Audio-to-text processing method and device, equipment and storage medium

Abstract

The invention provides an audio-to-text processing method and device, equipment and a storage medium. Relates to the technical field of artificial intelligence. The method comprises the following steps: acquiring user audio information, user speed information and logic verification information; segmenting and / or merging the user audio information according to the user speech speed information and the service feature identifier to obtain initial text information; obtaining user region information according to the initial text information; determining general voice feature information according to the user region information; generating calibration adjustment information according to the general voice feature information and the reference voice feature information; generating basic text information according to the initial text information and the calibration adjustment information; obtaining target text information according to the basic text information and the logic verification information; the target text information is used for indicating the text content which can be used for generating the user appeal work order to output the target text information, and the problem that the text content generated by the scheme in the prior art is inaccurate is solved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to an audio-to-text processing method, device, equipment and storage medium. Background Art

[0002] When users handle business over the phone, customer service personnel usually need to record the user's request in the work order system to facilitate the next step of processing.

[0003] In the prior art, the user's telephone recording is often recognized and processed through voice recognition technology to generate corresponding text content, and then a corresponding work order is generated in the work order system according to the text content.

[0004] However, the solutions in the prior art have the problem of inaccurate text content. Summary of the invention

[0005] The present application provides an audio-to-text processing method, device, equipment and storage medium to solve the problem of inaccurate text content generated by solutions in the prior art.

[0006] In a first aspect, the present application provides an audio-to-text processing method, comprising: obtaining user audio information, user speech speed information and logic verification information; wherein the user audio information represents the user information input by voice when the user handles business through telephone voice customer service; the user speech speed information represents the speaking speed of the user when inputting the user information by voice; the logic verification information represents the content characteristics of the user information; according to the user speech speed information and the business feature identifier, the user audio information is segmented and / or merged to obtain initial text information; wherein the business feature identifier is used to indicate words with business characteristics; according to the initial text information, the user region information is obtained; wherein the user region information is used to indicate the user The region where the user of the user is located; according to the user region information, determining general voice feature information; wherein the general voice feature information represents the regional accent characteristics corresponding to the region where the user is located; generating calibration adjustment information according to the general voice feature information and reference voice feature information; wherein the reference voice feature information represents reference voice pronunciation characteristics; the calibration adjustment information represents the calibration adjustment characteristics when calibrating and adjusting the initial text information; generating basic text information according to the initial text information and the calibration adjustment information; obtaining target text information according to the basic text information and the logic check information; the target text information is used to indicate the text content that can be used to generate a user request work order; outputting the target text information.

[0007] In one possible design, the generating of calibration adjustment information based on the general voice feature information and the reference voice feature information includes: generating pre-reference information based on pending business information, the pending business information representing target business features of the business to be handled by the user, and the pre-reference information representing reference content corresponding to the target business features; generating user voice feature information in response to user verification information uploaded by the user based on the pre-reference information, the user verification information representing audio content corresponding to the reference content, and the user voice feature information representing the user's personal accent features; generating the calibration adjustment information based on the user voice feature information, the general voice feature information and the reference voice feature information.

[0008] In a possible design, the target text information is obtained according to the basic text information and the logical verification information, including: obtaining user identity information according to the initial text information, the user identity information representing user identity characteristics of the user, the user identity characteristics including user age and user gender; obtaining first verification pass information according to the basic text information and the logical verification information, the first verification pass information representing that the text content of the basic text information matches the content characteristics of the user information; obtaining second verification pass information according to the basic text information, the user identity information and the first verification pass information, the second verification pass information representing that the text content of the basic text information matches the user identity characteristics of the user; obtaining the target text information according to the basic text information and the second verification pass information.

[0009] In a possible design, the target text information is obtained according to the basic text information and the logical verification information, including: inputting the basic text information and the first preset identifier into N information extraction models to obtain N text extraction information, the text extraction information represents the text content extracted based on the first preset identifier, and N is an odd number greater than or equal to 3; determining target extraction information according to the N text extraction information, the target extraction information is used to indicate the text extraction information with the most votes determined from the N text extraction information based on the voting results; and obtaining the target text information according to the target extraction information and the logical verification information.

[0010] In a possible design, the target text information is obtained according to the target extraction information and the logical verification information, including: determining a target information extraction model from the N information extraction models according to the target extraction information; inputting the basic text information and the second preset identifier into the target information extraction model to obtain full extraction information, wherein the full extraction information represents the text content extracted based on the second preset identifier; and obtaining the target text information according to the full extraction information and the logical verification information.

[0011] In one possible design, the method also includes: quantifying the user voice information and / or the user speech speed information to obtain user emotional information, wherein the user voice information is used to indicate the volume and / or pitch of the user when inputting the user information by voice, and the user emotional information represents the user's attention to the business; matching preset grading information based on the user emotional information to determine a target grading identifier, wherein the preset grading information represents the processing priority when processing user demands, and the target grading identifier is used to indicate the target processing priority corresponding to the user demand work order; marking the target text information based on the target grading identifier to generate graded text information, and the graded text information is used to indicate text content that can be used to generate the user demand work order and has the target processing priority.

[0012] In a second aspect, the present application provides an audio-to-text processing device, comprising:

[0013] An acquisition module is used to acquire user audio information, user speech speed information and logic verification information; wherein the user audio information represents the user information input by voice when the user handles business through telephone voice customer service; the user speech speed information represents the speaking speed of the user when inputting the user information by voice; and the logic verification information represents the content characteristics of the user information;

[0014] A first processing module, configured to segment and / or merge the user audio information according to the user speech rate information and the service feature identifier to obtain initial text information; wherein the service feature identifier is used to indicate a word with a service feature;

[0015] The second processing module is used to obtain user region information according to the initial text information; wherein the user region information is used to indicate the region where the user is located; determine general voice feature information according to the user region information; wherein the general voice feature information represents the regional accent features corresponding to the region where the user is located; generate calibration adjustment information according to the general voice feature information and reference voice feature information; wherein the reference voice feature information represents reference voice pronunciation features; and the calibration adjustment information represents calibration adjustment features when calibrating and adjusting the initial text information;

[0016] The second processing module is further used to generate basic text information according to the initial text information and the calibration adjustment information;

[0017] A determination module, configured to obtain target text information according to the basic text information and the logic verification information, wherein the logic verification information represents content characteristics of the user information, and the target text information is used to indicate text content that can be used to generate a user request work order;

[0018] An output module is used to output the target text information.

[0019] In a possible design, when the second processing module generates the calibration adjustment information based on the general voice feature information and the reference voice feature information, it is specifically used to: generate pre-reference information based on the pending business information, the pending business information represents the target business features of the business to be handled by the user, and the pre-reference information represents the reference content corresponding to the target business features; generate user voice feature information in response to user verification information uploaded by the user based on the pre-reference information, the user verification information represents the audio content corresponding to the reference content, and the user voice feature information represents the personal accent features of the user; generate the calibration adjustment information based on the user voice feature information, the general voice feature information and the reference voice feature information.

[0020] In a possible design, when the determination module obtains the target text information based on the basic text information and the logical verification information, it is specifically used to: obtain user identity information based on the initial text information, the user identity information representing the user identity characteristics of the user, and the user identity characteristics including the user age and the user gender; obtain first verification pass information based on the basic text information and the logical verification information, the first verification pass information representing that the text content of the basic text information matches the content characteristics of the user information; obtain second verification pass information based on the basic text information, the user identity information and the first verification pass information, the second verification pass information representing that the text content of the basic text information matches the user identity characteristics of the user; obtain the target text information based on the basic text information and the second verification pass information.

[0021] In a possible design, when the determination module obtains the target text information based on the basic text information and the logical verification information, it is specifically used to: input the basic text information and the first preset identifier into N information extraction models to obtain N text extraction information, the text extraction information represents the text content extracted based on the first preset identifier, and N is an odd number greater than or equal to 3; determine the target extraction information based on the N text extraction information, the target extraction information is used to indicate the text extraction information with the most votes determined from the N text extraction information based on the voting results; obtain the target text information based on the target extraction information and the logical verification information.

[0022] In a possible design, when the determination module obtains the target text information based on the target extraction information and the logical verification information, it is specifically used to: determine the target information extraction model from the N information extraction models based on the target extraction information; input the basic text information and the second preset identifier into the target information extraction model to obtain full extraction information, and the full extraction information represents the text content extracted based on the second preset identifier; obtain the target text information based on the full extraction information and the logical verification information.

[0023] In one possible design, the determination module is also used to quantify the user voice information and / or the user speaking speed information to obtain user emotional information, wherein the user voice information is used to indicate the volume and / or pitch of the user when inputting the user information by voice, and the user emotional information represents the user's attention to the business handling; based on the user emotional information, the preset grading information is matched to determine the target grading identifier, and the preset grading information represents the processing priority when processing the user's request, and the target grading identifier is used to indicate the target processing priority corresponding to the user request work order; based on the target grading identifier, the target text information is marked to generate graded text information, and the graded text information is used to indicate the text content that can be used to generate the user request work order and has the target processing priority.

[0024] In a third aspect, an embodiment of the present application provides an electronic device, comprising: at least one processor and a memory; the memory stores computer-executable instructions; the at least one processor executes the computer-executable instructions stored in the memory, so that the at least one processor executes the audio-to-text processing method described in the first aspect and various possible designs of the first aspect.

[0025] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, in which computer execution instructions are stored. When a processor executes the computer execution instructions, the audio-to-text processing method described in the first aspect and various possible designs of the first aspect is implemented.

[0026] In a fifth aspect, an embodiment of the present application provides a computer program product, including a computer program. When the computer program is executed by a processor, it implements the audio-to-text processing method described in the first aspect and various possible designs of the first aspect.

[0027] The audio-to-text processing method, device, equipment and storage medium provided in the present application obtain user audio information, user speech speed information and logic verification information; wherein the user audio information represents the user information input by voice when the user handles business through telephone voice customer service; the user speech speed information represents the speaking speed of the user when inputting the user information by voice; the logic verification information represents the content characteristics of the user information; according to the user speech speed information and the business feature identifier, the user audio information is segmented and / or merged to obtain initial text information; wherein the business feature identifier is used to indicate words with business characteristics; according to the initial text information, the user region information is obtained; wherein the user region information is used to indicate the The region where the user is located; according to the user region information, general voice feature information is determined; wherein the general voice feature information represents the regional accent characteristics corresponding to the region where the user is located; according to the general voice feature information and reference voice feature information, calibration adjustment information is generated; wherein the reference voice feature information represents reference voice pronunciation characteristics; the calibration adjustment information represents the calibration adjustment characteristics when calibrating and adjusting the initial text information; according to the initial text information and the calibration adjustment information, basic text information is generated; according to the basic text information and the logic check information, target text information is obtained; the target text information is used to indicate the text content that can be used to generate a user request work order; and the target text information is output. On the basis of converting the user audio information into text content corresponding to the initial text information, the user's regional accent features are obtained, and the difference between the user's regional accent features and the reference speech pronunciation features is obtained through vector calculation, that is, calibration adjustment information based on the regional accent features is obtained, and then the user's colloquial text content in the recognized text content is calibrated and adjusted based on the calibration adjustment information, so as to obtain basic text information. Furthermore, in combination with the content features of the user information, such as the number of coding digits of the user's user code and the number of telephone digits of the contact number, a logical check is performed on the basic text information. If the check fails, the text content corresponding to the initial text information is recalibrated and adjusted. If the check passes, the text content that can be used to generate a user request work order can be obtained, which solves the problem of inaccurate text content generated by the solution in the prior art. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0029] Figure 1 An application scenario diagram of the audio-to-text processing method provided in an embodiment of the present application;

[0030] Figure 2A flowchart of an audio-to-text processing method provided in one embodiment of the present application;

[0031] Figure 3 for Figure 2 A schematic diagram of specific implementation steps of step S105 in the illustrated embodiment;

[0032] Figure 4 A schematic diagram of an interactive scenario for generating user voice feature information provided by this application;

[0033] Figure 5 for Figure 2 A schematic diagram of the specific implementation steps of step S107 in the illustrated embodiment;

[0034] Figure 6 A flowchart of an audio-to-text processing method provided in another embodiment of the present application;

[0035] Figure 7 A schematic diagram of determining target extraction information provided by an embodiment of the present application;

[0036] Figure 8 for Figure 6 A schematic diagram of the specific implementation steps of step S209 in the illustrated embodiment;

[0037] Fig. 9 A schematic diagram of the structure of an audio-to-text processing device provided in one embodiment of the present application;

[0038] Fig.10 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application.

[0039] The above drawings have shown clear embodiments of the present application, which will be described in more detail later. These drawings and text descriptions are not intended to limit the scope of the present application in any way, but to illustrate the concept of the present application to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0040] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application. Instead, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the appended claims.

[0041] In the technical solution of this application, the collection, storage, use, processing, transmission, provision and disclosure of information such as financial data or user data involved shall comply with the provisions of relevant laws and regulations and shall not violate public order and good morals.

[0042] It should be noted that in the embodiments of the present application, certain software, components, models and other existing solutions in the industry may be mentioned, and they should be regarded as exemplary. Their purpose is only to illustrate the feasibility of implementing the technical solution of the present application, but it does not mean that the applicant has or will necessarily use the solution.

[0043] The application scenarios of the embodiments of the present application are explained below:

[0044] Figure 1 FIG. 1 is an application scenario diagram of the audio-to-text processing method provided in the embodiment of the present application. The audio-to-text processing method provided in the embodiment of the present application can be applied to the scenario of converting audio information into text information. For example, Figure 1 As shown, the execution subject of the method provided in the embodiment of the present application can be an electronic control unit, a server or a terminal device. The terminal device is used as the execution subject for explanation. When the user handles business through telephone voice customer service, the terminal device on the customer service staff side stores the user's phone recording, and then the terminal device identifies and processes the phone recording to generate corresponding text content, and then the work order system generates a work order for solving the user's demands based on the text content, wherein the text content includes user personal information and user demand information. The user personal information is used for customer service personnel to return the user when solving the user's demands, and the user demand information is used to introduce the user's specific demands in detail.

[0045] Currently, in the prior art, the user's telephone recordings are often recognized and processed through voice recognition technology to generate corresponding text content. That is, the audio content in the user's telephone recording is directly converted through voice recognition technology based on the user's pronunciation, that is, the user's spoken content cannot be directly converted into text content that can be directly called by the work order system. For example, the user code of user people_1 is 15333346. When user people_1 registers his or her personal user code by phone, the audio content spoken by the user is "one, five, four threes, four, six". In the prior art solution, the converted text content is "1, 5, four threes, 4, 6", which is not the text content "15333346" that can be directly called by the work order system. Therefore, there is a problem of inaccurate text content generated by the solution in the prior art.

[0046] The technical solution of the present application and how the technical solution of the present application solves the above-mentioned technical problems are described in detail below with specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present application will be described below in conjunction with the accompanying drawings.

[0047] Figure 2 A flowchart of an audio-to-text processing method provided in one embodiment of the present application is shown in FIG. Figure 2 As shown, the execution subject of the audio-to-text processing method provided in this embodiment may be an electronic control unit, a server or a terminal device. By way of example, this embodiment uses the terminal device as the execution subject of the method of this embodiment for explanation. The audio-to-text processing method provided in this embodiment includes the following steps:

[0048] Step S101, obtaining user audio information, user speech speed information and logic verification information, wherein the user audio information represents the user information input by voice when the user handles business through telephone voice customer service; the user speech speed information represents the speaking speed of the user when inputting user information by voice; and the logic verification information represents the content characteristics of the user information.

[0049] Step S102, segmenting and / or merging the user audio information according to the user speech rate information and the service feature identifier to obtain initial text information; wherein the service feature identifier is used to indicate a word with service features.

[0050] Exemplarily, the user audio information represents the user information input by voice when the user handles business through the telephone voice customer service, wherein the user information input by the user has been authorized by the user to be used for the storage, use, processing and processing of the user information by the method of the present application; then, the user audio information is processed by natural language processing technology to obtain the original text content corresponding to the user information input by voice, and then the original text content is segmented and / or merged based on the user's speaking speed and business feature identifier, so as to obtain the text content corresponding to the user information input by voice, that is, the initial text information; specifically, for example, the user code of user people_1 is 15333346, when user people_1 registers his personal user code by phone, the audio content spoken by the user is "my user code is one, five, four threes, four, six", then the audio content in the user audio information is processed by natural language processing technology, and the generated original text content is "text_1, text_2, my user code is 1, 5, four threes, 4, 6, text_3, text_4", wherein text_1, text_2, text_3, text_4, _2, text_3 and text_4 are used to represent other content said by the user, and then the text content corresponding to the user code is determined from the original text content based on the service feature identifier "user code": "My user code is 1, 5, 4 3, 4, 6", and further, text_1, text_2, text_3 and text_4 are segmented and / or merged based on the user's speaking speed. For example, when the user's speaking speed when inputting text_1 and text_2 is greater than or equal to the preset speed, text_1 is divided into text_1_1 and text_2. ext_1_2, divide text_2 into text_2_1 and text_2_2, and when the user's speaking speed is lower than the preset speed when inputting text_3 and text_4 by voice, merge text_3 and text_4 into text_34; and then obtain the text content based on the segmentation and / or merging of the original text content: "text_1_1, text_1_2, text_2_1, text_2_2, my user code is 1, 5, 4 3, 4, 6, text_34", that is, obtain the initial text information.

[0051] In the steps of this embodiment, the user information of voice input is segmented and / or merged according to the speaking speed and words with business characteristics, and the initial text information composed of fragmented text content is obtained. Then, when the text content is calibrated by calibration adjustment information in the subsequent step S106, the calibration accuracy and efficiency are improved. At the same time, the fragmented segmentation form improves the readability of the text content.

[0052] Step S103, obtaining user region information according to the initial text information; wherein the user region information is used to indicate the region where the user is located.

[0053] Step S104, determining general voice feature information according to the user's region information; wherein the general voice feature information represents the regional accent features corresponding to the user's region.

[0054] Exemplarily, the user region information is used to indicate the user's region. When the user handles business through the telephone voice customer service, the telephone voice customer service actively asks the user about the user's region, and the user answers the region during the call. Then, the terminal device processes and parses the initial text information to obtain the text content fragment of the user's answer about the region in the initial text information, that is, the user region information is obtained. Further, after determining the user region information, the corresponding general voice feature information can be matched, that is, the regional accent feature corresponding to the user's region is determined; specifically, for example, for the area served by the telephone voice customer service, according to the regional accent feature, it can be divided into area A, area B and area C, the general voice feature information corresponding to area A is info_A, the general voice feature information corresponding to area B is info_B, and the general voice feature information corresponding to area C is info_C. Then, based on the determined user region information, for example, the user's region is area C, that is, based on area C, the user's regional accent feature is determined to be the accent feature of area C, that is, the general voice feature information info_C is obtained.

[0055] Step S105, generating calibration adjustment information according to the general speech feature information and the reference speech feature information; wherein the reference speech feature information represents reference speech pronunciation features; and the calibration adjustment information represents calibration adjustment features when calibrating and adjusting the initial text information.

[0056] Exemplarily, the reference speech feature information is used to indicate reference speech pronunciation features, that is, after obtaining the general speech feature information, the corresponding calibration adjustment information is generated according to the difference between the regional accent features corresponding to the general speech feature information and the reference speech pronunciation features corresponding to the reference speech feature information; specifically, for example, at least M words are determined according to the general speech feature information, and then according to the regional accent features corresponding to the general speech feature information, general accent vectors vector_C_i (i=1,2,3,...,M) of the M words are generated respectively, and according to the reference speech pronunciation features corresponding to the reference speech feature information, reference accent vectors vector_R_i of the M words are generated respectively, and then, the vector distance calculation is performed on the general accent vector vector_C_i and the reference accent vector vector_R_i to obtain M vector distances dist_i; further, by summing and averaging the vector distances dist_i, the average vector distance dist_ei can be obtained, that is, the calibration adjustment information is obtained.

[0057] In the steps of this embodiment, on the basis of determining the regional accent characteristics of the user, the difference between the regional accent characteristics of the user and the reference speech pronunciation characteristics is obtained through vector calculation, that is, calibration adjustment information based on the regional accent characteristics is obtained. Then in step S106, when the initial text information is calibrated and adjusted by the calibration adjustment information, the text content based on the regional accent characteristics will be calibrated and adjusted to the text content corresponding to the reference speech pronunciation characteristics, which not only solves the problem of inaccurate text content generated by the solution in the prior art, but also improves the readability and readability of the text content of the subsequently generated work order.

[0058] In a possible implementation, the text content corresponding to the initial text information can also be adjusted according to the user's personal accent characteristics. Therefore, before generating the basic text information according to the initial text information and the calibration adjustment information, the audio-to-text processing method provided by an embodiment of the present application includes a process of determining the calibration adjustment information having the personal accent characteristics. Figure 3 for Figure 2 The schematic diagram of the specific implementation steps of step S105 in the embodiment shown is as follows: Figure 3 As shown, the specific implementation steps of step S105 include the following steps:

[0059] Step S1051, generating pre-reference information according to pending business information, wherein the pending business information represents target business characteristics of the business to be handled by the user, and the pre-reference information represents reference content corresponding to the target business characteristics.

[0060] Step S1052, in response to the user verification information uploaded by the user according to the pre-reference information, user voice feature information is generated, the user verification information represents the audio content corresponding to the reference content, and the user voice feature information represents the user's personal accent feature.

[0061] Exemplarily, the pending business information is used to indicate the target business characteristics of the business to be handled by the user, and the pre-reference information is used to indicate the reference content corresponding to the target business characteristics, that is, based on the target business characteristics of the business to be handled by the user, the reference content associated with the target business characteristics is generated; then the user recites the reference content corresponding to the target business characteristics fed back by the telephone voice customer service, and obtains the audio content corresponding to the reference content, that is, obtains the user verification information; then the terminal device responds to the user verification information to extract the user's personal accent characteristics, that is, generates the user voice feature information.

[0062] Figure 4 A schematic diagram of an interactive scenario for generating user voice feature information provided by this application, such as Figure 4 As shown, the user makes an outbound call to the telephone voice customer service through a personal device, and the server responds to the user's outbound call and interacts with the user; then the telephone voice customer service provided by the server sends a voice message to the user "Hello, what business do you want to report?"; the user replies to the voice message of the telephone voice customer service "I want to report the business problem of task_A business"; the server (telephone voice customer service) responds to the target business feature of the task_A business that the user wants to report, and generates reference content text_A corresponding to the target business feature, and then the telephone voice customer service feedbacks to the user "Hello, please recite the reference content text_A first, and after the recitation, describe your feedback content in detail"; then, the user recites and reads the reference content text_A reported by the telephone voice customer service, and obtains user verification information; the server receives and stores the user verification information, and the terminal device on the customer service staff side obtains the user verification information from the server, and then extracts the feature of the audio content corresponding to the user verification information, and obtains the user's personal accent feature, that is, generates user voice feature information. Among them, when the telephone voice customer service feeds back the reference content text_A to the user, it can be displayed on the display screen of the personal device, and then the user recites the reference content text_A displayed on the display screen to obtain the user verification information corresponding to the reference content text_A. The telephone voice customer service can also play the reference content text_A sentence by sentence through the audio output unit of the personal device, and prompt the user to repeat it after each sentence is played to obtain the user verification information corresponding to the reference content text_A.

[0063] Step S1053: Generate calibration adjustment information according to the user voice feature information, the general voice feature information and the reference voice feature information.

[0064] Exemplarily, the reference voice feature information is used to indicate the reference voice pronunciation feature. After obtaining the user voice feature information, the corresponding calibration adjustment information is generated according to the difference between the user's personal accent feature corresponding to the user voice feature information and the reference voice pronunciation feature corresponding to the reference voice feature information. Specifically, for example, at least M words are determined according to the user voice feature information, and then according to the user's personal accent feature corresponding to the user voice feature information, personal accent vectors vector_P_j (j=1, 2, 3, ..., M) of the M words are generated respectively. According to the reference voice pronunciation feature corresponding to the reference voice feature information, the personal accent vectors vector_P_j (j=1, 2, 3, ..., M) of the M words are generated respectively. Generate reference accent vectors vector_R_j for M words respectively, and then perform vector distance calculation between the personal accent vector vector_P_j and the reference accent vector vector_R_j to obtain M vector distances dist_j; further, by performing weighted calculation on the vector distance dist_j and the vector distance dist_i, the average vector distance dist_ej can be obtained, that is, the calibration adjustment information is obtained, wherein the vector distance dist_i is obtained by performing vector distance calculation between the general accent vector vector_C_i and the reference accent vector vector_R_i in the above embodiment.

[0065] In the steps of this embodiment, when the user handles the business, the audio content corresponding to the reference content related to the business handled by the user is obtained to determine the user's personal accent characteristics, and then the difference between the user's personal accent characteristics and the reference speech pronunciation characteristics is obtained through vector calculation, that is, calibration adjustment information based on the personal accent characteristics is obtained. Then in step S106, when the initial text information is calibrated and adjusted by the calibration adjustment information, the text content based on the personal accent characteristics will be calibrated and adjusted to the text content corresponding to the reference speech pronunciation characteristics, which not only solves the problem of inaccurate text content generated by the solution in the prior art, but also improves the readability and readability of the text content of the subsequently generated work order.

[0066] Step S106: Generate basic text information according to the initial text information and the calibration adjustment information.

[0067] Exemplarily, the calibration adjustment information is used to indicate the calibration adjustment features when calibrating the initial text information. Based on the calibration adjustment information, the spoken text content in the text content corresponding to the initial text information is processed, and the text content based on the regional accent characteristics is calibrated and adjusted to the text content corresponding to the reference speech pronunciation characteristics, thereby generating the corresponding basic text information; specifically, for example, the text content corresponding to the initial text information is "My user code is 1, 5, 4 3s, 4, 6", and based on the text content corresponding to the user code, the "4 3s" are determined to be the spoken text content, and then the spoken text content "4 3s" is processed based on the calibration adjustment information to obtain the processed text content "3333", and then the basic text information is obtained, and the text content corresponding to the basic text information is "My user code is 15333346". In one possible implementation, the generation of basic text information is achieved based on a pre-trained large language model. Exemplarily, the large language model instructions corresponding to the calibration adjustment information include "please identify the spoken text content from the text content corresponding to the initial text information; process the identified spoken text content to generate the corresponding text content; further, combine the text content corresponding to the initial text information to obtain the full amount of text content, and then calibrate the full amount of text content based on regional accent features to the full amount of text content corresponding to reference speech pronunciation features". Then, the text content corresponding to the initial text information and the large language model instructions corresponding to the calibration adjustment information are input into the pre-trained large language model. The pre-trained large language model processes and calibrates the initial text information in response to the large language model instructions, thereby generating the calibrated and adjusted basic text information.

[0068] Step S107, obtaining target text information according to the basic text information and the logic verification information; the target text information is used to indicate the text content that can be used to generate a user request work order.

[0069] Step S108: output the target text information.

[0070] Exemplarily, the logical verification information is used to indicate the content characteristics of the user information. By performing a logical verification on the text content corresponding to the basic text information according to the content characteristics of the user information, the text content that can be used to generate a user request work order can be obtained based on the passed verification, that is, the target text information. In a possible implementation, the number of digits of the user code of the user is 8, and the number of digits of the user's contact phone number is 10. The number of digits of the user code and the number of digits of the contact phone number are used as content features of the user information, that is, to determine the logic verification information; then after obtaining the basic text information, the user code and the contact phone number of the user are determined from the basic text information, and the number of digits of the user code in the basic text information and the number of digits of the contact phone number in the basic text information are verified according to the logic verification information; if the number of digits of the user code and the number of digits of the contact phone number are both verified, the text content that can be used to generate a user request work order is obtained; if the verification fails, return to step S106 to recalibrate and adjust the initial text information, and when calling the large language model for calibration and adjustment, provide the large language model with the specific reason for the failure of the logic verification, for example, the number of digits of the user code fails to be verified, and the number of digits of the code corresponding to the basic text information is greater than the preset number of digits (8 digits).

[0071] Furthermore, the terminal device outputs the target text information in the form of text or audio for the customer service personnel to review.

[0072] In another possible implementation, Figure 5 for Figure 2 The process diagram of the specific implementation steps of step S107 in the embodiment shown is as follows: Figure 5 As shown, the specific implementation steps of step S107 include:

[0073] Step S1071, obtaining user identity information according to the initial text information, wherein the user identity information represents the user identity characteristics of the user, and the user identity characteristics include the user age and the user gender.

[0074] Exemplarily, user identity information is used to indicate the user identity characteristics of the user, where the user identity characteristics include the user age and user gender; when the user handles business through telephone voice customer service, the telephone voice customer service sends a voice message to the user "Please reply with your age and gender by voice", and then after the user replies with his age and gender, the user age and user gender of the user are obtained, that is, the user identity information is obtained; then, after obtaining the initial text information based on the user audio information, the terminal device can obtain the user age and user gender of the user by processing the initial text information, that is, the user identity information. Among them, since different regions and different users have different ways of calculating age, the user age corresponding to the user identity characteristic can be an age range or a specific age value.

[0075] Step S1072: obtaining first verification pass information according to the basic text information and the logic verification information, wherein the first verification pass information indicates that the text content of the basic text information matches the content feature of the user information.

[0076] Exemplarily, the implementation method of the steps of this embodiment is to perform logical verification on the number of coding digits of the user code and the number of telephone digits of the contact number, which is the same as the implementation method of the above embodiment and will not be repeated here. On the basis that the text content of the basic text information matches the content characteristics of the user information, step S1073 is executed. If the text content of the basic text information does not match the content characteristics of the user information, return to execute step S106.

[0077] Step S1073, obtaining second verification information according to the basic text information, the user identity information and the first verification information, wherein the second verification information indicates that the text content of the basic text information matches the user identity feature.

[0078] Step S1074, obtaining target text information according to the basic text information and the second verification pass information.

[0079] Exemplarily, the number of encoding bits of the user code of the user is 14 bits, of which the 5th to 8th bits are used to indicate the user's year of birth, and the 12th to 13th bits are used to indicate the user's gender; then, on the basis that the text content of the basic text information matches the content characteristics of the user information, that is, based on the first verification pass information, the user code of the user is obtained from the basic text information, and the first user code segment corresponding to the 5th to 8th bits and the second user code segment corresponding to the 12th to 13th bits are extracted from the user code; then, the user's target age can be determined according to the first user code segment, and the user's target gender can be determined according to the second user code segment, and then the user's target age and target gender are verified according to the user identity feature. If the user's target age is within the age range of the user age corresponding to the user identity feature (or, the error between the user's target age and the user age corresponding to the user identity feature is within the allowable error range), and the user's target gender is consistent with the user gender corresponding to the user identity feature, then it is determined that the text content of the basic text information matches the user's user identity feature, that is, the second verification pass information is obtained, and then the text content that can be used to generate a user request work order is obtained. If the verification fails, return to step S106 to recalibrate the initial text information, and when calling the large language model for calibration, provide the large language model with the specific reason why the logic verification failed, for example, the verification of the 12th to 13th digits of the user code failed, and the user's gender needs to be rechecked.

[0080] Specifically, for example, the first user code segment corresponding to the 5th to 8th bits is 1972, and the current year is 2024. The target age of the user is the difference between 2024 and 1972, that is, the target age of the user is 52 years old. When the second user code segment corresponding to the 12th to 13th bits is 01, the target gender of the user is male. When the second user code segment corresponding to the 12th to 13th bits is 02, the target gender of the user is female. Furthermore, based on the first user code segment corresponding to the 5th to 8th bits is 1965, it is determined that the target age of the user is 59 years old, and the second user code segment corresponding to the 12th to 13th bits is 03. The segment is 01, and the target gender of the user is determined to be male; the age range of the user determined based on the user identity information is 55 to 65 years old, and the user gender is male, that is, the user is a middle-aged male. Since the target age of the user (59 years old) is within the age range of the user age corresponding to the user identity feature (57 to 61 years old), the target gender of the user (male) is consistent with the user gender (male) corresponding to the user identity feature, so it is determined that the text content of the basic text information matches the user identity feature of the user, that is, the second verification pass information is obtained, and then the text content that can be used to generate a user request work order is obtained.

[0081] In the steps of this embodiment, by performing a verification based on the content characteristics of the user information and / or a logical verification based on the user identity characteristics of the user, the accuracy of the text content is improved, and the problem of inaccurate text content generated by the solutions in the prior art is solved.

[0082] On this basis, the audio-to-text processing method provided in the embodiment of the present application also includes: quantifying the user voice information and / or the user speaking speed information to obtain the user emotional information, the user voice information is used to indicate the volume and / or pitch of the user when inputting user information through voice, and the user emotional information represents the user's attention to the business; matching the preset grading information based on the user emotional information, determining the target grading identifier, the preset grading information represents the processing priority when processing the user's request, and the target grading identifier is used to indicate the target processing priority corresponding to the user request work order; marking the target text information based on the target grading identifier to generate graded text information, and the graded text information is used to indicate the text content that can be used to generate the user request work order and has the target processing priority.

[0083] In a possible implementation, the user voice information and / or the user speech speed information is quantified, and the volume of the user when inputting user information through voice is divided into volu_1, volu_2 and volu_3, the pitch is divided into tone_1, tone_2 and tone_3, and the speaking speed is divided into velo_1, velo_2 and velo_3, wherein the louder the volume, the higher the user's attention to the business, the higher the pitch, the higher the user's attention to the business, and the faster the speaking speed, the higher the user's attention to the business; and then by quantifying the volume, the user's voice information is divided into volu_1, volu_2 and volu_3, the pitch is divided into tone_1, tone_2 and tone_3, and the speaking speed is divided into velo_1, velo_2 and velo_3. The volume, pitch and speaking speed are weightedly calculated to obtain a comprehensive quantitative value. The larger the comprehensive quantitative value is, the higher the user's attention to the service is, that is, the user's emotional information is obtained; then, the preset hierarchical information represents the processing priority when processing user demands. For example, the numerical range of the comprehensive quantitative value corresponding to the third priority is (num_1, num_2), the numerical range of the comprehensive quantitative value corresponding to the second priority is (num_2, num_3), and the numerical range of the comprehensive quantitative value corresponding to the first priority is (num_3, num_4). Then, the corresponding comprehensive quantitative value is determined according to the comprehensive quantitative value num_0. For example, num_0 is between num_2 and num_3, so the second priority is determined, that is, the target processing priority is obtained, that is, the target hierarchical identifier is determined, wherein the first priority is the priority to be processed first, followed by the second priority, and finally the third priority; further, by marking the target text information with the target hierarchical identifier, the text content with the target processing priority that can be used to generate the user request work order can be obtained, for example, the target hierarchical identifier corresponding to the target text information cont_1 is the second priority, the target hierarchical identifier corresponding to the target text information cont_2 is the third priority, and the target hierarchical identifier corresponding to the target text information cont_3 is the first priority, and then the hierarchical text information cont_1_2 corresponding to the target text information cont_1, the hierarchical text information cont_2_3 corresponding to the target text information cont_2, and the hierarchical text information cont_3_1 corresponding to the target text information cont_3 are obtained, that is, the processing order of the above three hierarchical text information is obtained: the hierarchical text information cont_3_1 is processed first, the hierarchical text information cont_1_2 is processed second, and the hierarchical text information cont_1_2 is processed last.In this embodiment, by obtaining user audio information, user speech speed information and logic verification information; wherein the user audio information represents the user information input by voice when the user handles business through telephone voice customer service; the user speech speed information represents the speaking speed of the user when inputting user information by voice; the logic verification information represents the content characteristics of the user information; according to the user speech speed information and the business feature identifier, the user audio information is segmented and / or merged to obtain initial text information; wherein the business feature identifier is used to indicate words with business features; according to the initial text information, the user region information is obtained; wherein the user region information is used to indicate the user's user region ; Determine general voice feature information based on user region information; wherein the general voice feature information represents the regional accent features corresponding to the user's region; generate calibration adjustment information based on the general voice feature information and reference voice feature information; wherein the reference voice feature information represents reference voice pronunciation features; the calibration adjustment information represents the calibration adjustment features when calibrating and adjusting the initial text information; generate basic text information based on the initial text information and the calibration adjustment information; obtain target text information based on the basic text information and the logic verification information; the target text information is used to indicate text content that can be used to generate a user request work order; output the target text information. On the basis of converting the user audio information into text content corresponding to the initial text information, the user's regional accent features are obtained, and the difference between the user's regional accent features and the reference speech pronunciation features is obtained through vector calculation, that is, calibration adjustment information based on the regional accent features is obtained, and then the user's colloquial text content in the recognized text content is calibrated and adjusted based on the calibration adjustment information, so as to obtain basic text information. Furthermore, in combination with the content features of the user information, such as the number of coding digits of the user's user code and the number of telephone digits of the contact number, a logical check is performed on the basic text information. If the check fails, the text content corresponding to the initial text information is recalibrated and adjusted. If the check passes, the text content that can be used to generate a user request work order can be obtained, which solves the problem of inaccurate text content generated by the solution in the prior art.

[0084] Figure 6 A flowchart of an audio-to-text processing method provided in another embodiment of the present application is shown in FIG. Figure 6 As shown, the audio-to-text processing method provided in this embodiment is Figure 2 Based on the audio-to-text processing method provided in the illustrated embodiment, step 107 is further refined, and the audio-to-text processing method provided in this embodiment includes the following steps:

[0085] Step S201, obtaining user audio information, user speech speed information and logic verification information; wherein the user audio information represents the user information input by voice when the user handles business through telephone voice customer service; the user speech speed information represents the speaking speed of the user when inputting user information by voice; and the logic verification information represents the content characteristics of the user information.

[0086] Step S202, segmenting and / or merging the user audio information according to the user speech rate information and the service feature identifier to obtain initial text information; wherein the service feature identifier is used to indicate a word with service features.

[0087] Step S203, obtaining user region information according to the initial text information; wherein the user region information is used to indicate the region where the user is located.

[0088] Step S204, determining general voice feature information according to the user's region information; wherein the general voice feature information represents the regional accent features corresponding to the user's region.

[0089] Step S205, generating calibration adjustment information according to the general speech feature information and the reference speech feature information; wherein the reference speech feature information represents reference speech pronunciation features; and the calibration adjustment information represents calibration adjustment features when calibrating and adjusting the initial text information.

[0090] Step S206: Generate basic text information according to the initial text information and the calibration adjustment information.

[0091] Step S207: input the basic text information and the first preset identifier into N information extraction models to obtain N text extraction information, where the text extraction information represents the text content extracted based on the first preset identifier, and N is an odd number greater than or equal to 3.

[0092] Step S208: determining target extraction information according to the N text extraction information, where the target extraction information is used to indicate the text extraction information with the most votes determined from the N text extraction information based on the voting result.

[0093] Exemplarily, the text extraction information is used to indicate the text content extracted by the information extraction model from the basic text information based on the first preset identifier; that is, the basic text information and the first preset identifier are input into N information extraction models to obtain N text extraction information corresponding to the first preset identifier, where N is an odd number greater than or equal to 3; then, the N text extraction information corresponding to each first preset identifier is voted, and the text extraction information with the most votes among the N text extraction information is determined as the target extraction information.

[0094] In one possible implementation, Figure 7A schematic diagram of determining target extraction information provided by an embodiment of the present application, such as Figure 7 As shown, the first preset identifier includes a first preset identifier L_1, a first preset identifier L_2 and a first preset identifier L_3, and the information extraction model includes an information extraction model M_1, an information extraction model M_2 and an information extraction model M_3; based on the first preset identifier L_1, the text extraction information extracted from the basic text information by the information extraction model M_1, the information extraction model M_2 and the information extraction model M_3 is S_1_1, S_1_1, S_1_2 respectively; based on the first preset identifier L_2, the text extraction information extracted from the basic text information by the information extraction model M_1, the information extraction model M_2 and the information extraction model M_3 is S_2_1, S_2_2, S_2_2 respectively; based on the first preset identifier L_3, the text extraction information extracted from the basic text information by the information extraction model M_1, the information extraction model M_2 and the information extraction model M_3 is S_3 _1, S_3_1, S_3_2; then, vote on the three text extraction information corresponding to each first preset identifier, and then obtain the voting result. The text extraction information S_1_1 gets two votes and the text extraction information S_1_2 gets one vote, so the text extraction information S_1_1 is determined as the target extraction information corresponding to the first preset identifier L_1, the text extraction information S_2_1 gets one vote and the text extraction information S_2_2 gets two votes, so the text extraction information S_2_2 is determined as the target extraction information corresponding to the first preset identifier L_2, the text extraction information S_3_1 gets two votes and the text extraction information S_3_2 gets one vote, so the text extraction information S_3_1 is determined as the target extraction information corresponding to the first preset identifier L_3; then, the target extraction information corresponding to the basic text information is obtained according to the text extraction information S_1_1, the text extraction information S_2_2, and the text extraction information S_3_1.

[0095] Step S209, obtaining target text information according to the target extraction information and logic verification information.

[0096] Step S2010, output target text information.

[0097] In a possible implementation, the target extraction information is logically verified by the logic verification information. If the verification passes, the target text information is obtained. If the verification fails, the process returns to step S206 to recalibrate the initial text information. The implementation of the logic verification process is the same as that of the present application. Figure 2 The implementation method of step S107 in the illustrated embodiment is the same and will not be described in detail here.

[0098] In the steps of this embodiment, basic text information is extracted based on the first preset identifier through N information extraction models, and then the target extraction information corresponding to each first preset identifier is obtained based on the voting results, and then the target extraction information corresponding to each first preset identifier is summarized to obtain the target extraction information corresponding to the basic text information. This realizes determining the target extraction information corresponding to each first preset identifier based on different information extraction models, avoids errors when using one information extraction model to extract basic text information, provides accurate text information for the subsequent logic verification process, and solves the problem of inaccurate text content generated by the solution in the prior art.

[0099] In another possible implementation, Figure 8 for Figure 6 The process diagram of the specific implementation steps of step S209 in the embodiment shown is as follows: Figure 8 As shown, the specific implementation steps of step S209 include:

[0100] Step S2091, determining a target information extraction model from N information extraction models according to the target extraction information.

[0101] Step S2092: input the basic text information and the second preset identifier into the target information extraction model to obtain full extraction information, where the full extraction information represents the text content extracted based on the second preset identifier.

[0102] Exemplarily, the first preset identifier is used to determine a target information extraction model for extracting information from the basic text information. Specifically, for example, text extraction information is extracted from the basic text information based on the first preset identifier, and then the target extraction information is obtained by voting on the text extraction information, and then the target information extraction model is determined according to the target extraction information; exemplary, Figure 7 As shown, for the first preset identifier L_1, its target extraction information is text extraction information S_1_1, which is extracted by information extraction model M_1 and information extraction model M_3; for the first preset identifier L_2, its target extraction information is text extraction information S_2_2, which is extracted by information extraction model M_2 and information extraction model M_3; for the first preset identifier L_3, its target extraction information is text extraction information S_3_1, which is extracted by information extraction model M_1 and information extraction model M_2. Since the target extraction information corresponding to the above three first preset identifiers can be extracted by information extraction model M_1, information extraction model M_1 is determined as the target information extraction model. Further, on the basis of determining the target information extraction model, the target information extraction model performs text extraction on the basic text information according to the second preset identifier to obtain the full amount of extracted information.

[0103] Step S2093, obtaining target text information based on the full extraction information and logic verification information.

[0104] Exemplarily, the full amount of extracted information is logically checked by the logic check information. If the check passes, the target text information is obtained. If the check fails, the process returns to step S206 to recalibrate the initial text information. The implementation method of the logic check process is the same as that of the present application. Figure 2 The implementation method of step S107 in the illustrated embodiment is the same and will not be described in detail here.

[0105] It is understandable that in Figure 8 In the illustrated embodiment, the full amount of extraction information determined based on the second preset identifier is used to determine the target text information; while the target extraction information determined based on the first preset identifier is used to determine the target information extraction model, and is not used to determine the target text information; the number of identifiers of the second preset identifier is greater than the number of identifiers of the first preset identifier.

[0106] In the steps of this embodiment, text extraction is performed on basic text information through a small number of first preset identifiers, and then the target information extraction model is quickly and accurately determined based on the target extraction information, and then the basic text information is fully extracted through the second preset identifier based on the determined target information extraction model to obtain full extraction information, which provides accurate text information for the subsequent logical verification process, and not only solves the problem of inaccurate text content generated by the solution in the prior art, but also improves the speed of extracting text information.

[0107] In this embodiment, the implementation method of steps S201-S206 and step S2010 is the same as that of the present application. Figure 2 The implementation methods of steps S101 to S106 and step S108 in the illustrated embodiment are the same and will not be described in detail herein.

[0108] Fig. 9 A schematic diagram of the structure of an audio-to-text processing device provided in one embodiment of the present application is shown in FIG. Fig. 9 As shown, the audio-to-text processing device 3 provided in this embodiment includes:

[0109] The acquisition module 30 is used to acquire user audio information, user speech speed information and logic verification information; wherein the user audio information represents the user information input by voice when the user handles the business through the telephone voice customer service; the user speech speed information represents the speaking speed of the user when the user information is input by voice; the logic verification information represents the content characteristics of the user information;

[0110] The first processing module 31 is used to segment and / or merge the user audio information according to the user speech rate information and the service feature identifier to obtain the initial text information; wherein the service feature identifier is used to indicate a word with service features;

[0111] The second processing module 32 is used to obtain user region information according to the initial text information; wherein the user region information is used to indicate the user's region; determine general voice feature information according to the user region information; wherein the general voice feature information represents the regional accent features corresponding to the user's region; generate calibration adjustment information according to the general voice feature information and reference voice feature information; wherein the reference voice feature information represents reference voice pronunciation features; and the calibration adjustment information represents calibration adjustment features when calibrating and adjusting the initial text information;

[0112] The second processing module 32 is further used to generate basic text information according to the initial text information and the calibration adjustment information, wherein the calibration adjustment information represents the calibration adjustment characteristics when the calibration adjustment is performed on the initial text information;

[0113] A determination module 33, configured to obtain target text information according to basic text information and logic verification information, wherein the logic verification information represents content characteristics of user information, and the target text information is used to indicate text content that can be used to generate a user request ticket;

[0114] The output module 34 is used to output the target text information.

[0115] In one possible design, when the second processing module 32 generates calibration adjustment information based on general voice feature information and reference voice feature information, it is specifically used to: generate pre-reference information based on pending business information, the pending business information represents the target business features of the business to be handled by the user, and the pre-reference information represents the reference content corresponding to the target business features; generate user voice feature information in response to user verification information uploaded by the user based on the pre-reference information, the user verification information represents the audio content corresponding to the reference content, and the user voice feature information represents the user's personal accent features; generate calibration adjustment information based on the user voice feature information, general voice feature information and reference voice feature information.

[0116] In one possible design, when the determination module 33 obtains the target text information based on the basic text information and the logical verification information, it is specifically used to: obtain user identity information based on the initial text information, the user identity information represents the user identity characteristics of the user, and the user identity characteristics include the user age and the user gender; obtain first verification pass information based on the basic text information and the logical verification information, the first verification pass information represents that the text content of the basic text information matches the content characteristics of the user information; obtain second verification pass information based on the basic text information, the user identity information and the first verification pass information, the second verification pass information represents that the text content of the basic text information matches the user identity characteristics of the user; obtain the target text information based on the basic text information and the second verification pass information.

[0117] In one possible design, when the determination module 33 obtains the target text information based on the basic text information and the logical verification information, it is specifically used to: input the basic text information and the first preset identifier into N information extraction models to obtain N text extraction information, the text extraction information represents the text content extracted based on the first preset identifier, and N is an odd number greater than or equal to 3; determine the target extraction information based on the N text extraction information, the target extraction information is used to indicate the text extraction information with the most votes determined from the N text extraction information based on the voting results; obtain the target text information based on the target extraction information and the logical verification information.

[0118] In one possible design, when the determination module 33 obtains the target text information based on the target extraction information and the logical verification information, it is specifically used to: determine the target information extraction model from N information extraction models based on the target extraction information; input the basic text information and the second preset identifier into the target information extraction model to obtain the full extraction information, and the full extraction information represents the text content extracted based on the second preset identifier; obtain the target text information based on the full extraction information and the logical verification information.

[0119] In one possible design, the determination module 33 is also used to quantify the user voice information and / or the user speech speed information to obtain user emotional information, where the user voice information is used to indicate the volume and / or pitch of the user's voice when inputting user information through voice, and the user emotional information represents the user's attention to handling the business; based on the user emotional information, the preset grading information is matched to determine the target grading identifier, the preset grading information represents the processing priority when processing the user's request, and the target grading identifier is used to indicate the target processing priority corresponding to the user request work order; based on the target grading identifier, the target text information is marked to generate graded text information, and the graded text information is used to indicate text content that can be used to generate a user request work order and has a target processing priority.

[0120] The acquisition module 30, the first processing module 31, the second processing module 32, the determination module 33 and the output module 34 are connected in sequence. The audio-to-text processing device 3 provided in this embodiment can be executed as follows: Figure 2-Figure 8 The technical solutions of any of the method embodiments shown have similar implementation principles and technical effects, which will not be described in detail here.

[0121] The audio-to-text processing device provided in the embodiment of the present application can be used to execute the technical solution of the audio-to-text processing method in the above embodiment. Its implementation principle and technical effect are similar and will not be repeated here.

[0122] It should be noted that it should be understood that the division of the various modules of the above device is only a division of logical functions. In actual implementation, they can be fully or partially integrated into one physical entity, or they can be physically separated. And these modules can all be implemented in the form of software called by processing elements; they can also be all implemented in the form of hardware; some modules can also be implemented in the form of software called by processing elements, and some modules can be implemented in the form of hardware. For example, the first processing module 31 can be a separately established processing element, or it can be integrated in a chip of the above device. In addition, it can also be stored in the memory of the above device in the form of program code, and called and executed by a processing element of the above device. The function of the first processing module 31. The implementation of other modules is similar. In addition, all or part of these modules can be integrated together, or they can be implemented independently. The processing element here can be an integrated circuit with signal processing capabilities. In the implementation process, each step of the above method or each module above can be completed by an integrated logic circuit of hardware in the processor element or instructions in the form of software.

[0123] Fig.10 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. Fig.10 As shown, the electronic device may include: a transceiver 121 , a processor 122 , and a memory 123 .

[0124] The processor 122 executes the computer execution instructions stored in the memory, so that the processor 122 executes the scheme in the above embodiment. The processor 122 can be a general-purpose processor, including a central processing unit CPU, a network processor (network processor, NP), etc.; it can also be a digital signal processor DSP, an application-specific integrated circuit ASIC, a field programmable gate array FPGA or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components.

[0125] The memory 123 is connected to the processor 122 via a system bus and completes communication between them. The memory 123 is used to store computer program instructions.

[0126] The transceiver 121 may be used to obtain tasks to be executed and configuration information of the tasks to be executed.

[0127] The system bus can be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. The system bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus. The transceiver is used to realize the communication between the database access device and other computers (such as clients, read-write libraries, and read-only libraries). The memory may include random access memory (RAM) and may also include non-volatile memory.

[0128] The electronic device provided in the embodiment of the present application may be the terminal device of the above embodiment.

[0129] An embodiment of the present application also provides a chip for running instructions, which is used to execute the technical solution of the audio-to-text processing method in the above embodiment.

[0130] An embodiment of the present application also provides a computer-readable storage medium, in which computer instructions are stored. When the computer instructions are executed on a computer, the computer executes the technical solution of the audio-to-text processing method of the above embodiment.

[0131] An embodiment of the present application also provides a computer program product, which includes a computer program stored in a computer-readable storage medium. At least one processor can read the computer program from the computer-readable storage medium. When at least one processor executes the computer program, the technical solution of the audio-to-text processing method in the above embodiment can be implemented.

[0132] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of modules is only a logical function division, and there may be other division methods in actual implementation, such as multiple modules can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or modules, which can be electrical, mechanical or other forms.

[0133] 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 on multiple network units. Some or all of the modules may be selected according to actual needs to implement the solution of this embodiment.

[0134] In addition, each functional module in each embodiment of the present application can be integrated into one processing unit, or each module can exist physically separately, or two or more modules can be integrated into one unit. The above-mentioned module-composed unit can be implemented in the form of hardware or in the form of hardware plus software functional units.

[0135] The above-mentioned integrated module implemented in the form of a software function module can be stored in a computer-readable storage medium. The above-mentioned software function module is stored in a storage medium, including a number of instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor to perform some steps of the methods of various embodiments of the present application.

[0136] It should be understood that the processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), etc. A general-purpose processor may be a microprocessor or any conventional processor. The steps of the method disclosed in the invention may be directly implemented as being executed by a hardware processor, or may be implemented by a combination of hardware and software modules in the processor.

[0137] The memory may include a high-speed RAM memory, and may also include a non-volatile storage NVM, such as at least one disk memory, and may also be a USB flash drive, a mobile hard disk, a read-only memory, a magnetic disk or an optical disk, etc.

[0138] The bus may be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, the bus in the drawings of the present application is not limited to only one bus or one type of bus.

[0139] The above storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk. The storage medium can be any available medium that can be accessed by a general or special purpose computer.

[0140] An exemplary storage medium is coupled to a processor so that the processor can read information from the storage medium and write information to the storage medium. Of course, the storage medium can also be a component of the processor. The processor and the storage medium can be located in an application specific integrated circuit (ASIC). Of course, the processor and the storage medium can also exist as discrete components in an electronic control unit or a main control device.

[0141] Those skilled in the art can understand that all or part of the steps of implementing the above-mentioned method embodiments can be completed by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, the steps of the above-mentioned method embodiments are executed; and the aforementioned storage medium includes: ROM, RAM, disk or optical disk and other media that can store program codes.

[0142] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. An audio-to-text processing method, characterized in that: include: Obtain user audio information, user speech speed information and logic verification information; The user audio information represents the user information input by voice when the user handles business through telephone voice customer service; The user speech speed information represents the speaking speed of the user when inputting the user information by voice; the logic check information represents the content characteristics of the user information; According to the user speech speed information and the service feature identifier, the user audio information is segmented and / or merged to obtain initial text information; wherein the service feature identifier is used to indicate a word with a service feature; According to the initial text information, user region information is obtained; wherein the user region information is used to indicate the region where the user is located; Determining general voice feature information according to the user region information; wherein the general voice feature information represents regional accent features corresponding to the region where the user is located; Generating calibration adjustment information according to the general speech feature information and the reference speech feature information; wherein the reference speech feature information represents reference speech pronunciation features; and the calibration adjustment information represents calibration adjustment features when calibrating and adjusting the initial text information; Generate basic text information according to the initial text information and the calibration adjustment information; Obtaining target text information according to the basic text information and the logic verification information; the target text information is used to indicate text content that can be used to generate a user request work order; The target text information is output.

2. The method according to claim 1, characterized in that The step of generating calibration adjustment information according to the general speech feature information and the reference speech feature information includes: Generate pre-reference information according to the pending business information, wherein the pending business information represents the target business characteristics of the business to be handled by the user, and the pre-reference information represents the reference content corresponding to the target business characteristics; In response to user verification information uploaded by the user according to the pre-set reference information, generating user voice feature information, wherein the user verification information represents the audio content corresponding to the reference content, and the user voice feature information represents the personal accent feature of the user; The calibration adjustment information is generated according to the user voice feature information, the general voice feature information and the reference voice feature information.

3. The method according to claim 1, characterized in that The step of obtaining target text information according to the basic text information and the logic verification information includes: Obtaining user identity information according to the initial text information, wherein the user identity information represents user identity characteristics of the user, and the user identity characteristics include user age and user gender; Obtaining first verification pass information according to the basic text information and the logic verification information, wherein the first verification pass information indicates that the text content of the basic text information matches the content feature of the user information; Obtaining second verification information according to the basic text information, the user identity information and the first verification information, wherein the second verification information indicates that the text content of the basic text information matches the user identity feature of the user; The target text information is obtained according to the basic text information and the second verification pass information.

4. The method according to claim 1, characterized in that: The step of obtaining target text information according to the basic text information and the logic verification information includes: Inputting the basic text information and the first preset identifier into N information extraction models to obtain N text extraction information, wherein the text extraction information represents the text content extracted based on the first preset identifier, and N is an odd number greater than or equal to 3; Determine target extraction information according to the N text extraction information, wherein the target extraction information is used to indicate the text extraction information with the most votes determined from the N text extraction information based on the voting result; The target text information is obtained according to the target extraction information and the logic verification information.

5. The method according to claim 4, characterized in that The step of obtaining the target text information according to the target extraction information and the logic verification information includes: Determining a target information extraction model from the N information extraction models according to the target extraction information; Inputting the basic text information and the second preset identifier into the target information extraction model to obtain full extraction information, wherein the full extraction information represents the text content extracted based on the second preset identifier; The target text information is obtained according to the full extraction information and the logic verification information.

6. The method according to claim 1, characterized in that The method further comprises: Quantifying the user voice information and / or the user speech speed information to obtain user emotion information, wherein the user voice information is used to indicate the volume and / or pitch of the user's voice when inputting the user information by voice, and the user emotion information represents the user's attention to the handling of the service; Based on the user emotion information matching preset classification information, determining a target classification identifier, the preset classification information represents the processing priority when processing the user demand, and the target classification identifier is used to indicate the target processing priority corresponding to the user demand work order; The target text information is marked based on the target classification identifier to generate classified text information, where the classified text information is used to indicate text content that can be used to generate the user request work order and has the target processing priority.

7. An audio-to-text processing device, characterized in that: include: The acquisition module is used to obtain user audio information, user speech speed information and logic verification information; The user audio information represents the user information input by voice when the user handles business through telephone voice customer service; The user speech speed information represents the speaking speed of the user when inputting the user information by voice; the logic check information represents the content characteristics of the user information; A first processing module is used to segment and / or merge the user audio information according to the user speech rate information and the service feature identifier to obtain initial text information; wherein the service feature identifier is used to indicate a word with a service feature; The second processing module is used to obtain user region information according to the initial text information; wherein the user region information is used to indicate the region where the user is located; determine general voice feature information according to the user region information; wherein the general voice feature information represents the regional accent features corresponding to the region where the user is located; generate calibration adjustment information according to the general voice feature information and reference voice feature information; wherein the reference voice feature information represents reference voice pronunciation features; and the calibration adjustment information represents calibration adjustment features when calibrating and adjusting the initial text information; The second processing module is further used to generate basic text information according to the initial text information and the calibration adjustment information; A determination module, configured to obtain target text information according to the basic text information and the logic verification information, wherein the logic verification information represents content characteristics of the user information, and the target text information is used to indicate text content that can be used to generate a user request work order; An output module is used to output the target text information.

8. An electronic device, characterized in that: include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory to implement the method according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, which are used to implement the method according to any one of claims 1 to 6 when executed by a processor.

10. A computer program product, characterized in that The invention comprises a computer program, which implements the method according to any one of claims 1 to 6 when being executed by a processor.

Images