features, according to the user voice features, retrieve the matching voice features from the water service voice database to detect the feature similarity between the user voice features and the matching voice features, construct a water service knowledge graph, and obtain inquiry information voice and encrypted user data according to the feature similarity and the water service knowledge graph; A user interaction module, used to obtain updated voice according to the inquiry information voice and the user, use the updated voice as the user voice, return to the step of performing voice signal processing on the user voice, obtain the number of inquiries of the inquiry information voice, compare the number of inquiries with the preset maximum number of inquiries, if it is confirmed that the number of inquiries is greater than the preset maximum number of inquiries, then use the semantic understanding enhancement unit to perform in-depth analysis on the user voice to obtain the analyzed language data, and provide a solution according to the analyzed language data; A system optimization module, used to respectively summarize the encrypted user data, the analyzed language data and the solution to obtain an encrypted user data set, an analyzed language data set and a solution set, wherein the analyzed language data and the solution are in one-to-one correspondence, update the water service voice database according to the analyzed language data set and the solution set to obtain an optimized water service voice database, and complete the AI-based water service intelligent voice customer service response based on the encrypted user data set and the optimized water service voice database.

Images