Voice call quality inspection method, system, device and storage medium

By introducing QMQ message queue components and audio recognition models in the voice call quality inspection in the OTA industry, real-time recognition and transfer of voice call data is realized, the problems of low processing efficiency and high cost are solved, and the quality inspection efficiency and real-time are improved sex.

CN112509582BActive Publication Date: 2024-11-29CTRIP COMP TECH SHANGHAI
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202011325106.0
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2020-11-24
Publication Date
2024-11-29
Estimated Expiration
2040-11-24

AI Technical Summary

Technical Problem

In the prior art, the quality inspection and processing of voice calls in the OTA industry is low, poor real-time and high cost, making it difficult to quickly identify and translate a large number of call recordings, affecting the quality inspection efficiency.

Method used

Adopt the real-time data reception mechanism based on the QMQ message queue component, and process voice call data through automatic speech recognition, generate transcribed text and perform quality inspection. The audio recognition model is used to recognize voice data as transcribed text, and text breaking and smoothing Process to improve quality inspection efficiency.

Benefits of technology

It realizes the real-time and efficiency improvement of voice call quality inspection, reduces the cost of quality inspection, can quickly identify and process a large number of call recordings, and improves the response speed and accuracy of the quality inspection platform.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN112509582B_ABST
    Figure CN112509582B_ABST
Patent Text Reader

Abstract

The present invention discloses a voice call quality inspection method, system, device and storage medium, wherein the voice call quality inspection method comprises: receiving voice call data in real time based on a QMQ mechanism; performing automatic voice recognition processing on the voice call data to obtain corresponding transcribed text; and performing quality inspection processing based on the transcribed text. The present invention ensures the stability of the voice call data acquisition process by introducing an asynchronous and synchronous QMQ mechanism in a voice recognition architecture, obtains the corresponding transcribed text by performing automatic voice recognition processing on the voice call data, and completes the quality inspection processing based on the transcribed text, thereby solving the problems of low voice call quality inspection efficiency, insufficient real-time performance and high cost.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of speech recognition technology, and in particular to a method, system, device and storage medium for quality inspection of voice calls. Background Art

[0002] Currently, the OTA (Online Travel Agency) industry has a large number of manual telephone customer service. In order to determine whether the telephone customer service has any illegal or abnormal speech behavior when making voice calls with the target hotel or guests, such as verbal abuse, ridicule, evasion, excessive promises, etc., a quality inspection system for voice calls is needed to regulate the behavior of customer service personnel and thus improve the quality of customer service. The traditional voice quality inspection system, first, has quality inspectors randomly listen to a certain proportion of telephone recordings for inspection. The number of calls listened to in this way is limited, and the coverage rate of recording sampling is low. At present, each employee listens to about 3 hours of recorded calls a day and records the sampling content, which is inefficient; and the subjectivity of manual listening leads to the problem of inconsistent quality inspection standards, and many risk information is difficult to discover, and the quality inspection cost is relatively high. The other is delayed intelligent voice quality inspection, but the quality inspection platform needs to transcribe the entire call after the call is completed. For up to 500,000 call recordings per day, it is difficult for the voice recognition service to quickly transcribe and forward them to the quality inspection platform, which delays the quality inspection process and affects the efficiency of the quality inspection process. When encountering abnormal situations, the intervention time is late. In summary, both voice quality inspection methods have their own problems, resulting in poor user experience. Summary of the invention

[0003] The technical problem to be solved by the present invention is to overcome the defects of low processing efficiency, poor real-time performance and high cost in the voice call quality inspection in the OTA industry in the prior art, and to provide a voice call quality inspection method, system, device and storage medium that can improve the voice call quality inspection efficiency and reduce the quality inspection cost while ensuring real-time processing.

[0004] The present invention solves the above technical problems through the following technical solutions:

[0005] The present invention provides a voice call quality inspection method, the quality inspection method comprising:

[0006] Receive voice call data in real time based on QMQ (Q-Message Quene, “Qunar” message queue component) mechanism;

[0007] Performing automatic speech recognition processing on the voice call data to obtain corresponding transcribed text;

[0008] A quality inspection process is performed based on the transcribed text.

[0009] Preferably, the step of receiving voice call data in real time based on the QMQ mechanism includes:

[0010] The producer service receives the voice call data in real time, and sends the voice call basic information of the voice call data to the QMQ message middleware;

[0011] The consumer service extracts the basic voice call information from the QMQ message middleware.

[0012] Preferably, the step of performing automatic speech recognition processing on the voice call data to obtain corresponding transcribed text includes:

[0013] Retrieving the corresponding voice call data through the voice call basic information;

[0014] The audio recognition model is used to recognize the voice call data into the corresponding transcribed text.

[0015] Preferably, the step of using the audio recognition model to identify the voice call data into the corresponding transcribed text includes:

[0016] Splitting the voice call data into a corresponding number of single-channel data according to the number of channels included;

[0017] Cutting the single-channel data according to the detected silence points to obtain single-slice data of voice calls;

[0018] Extracting audio features from the voice call single-piece data using a feature extraction model;

[0019] The audio features are input into the voiceprint model for processing and then the corresponding transcribed text is transcribed.

[0020] Preferably, the step of performing automatic speech recognition processing on the voice call data and the step of performing quality inspection processing based on the transcribed text further include:

[0021] The transcribed text is subjected to text segmentation processing and / or text smoothing processing.

[0022] Preferably, the basic voice call information includes a voice identifier.

[0023] Preferably, the audio features at least include mfcc (Mel Frequency Cepstrum Coefficient) features and pitch (an audio feature reflecting pitch in acoustics) features.

[0024] Preferably, before the step of splitting the voice call data into a corresponding number of single-channel data according to the number of included channels, the step further includes:

[0025] The voice call data is decompressed, and the decompressed voice call data at least includes a PCM (Pulse Code Modulation, an audio format that can be post-processed) format.

[0026] Preferably, when the voice call data is dual-channel, the voice call data is split into first single-channel data and second single-channel data; the first single-channel data is used to represent the user recording, and the second single-channel data is used to represent the customer service recording.

[0027] The present invention also provides a voice call quality inspection system, the quality inspection system comprising:

[0028] The receiving module is used to receive voice call data in real time based on the QMQ mechanism;

[0029] A recognition module, used for performing automatic speech recognition processing on the voice call data to obtain corresponding transcribed text;

[0030] A quality inspection module is used to perform quality inspection based on the transcribed text.

[0031] Preferably, the receiving module comprises:

[0032] A producer service unit, configured to receive the voice call data in real time by using a producer service, and send voice call basic information of the voice call data to a QMQ message middleware;

[0033] The consumer service unit is used to extract the basic voice call information from the QMQ message middleware using consumer services.

[0034] Preferably, the identification module comprises:

[0035] A call data retrieving unit, configured to retrieve the corresponding voice call data through the voice call basic information;

[0036] A call data recognition unit is used to use an audio recognition model to recognize the voice call data into the corresponding transcribed text.

[0037] Preferably, the call data identification unit includes:

[0038] A channel splitting subunit, used to split the voice call data into a corresponding number of single-channel data according to the number of included channels;

[0039] An audio cutting subunit, used for cutting the single channel data according to the detected silence points to obtain single-slice data of voice calls;

[0040] A feature extraction subunit, configured to extract audio features from the voice call single-piece data using a feature extraction model;

[0041] The text transcription subunit is used to input the audio features into the voiceprint model for processing and then transcribe the corresponding transcribed text.

[0042] Preferably, the quality inspection system further comprises a text processing module for performing text segmentation processing and / or text smoothing processing on the transcribed text.

[0043] Preferably, the basic voice call information includes a voice identifier.

[0044] Preferably, the audio features at least include mfcc features and pitch features.

[0045] Preferably, the call data identification unit further includes:

[0046] The decompression subunit is used to decompress the voice call data, and the decompressed voice call data at least includes a PCM format.

[0047] Preferably, when the voice call data is dual-channel, the channel splitting subunit is used to split the voice call data into first single-channel data and second single-channel data; the first single-channel data is used to represent the user recording, and the second single-channel data is used to represent the customer service recording.

[0048] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned voice call quality inspection method when executing the computer program.

[0049] The present invention also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the above-mentioned voice call quality inspection method are implemented.

[0050] The positive progressive effect of the present invention is that the present invention provides a voice call quality inspection method, system, device and storage medium. Compared with the prior art, the voice call quality inspection method provided by the present invention introduces an asynchronous and synchronous QMQ mechanism in the voice recognition architecture to ensure the stability of the voice call data acquisition process, and obtains the corresponding transcribed text by performing automatic voice recognition processing on the voice call data, and completes the quality inspection processing based on the transcribed text, thereby solving the problems of low efficiency, insufficient real-time performance and high cost of voice call quality inspection. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 This is a flow chart of a voice call quality inspection method according to Embodiment 1 of the present invention.

[0052] Figure 2 This is a flow chart of step S1 in the voice call quality inspection method of embodiment 1 of the present invention.

[0053] Figure 3 This is a flow chart of step S2 in the voice call quality inspection method of embodiment 1 of the present invention.

[0054] Figure 4 This is a flow chart of step S22 in the voice call quality inspection method of embodiment 1 of the present invention.

[0055] Figure 5 This is an overall schematic diagram of the modules of the voice call quality inspection system according to Example 2 of the present invention.

[0056] Figure 6 This is a structural block diagram of an electronic device according to Embodiment 3 of the present invention. DETAILED DESCRIPTION

[0057] The present invention is further described below by way of examples, but the present invention is not limited to the scope of the examples.

[0058] Example 1

[0059] See also Figure 1 This embodiment provides a method for quality inspection of voice calls, which specifically includes the following steps:

[0060] Step S1. Receive voice call data in real time based on the QMQ mechanism.

[0061] Step S2: Perform automatic speech recognition processing on the voice call data to obtain corresponding transcribed text.

[0062] Through step S2, a transcribed text corresponding to the input voice call audio data will be obtained.

[0063] Step S3: performing text segmentation processing and / or text smoothing processing on the transcribed text.

[0064] In step S3, since the transcription result is a borderless text, that is, a text without punctuation format and all Chinese characters excluding numbers, it is necessary to first add boundary processing to form a normal sentence segmentation text content before subsequent processing. The specific implementation method is: the segmentation processing sends the transcribed text into the segmentation model, and automatically configures punctuation marks according to the text to improve the readability and quality inspection efficiency of the text; the smoothing processing inputs the segmentation result into the smoothing model to convert the quantifiers in the text, such as: "March 20th" to "March 20th", which can further improve the efficiency of quality inspection.

[0065] Step S4: Perform quality inspection based on the transcribed text.

[0066] This embodiment is implemented based on the overall architecture of the QMQ mechanism; wherein step S2 represents the detailed process of speech recognition, and step S3 represents the post-processing process of the transcribed text, which mainly includes "sentence segmentation" and "smoothing" of the obtained transcribed text.

[0067] Step S4 determines whether there are any non-compliant situations in the voice call based on the processed formatted transcribed text. For example, the processed transcribed text is compared with the speech template and the illegal keywords to find the illegal calls and locate the roles and timing, or make a secondary correction and judgment, and report the problematic calls to the downstream system to complete the quality inspection process.

[0068] See also Figure 2 In this embodiment, step S1 specifically includes:

[0069] Step S11. The producer service receives the voice call data in real time, and sends the voice call basic information of the voice call data to the QMQ message middleware.

[0070] Step S12: The consumer service extracts the basic voice call information from the QMQ message middleware.

[0071] Specifically, the QMQ mechanism is an open source framework consisting of producer services and consumer services. The producer service is used to receive a large amount of basic voice call information sent by the upstream. The main purpose of designing the producer service is to buffer the pressure of the service interface and prevent the service from crashing due to a surge in access volume in a certain period of time, thereby temporarily relieving service pressure and load balancing, ensuring that the energy consumption of the processing terminals is similar and avoiding excessive differences in energy consumption between terminals.

[0072] See also Figure 3 In a preferred embodiment, step S2 specifically includes:

[0073] Step S21. Retrieve corresponding voice call data through voice call basic information.

[0074] Specifically, the basic information of voice calls includes voice identifiers, service module identifiers, timing identifiers, etc. Audio data is pulled through the ceph (distributed file system) data center interface. Since the pulled audio data is usually stored in the form of compressed files due to space requirements, it is necessary to decompress it into processable PCM format audio data, for example, FFMpeg (an open source audio format processing software) can be used to achieve this.

[0075] Step S22: Use the audio recognition model to recognize the voice call data into corresponding transcribed text.

[0076] Specifically, the consumer service connects the audio recognition model, including voice call information reception, audio retrieval, data decompression, audio channel separation, audio cutting, audio feature extraction and audio decoding modules, until it extends to the subsequent text segmentation or text smoothing processing module, to identify the voice call data into the corresponding transcribed text.

[0077] See also Figure 4 As a preferred implementation, step S22 includes:

[0078] Step S221: Split the voice call data into a corresponding number of single-channel data according to the number of included channels.

[0079] Step S222: Cut the single-channel data according to the detected silence points to obtain single-slice data of the voice call.

[0080] Step S223: Use the feature extraction model to extract audio features from the voice call single piece of data.

[0081] Step S224: Input the audio features into the voiceprint model for processing and transcribe the corresponding transcribed text.

[0082] Specifically, in step 221, since the recording file is multi-channel, it needs to be separated into single channels for separate processing. For example, when the call data is dual-channel, it is split into first single-channel data and second single-channel data according to the cross-storage method of the audio to respectively store the recordings of the user and the customer service staff, and relevant matching is performed through the timeline.

[0083] For step 222, the entire audio is cut using silence point detection, that is, audio with a longer duration needs to be cut into small segments for prediction. Specifically, the silence points of each speech interval are dynamically detected through an energy threshold. The silence point detection statistics the energy mean of a segment of audio (for example, the first 50 frames) as the cutting threshold for the recording, and then the cutting point is found by frame detection to cut the entire audio.

[0084] For step 223, it is necessary to extract common audio features such as mfcc features, pitch features, iverse (an audio vector that characterizes speaker differences) features, and concatenate them. At this time, for step 224, the concatenated audio features are input into the voiceprint model for processing and then transcribed into the corresponding transcribed text.

[0085] The voice call quality inspection method provided in this embodiment ensures the stability of the voice call data acquisition process by introducing an asynchronous and synchronous QMQ mechanism in the voice recognition architecture, obtains the corresponding transcribed text by performing automatic voice recognition processing on the voice call data, and realizes lower-latency transcription by connecting various main voice recognition modules in series. The quality inspection processing is completed based on the transcribed text, thereby improving the real-time transcription rate. The quality inspection results can be delivered to subsequent processes for use as soon as possible, so that the quality inspection platform personnel can quickly discover problems and prevent accidents. It can play a very important role in large-scale call centers.

[0086] Example 2

[0087] See also Figure 5 This embodiment provides a voice call quality inspection system, which specifically includes a receiving module 1, a recognition module 2, a text processing module 3 and a quality inspection module 4. Among them, the receiving module 1 includes a producer service unit 5 and a consumer service unit 6; the recognition module 2 includes a call data retrieval unit 7 and a call data recognition unit 8; the call data recognition unit 8 is divided into four subunits according to the function of audio recognition: a channel splitting subunit 801, an audio cutting subunit 802, a feature extraction subunit 803, and a text transcription subunit 804. The following is a specific functional description and explanation of the above modules or units in the quality inspection system.

[0088] The quality inspection system of this embodiment is implemented based on the overall architecture of the QMQ mechanism, wherein the receiving module 1 is used to receive voice call data in real time based on the QMQ mechanism; the recognition module 2 is used to process the specific process of voice recognition, that is, to perform automatic voice recognition processing on the voice call data to obtain the corresponding formatted transcribed text.

[0089] The text processing module 3 is used to perform text segmentation processing and / or text smoothing processing on the transcribed text. Since the transcription result is a borderless text, that is, a text without punctuation format and all Chinese characters do not include numbers, the text processing module 3 first needs to add boundary processing to form a normal sentence segmentation text content before subsequent processing. Specifically, the segmentation processing sends the transcribed text into the segmentation model, and automatically configures punctuation marks according to the text to improve the readability and quality inspection efficiency of the text; the smoothing processing inputs the segmentation result into the smoothing model to convert the quantifiers in the text, such as: "March 20th" to "March 20th", which can also improve the efficiency of quality inspection.

[0090] The quality inspection module 4 determines whether there are any non-compliant situations in the voice call based on the processed formatted transcribed text. For example, the processed transcribed text is compared with the speech template and the illegal keywords to find the illegal calls and locate the roles and timing, or make a secondary correction and judgment, and report the problematic calls to the downstream system.

[0091] The QMQ mechanism adopted in this embodiment is an open source framework consisting of producer services and consumer services. Among them, the producer service is used to receive a large amount of basic voice call information sent by the upstream. The main purpose of designing the producer service is to buffer the pressure of the service interface and prevent the service from crashing due to a surge in the number of visits in a certain period of time, thereby playing a role in temporarily relieving service pressure and load balancing, ensuring that the energy consumption of the processing terminals is similar and avoiding excessive differences in energy consumption between terminals.

[0092] In this embodiment, a producer service unit 5 is provided for receiving the voice call data in real time by using a producer service, and sending the voice call basic information of the voice call data to the QMQ message middleware; a consumer service unit 6 is provided for extracting the voice call basic information from the QMQ message middleware by using a consumer service. Through the mutual communication between the producer service unit 5 and the consumer service unit 6, the stability of the service is increased and the transcription efficiency is improved.

[0093] As a preferred embodiment, the identification module 2 includes a call data retrieval unit 7 and a call data identification unit 8. The call data retrieval unit 7 is used to retrieve the corresponding voice call data through the basic information of the voice call, for example, by reading the voice identifier, the service module identifier, the timing identifier, etc., to pull the audio data from the ceph data center interface. Since the pulled audio data is usually stored in the form of a compressed file due to space requirements, the call data retrieval unit 7 needs to decompress it into audio data in a processable PCM format, for example, it can be implemented using FFMpeg.

[0094] The call data recognition unit 8 is used to use the audio recognition model to recognize the voice call data into the corresponding transcribed text. Specifically, the call data recognition unit 8 uses the consumer service in series with the audio recognition model, including voice call information reception, audio retrieval, data decompression, audio channel separation, audio cutting, audio feature extraction and audio decoding modules to recognize the voice call data into the corresponding formatted transcribed text.

[0095] In this embodiment, the call data identification unit 8 includes the following four subunits:

[0096] The channel splitting subunit 801 is used to split the voice call data into a corresponding number of single-channel data according to the number of included channels.

[0097] The audio cutting subunit 802 is used to cut the single channel data according to the detected silence points to obtain the single-slice data of the voice call.

[0098] The feature extraction subunit 803 is used to extract audio features from the voice call single-piece data using a feature extraction model.

[0099] The text transcription subunit 804 is used to input the audio features into the voiceprint model for processing and then transcribe the corresponding transcribed text.

[0100] Specifically, the recording file obtained by the channel splitting sub-unit 801 is multi-channel and needs to be separated into single channels for separate processing. For example, when the call data is dual-channel, it is split into first single-channel data and second single-channel data according to the cross-storage method of the audio to respectively store the recordings of the user and the customer service staff, and relevant matching is performed through the timeline.

[0101] The audio cutting subunit 802 uses silence point detection to cut the entire audio, that is, for audio with a longer duration, it needs to be cut into small audio segments for prediction. Specifically, the silence point of each speech interval is dynamically detected through the energy threshold. The silence point detection statistics the energy mean of a segment of audio (for example, the first 50 frames) as the cutting threshold of the recording, and then detects frame by frame to find the cutting point to cut the entire audio.

[0102] The feature extraction subunit 803 extracts commonly used audio features such as mfcc features, pitch features, and iverter features, and concatenates them.

[0103] The voice call quality inspection system provided in this embodiment ensures the stability of the voice call data acquisition process by introducing an asynchronous and synchronous QMQ mechanism in the voice recognition architecture, obtains the corresponding transcribed text by performing automatic voice recognition processing on the voice call data, and completes the quality inspection processing based on the transcribed text, thereby solving the problems of low efficiency, lack of real-time performance and high cost of voice call quality inspection.

[0104] Example 3

[0105] Figure 6 This is a schematic diagram of the structure of an electronic device provided in Embodiment 3 of the present invention. The electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the voice call quality inspection method in Embodiment 1 is implemented. Figure 6 The electronic device 30 shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0106] like Figure 6 As shown, the electronic device 30 may be in the form of a general-purpose computing device, for example, it may be a server device. The components of the electronic device 30 may include, but are not limited to: at least one processor 31, at least one memory 32, and a bus 33 connecting different system components (including the memory 32 and the processor 31).

[0107] The bus 33 includes a data bus, an address bus, and a control bus.

[0108] The memory 32 may include a volatile memory, such as a random access memory (RAM) 321 and / or a cache memory 322 , and may further include a read-only memory (ROM) 323 .

[0109] The memory 32 may also include a program / utility 325 having a set (at least one) of program modules 324, such program modules 324 including but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment.

[0110] The processor 31 executes various functional applications and data processing by running the computer programs stored in the memory 32, such as the voice call quality inspection method in Embodiment 1 of the present invention.

[0111] The electronic device 30 may also communicate with one or more external devices 34 (e.g., keyboards, pointing devices, etc.). Such communication may be performed via an input / output (I / O) interface 35. Furthermore, the model generating device 30 may also communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) via a network adapter 36. Figure 6 As shown, the network adapter 36 communicates with other modules of the model-generated device 30 via the bus 33. It should be understood that, although not shown in the figure, other hardware and / or software modules can be used in conjunction with the model-generated device 30, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID (disk array) systems, tape drives, and data backup storage systems, etc.

[0112] It should be noted that although several units / modules or sub-units / modules of the electronic device are mentioned in the above detailed description, this division is merely exemplary and not mandatory. In fact, according to an embodiment of the present invention, the features and functions of two or more units / modules described above can be embodied in one unit / module. Conversely, the features and functions of one unit / module described above can be further divided into multiple units / modules to be embodied.

[0113] Example 4

[0114] This embodiment provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the steps in the voice call quality inspection method in Embodiment 1 are implemented.

[0115] The readable storage medium may include but is not limited to: a portable disk, a hard disk, a random access memory, a read-only memory, an erasable programmable read-only memory, an optical storage device, a magnetic storage device or any suitable combination of the above.

[0116] In a possible implementation manner, the present invention can also be implemented in the form of a program product, which includes program code. When the program product runs on a terminal device, the program code is used to enable the terminal device to execute the steps in the voice call quality inspection method in Example 1.

[0117] Among them, the program code for executing the present invention can be written in any combination of one or more programming languages, and the program code can be executed completely on the user device, partially on the user device, as an independent software package, partially on the user device and partially on a remote device, or completely on the remote device.

[0118] Although the specific embodiments of the present invention are described above, it should be understood by those skilled in the art that this is only for illustration and the protection scope of the present invention is defined by the appended claims. Those skilled in the art may make various changes or modifications to these embodiments without departing from the principles and essence of the present invention, but these changes and modifications all fall within the protection scope of the present invention.

Claims

1. A method for quality inspection of voice calls, characterized in that: The quality inspection method comprises: Receive voice call data in real time based on QMQ mechanism; Performing automatic speech recognition processing on the voice call data to obtain corresponding transcribed text; Performing quality inspection based on the transcribed text; The step of performing quality inspection based on the transcribed text comprises: Performing data comparison on the transcribed text according to the speech template and the illegal keywords, obtaining illegal call data and locating the roles and timing, or performing secondary correction and determination, and outputting the illegal call data to complete the quality inspection process; The step of receiving voice call data in real time based on the QMQ mechanism includes: The producer service receives the voice call data in real time, and sends the voice call basic information of the voice call data to the QMQ message middleware; The consumer service extracts the basic voice call information from the QMQ message middleware; The step of performing automatic speech recognition processing on the voice call data to obtain corresponding transcribed text includes: Retrieve the corresponding voice call data from the ceph data center interface through the voice call basic information; The voice call data includes a voice identifier, a service module identifier and a timing identifier; The audio recognition model is used to recognize the voice call data into the corresponding transcribed text.

2. The voice call quality inspection method according to claim 1, characterized in that: The step of using the audio recognition model to identify the voice call data into the corresponding transcribed text includes: Splitting the voice call data into a corresponding number of single-channel data according to the number of channels included; Cutting the single-channel data according to the detected silence points to obtain single-slice data of voice calls; Extracting audio features from the voice call single-piece data using a feature extraction model; The audio features are input into the voiceprint model for processing and then the corresponding transcribed text is transcribed.

3. The voice call quality inspection method according to claim 1, characterized in that: The step of automatically performing speech recognition processing on the voice call data and the step of performing quality inspection processing based on the transcribed text also include: The transcribed text is subjected to text segmentation processing and / or text smoothing processing.

4. A voice call quality inspection system, characterized in that: The quality inspection system comprises: The receiving module is used to receive voice call data in real time based on the QMQ mechanism; A recognition module, used for performing automatic speech recognition processing on the voice call data to obtain corresponding transcribed text; A quality inspection module, used for performing quality inspection based on the transcribed text; The quality inspection module is also used to compare the transcribed text with the speech template and the illegal keywords, obtain the illegal call data and locate the role and time sequence, or perform secondary correction and determination, and output the illegal call data; The receiving module comprises: A producer service unit, configured to receive the voice call data in real time by using a producer service, and send voice call basic information of the voice call data to a QMQ message middleware; A consumer service unit, configured to extract the voice call basic information from the QMQ message middleware using a consumer service; The identification module comprises: A call data retrieving unit, configured to retrieve the corresponding voice call data from a ceph data center interface through the voice call basic information; The voice call data includes a voice identifier, a service module identifier and a timing identifier; A call data recognition unit is used to use an audio recognition model to recognize the voice call data into the corresponding transcribed text.

5. The voice call quality inspection system according to claim 4, characterized in that: The call data identification unit comprises: A channel splitting subunit, used to split the voice call data into a corresponding number of single-channel data according to the number of included channels; An audio cutting subunit, used for cutting the single channel data according to the detected silence points to obtain single-slice data of voice calls; A feature extraction subunit, configured to extract audio features from the voice call single-piece data using a feature extraction model; The text transcription subunit is used to input the audio features into the voiceprint model for processing and then transcribe the corresponding transcribed text.

6. The voice call quality inspection system as claimed in claim 4, characterized in that: The quality inspection system also includes: The text processing module is used to perform text segmentation processing and / or text smoothing processing on the transcribed text.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the voice call quality inspection method according to any one of claims 1 to 3 is implemented.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the voice call quality inspection method according to any one of claims 1 to 3 are implemented.

Citation Information

Patent Citations

  • Data quality inspection method and device for recorded data and storage medium

    CN110598008A

  • Dialogue type voice recognition method and system, electronic equipment and storage medium

    CN111508498A