A method and system for recording quality inspection

Abstract

The embodiments of this specification mainly relate to the field of audio recording quality inspection technology, specifically a method and system for audio recording quality inspection. A method for audio recording quality inspection, comprising: step S2. receiving audio recordings to be inspected and quality inspection rules uploaded by users, and creating a quality inspection task bar for the received audio recordings to be inspected and quality inspection rules, the quality inspection task bar including the task name and task status, and setting the task status of the quality inspection task bar to the first task status; step S4. searching for the quality inspection task bar with the task status being the first task status, and converting the audio recording to be inspected corresponding to the found quality inspection task bar into a text to be inspected, and setting the task status of the quality inspection task bar to the second task status. The audio recording quality inspection method and system of the embodiments of this specification can help call centers conduct audio recording quality inspection. When there is a demand for audio recording quality inspection, the call center only needs to upload the audio recording to be inspected and the quality inspection rules to the audio recording quality inspection system, and the audio recording quality inspection system can comprehensively feedback the audio recording quality inspection results.

Description

Technical Field

[0001] The embodiments of this specification mainly relate to the field of recording quality inspection technology, and specifically to a recording quality inspection method and system. Background Art

[0002] A call center is a centralized location or system that manages and processes large numbers of telephone calls. It's set up to provide customer service, technical support, sales, and marketing services. A call center typically consists of a team of customer service representatives or customer support representatives who answer calls from customers and provide assistance and answers.

[0003] In order to monitor and evaluate the quality of telephone conversations with customer service representatives or customer support representatives, it is necessary to conduct quality control of call center call recordings. This can help improve service quality, ensure regulatory compliance, resolve complaints and disputes, and provide performance evaluation and reward and punishment mechanisms.

[0004] In the existing recording quality control method, call center quality inspectors use auxiliary equipment such as voice detection servers to check the recordings. However, due to the relatively low quality control efficiency of quality inspectors, when call centers have a large number of recordings to be checked, quality inspectors can only randomly or purposefully select a portion of the recordings to be checked, making it difficult for quality inspectors to complete the quality control of all the recordings. Summary of the Invention

[0005] The embodiments of this specification address the problems existing in the prior art and propose a recording quality inspection method and system. The technical solutions are as follows:

[0006] In a first aspect, an embodiment of this specification provides a recording quality inspection method, comprising:

[0007] Step S2. Receive the recordings and quality inspection rules uploaded by the user for quality inspection, and create a quality inspection task bar for the received recordings and quality inspection rules for quality inspection. The quality inspection task bar includes the task name and task status, and the task status of the quality inspection task bar is set to the first task status;

[0008] Step S4. Find the quality inspection task bar with the first task status, convert the quality inspection recording corresponding to the found quality inspection task bar into the quality inspection text, and set the task status of the quality inspection task bar to the second task status;

[0009] Step S6. Find a quality inspection task bar with a task status of the second task status, perform quality inspection on the to-be-inspected text corresponding to the found quality inspection task bar according to the quality inspection rules to obtain a quality inspection result, and set the task status of the quality inspection task bar to the third task status;

[0010] Step S8. Search for a quality inspection task bar whose task status is the third task status, send the quality inspection result corresponding to the found quality inspection task bar to the user, and set the task status of the quality inspection task bar to the fourth task status.

[0011] Preferably, step S2 specifically includes:

[0012] Step S21. Receive a signal to start uploading the recording to be inspected;

[0013] Step S22. Create a quality inspection folder to store the recording data package and quality inspection rules uploaded by the user, and a quality inspection folder stores a recording data package and the quality inspection rules corresponding to the recording data package. A recording data includes multiple recordings to be quality inspected, and the quality inspection rules are stored through the quality inspection rule text;

[0014] Step S23: receiving a signal indicating that uploading of the recording to be inspected has ended;

[0015] Step S24. Create a quality inspection task bar according to the quality inspection folder. One quality inspection task bar is created for each quality inspection folder. The task bar name of the quality inspection task bar is the same as the folder name of the quality inspection folder.

[0016] Step S25: Set the task status of the created quality inspection task bar to the first task status.

[0017] Preferably, step S4 specifically includes:

[0018] Step S41. Find a quality inspection task bar whose task status is the first task status;

[0019] Step S42. Determine the number of times the quality inspection task bar is found in the first task state. If the number is 1, step S43 is triggered and the process returns to step S41. Otherwise, the process returns to step S41.

[0020] Step S43. Obtain a quality inspection folder whose folder name is the same as the task bar name of the quality inspection task bar;

[0021] Step S44. Convert the recordings to be inspected in the inspection folder into texts to be inspected that correspond one to one with the recordings to be inspected;

[0022] Step S45: Set the task status of the quality inspection task bar to the second task status.

[0023] Preferably, step S44 specifically includes:

[0024] S441. Start the recording to be inspected, and identify the first and second voiceprints through the recording to be inspected;

[0025] S442. Restart the recording to be inspected, convert the recording to be inspected into text, and when the recording to be inspected changes from the first voiceprint to the second voiceprint or from the second voiceprint to the first voiceprint, wrap the text;

[0026] S443. Input a line of text into the machine learning model to obtain a text phrase corresponding to the line of text; repeat step S443 until all lines of text have corresponding text phrases;

[0027] S444. Store all text phrases in the text to obtain the text to be quality checked.

[0028] Preferably, step S6 includes:

[0029] Step S61: Find a quality inspection task bar whose task status is the second task status;

[0030] Step S62. Determine the number of times the quality inspection task bar is found in the second task state. If the number is 1, step S63 is triggered and the process returns to step S61. Otherwise, the process returns to step S61.

[0031] Step S63. Obtain a quality inspection folder whose folder name is the same as the task bar name of the quality inspection task bar;

[0032] Step S64: The documents to be inspected in the quality inspection folder are inspected in turn according to the quality inspection rules to obtain quality inspection results corresponding to the documents to be inspected, and all quality inspection results are stored in the same quality inspection result document;

[0033] Step S65. Set the task status of the quality inspection task bar to the third task status;

[0034] Step S64 specifically includes:

[0035] Step S641: Obtain a text to be inspected, and obtain a phrase to be inspected in the text to be inspected;

[0036] Step S642: Obtain a quality inspection rule from the quality inspection rule text, where the quality inspection rule includes a quality inspection phrase and a quality inspection tag;

[0037] Step S643. Match the phrase to be inspected with the inspection phrases in the inspection rules. If the phrase to be inspected matches at least one inspection phrase, the text to be inspected is determined to be abnormal and the inspection label of the inspection rule is assigned to the text to be inspected. Otherwise, the text to be inspected is determined to be normal.

[0038] Step S644. Determine whether there is any quality inspection rule in the quality inspection rule text that does not match the text to be inspected. If there is any quality inspection rule that does not match, return to step S642; otherwise, proceed to step S645.

[0039] Step S645. When the text to be inspected is determined to be normal each time, the quality inspection result of the text to be inspected is determined to be normal; when the text to be inspected is determined to be abnormal at least once, the quality inspection result of the text to be inspected is determined to be abnormal, and a quality inspection label is assigned to the quality inspection result.

[0040] In a second aspect, an embodiment of this specification provides a recording quality inspection system, including:

[0041] A quality inspection task bar creation module is used to receive the recordings to be inspected and the quality inspection rules uploaded by the user, and create a quality inspection task bar for the received recordings to be inspected and the quality inspection rules. The quality inspection task bar includes a task name and a task status, and the task status of the quality inspection task bar is set to the first task status;

[0042] The to-be-inspected text conversion module is used to search for a quality inspection task bar with a task status of the first task status, convert the to-be-inspected recording corresponding to the found quality inspection task bar into the to-be-inspected text, and set the task status of the quality inspection task bar to the second task status;

[0043] The quality inspection result acquisition module is used to search for a quality inspection task bar with a task status of the second task status, and perform quality inspection on the to-be-inspected text corresponding to the found quality inspection task bar according to the quality inspection rules to obtain a quality inspection result, and set the task status of the quality inspection task bar to the third task status;

[0044] The quality inspection result sending module is used to find the quality inspection task bar with the task status of the third task status, send the quality inspection result corresponding to the found quality inspection task bar to the user, and set the task status of the quality inspection task bar to the fourth task status.

[0045] Preferably, the quality inspection task bar creation module includes:

[0046] A recording upload start signal receiving unit, used to receive a recording upload start signal to be inspected;

[0047] The recording and rule storage unit is used to create a quality inspection folder to store the recording data package and quality inspection rules uploaded by the user. A quality inspection folder stores a recording data package and the quality inspection rules corresponding to the recording data package. A recording data includes multiple recordings to be quality inspected, and the quality inspection rules are stored through the quality inspection rule text;

[0048] A recording upload completion signal receiving unit, used to receive a signal indicating the completion of uploading the recording to be inspected;

[0049] A quality inspection task bar creation unit is used to create a quality inspection task bar based on a quality inspection folder. One quality inspection task bar is created for each quality inspection folder, and the task bar name of the quality inspection task bar is the same as the folder name of the quality inspection folder.

[0050] The first task state setting unit is used to set the task state of the created quality inspection task bar to the first task state.

[0051] Preferably, the module for converting text to be inspected includes:

[0052] A first task bar searching unit, configured to search for a quality inspection task bar whose task status is the first task status;

[0053] A first task bar search times determination unit, configured to determine the number of times the quality inspection task bar has been found when the task state is the first task state;

[0054] The first quality inspection folder obtaining unit is used to obtain a quality inspection folder whose folder name is the same as the task bar name of the quality inspection task bar;

[0055] A text conversion unit, used to convert the recordings to be inspected in the quality inspection folder into texts to be inspected corresponding to the recordings to be inspected;

[0056] The second task state setting unit is used to set the task state of the quality inspection task bar to the second task state.

[0057] Preferably, the text conversion unit includes:

[0058] The voiceprint recognition subunit is used to start the recording to be inspected and identify the first voiceprint and the second voiceprint through the recording to be inspected;

[0059] The text acquisition subunit is used to restart the recording to be inspected, convert the recording to be inspected into text, and wrap the text when the recording to be inspected changes from the first voiceprint to the second voiceprint or from the second voiceprint to the first voiceprint;

[0060] A text phrase acquisition subunit, configured to input a line of text into a machine learning model to obtain a text phrase corresponding to the line of text;

[0061] The text to be inspected forms a subunit, which is used to store all text phrases in the text to obtain the text to be inspected.

[0062] Preferably, the quality inspection result acquisition module includes:

[0063] A second task bar searching unit, configured to search for a quality inspection task bar whose task status is the second task status;

[0064] A second task bar search times determination unit, used to determine the number of times the quality inspection task bar is found when the task state is the second task state;

[0065] The second quality inspection folder obtaining unit is used to obtain a quality inspection folder whose folder name is the same as the task bar name of the quality inspection task bar;

[0066] The quality inspection result acquisition unit is used to perform quality inspection on the documents to be inspected in the quality inspection folder in turn according to the quality inspection rules to obtain quality inspection results corresponding to the documents to be inspected, and all quality inspection results are stored in the same quality inspection result text;

[0067] A third task state setting unit, configured to set the task state of the quality inspection task bar to a third task state;

[0068] The quality inspection result acquisition unit includes:

[0069] The phrase to be inspected obtaining subunit is used to obtain a text to be inspected and obtain the phrase to be inspected in the text to be inspected;

[0070] The quality inspection rule acquisition subunit is used to obtain a quality inspection rule in the quality inspection rule text. The quality inspection rule includes a quality inspection phrase and a quality inspection tag;

[0071] The subunit for determining the text to be inspected is used to match the phrase to be inspected with the quality inspection phrases in the quality inspection rules. When the phrase to be inspected matches at least one quality inspection phrase, the text to be inspected is determined to be abnormal and the quality inspection label of the quality inspection rule is assigned to the text to be inspected. Otherwise, the text to be inspected is determined to be normal.

[0072] The quality inspection rule text determination subunit is used to determine whether there is a quality inspection rule in the quality inspection rule text that does not match the text to be quality inspected;

[0073] The quality inspection result determination subunit is used to determine the quality inspection result of the text to be quality inspected as normal when the text to be quality inspected is determined to be normal each time; when the text to be quality inspected is determined to be abnormal at least once, the quality inspection result of the text to be quality inspected is determined to be abnormal, and a quality inspection label is assigned to the quality inspection result of the text to be quality inspected.

[0074] Beneficial effects

[0075] The recording quality inspection method and system of the embodiments of this specification can help call centers perform recording quality inspection. When there is a need for recording quality inspection, the call center only needs to upload the recording to be inspected and the quality inspection rules to the recording quality inspection system. After receiving the uploaded recording to be inspected and the quality inspection rules, the recording quality inspection system can create corresponding quality inspection folders and quality inspection task bars for the recording to be inspected and the quality inspection rules, and first convert the recording to be inspected into a text to be inspected through the quality inspection folder and the quality inspection task bar, and then perform quality inspection on the text to be inspected to obtain the quality inspection results, and finally feedback the quality inspection results to the call center. No matter how many recordings to be inspected the call center has, the recording quality inspection system can comprehensively and relatively quickly feedback the recording quality inspection results. BRIEF DESCRIPTION OF THE DRAWINGS

[0076] Figure 1A flowchart of a recording quality inspection method provided in an embodiment of this specification;

[0077] Figure 2 This is a structural diagram of the recording quality inspection system provided in the embodiments of this specification. DETAILED DESCRIPTION

[0078] Hereinafter, embodiments of the present specification will be described in more detail with reference to the accompanying drawings.

[0079] Example 1:

[0080] A recording quality inspection method, such as Figure 1 As shown, including:

[0081] Step S2. Receive the recordings and quality inspection rules uploaded by the user, and create a quality inspection task bar for the received recordings and quality inspection rules. The quality inspection task bar includes a task name and a task status, and the task status of the quality inspection task bar is set to the first task status.

[0082] The recording quality inspection method of this embodiment can be used by different call centers. When a call center has a recording quality inspection demand, it can upload the recording to be inspected to the recording quality inspection system. The recording quality inspection system uses the recording quality inspection method of this embodiment to inspect the recording to be inspected uploaded by the call center and obtain the quality inspection results. Finally, the recording quality inspection system feeds back the quality inspection results to the corresponding call center.

[0083] When a call center needs to perform audio recording quality control, it first places all recordings to be checked into a single data package, pre-configuring the format as required. The system then opens and clicks "Start Uploading Recordings." A data package upload area and a QC rule entry area appear on the system interface. Call center operators simply drag and drop the data package to be checked into the data package upload area, then enter the QC rule entry area.

[0084] Quality control rules consist of quality control phrases and quality control tags. A quality control rule consists of one quality control tag and multiple quality control phrases. A quality control rule can be associated with abusive language. For example, a quality control tag for a quality control rule might be "abusive"; its quality control phrases might be "idiot," "fool," "really stupid," or "too stupid." Quality control rules can also be associated with trade secrets, legally prohibited terms, and so on. Once the call center operator has configured all the quality control rules, they press the "Upload" button, and the recording quality control data package and the corresponding quality control rules are uploaded to the recording quality control system.

[0085] If the call center has other recording quality inspection data packages to upload, they can proceed in the same way: simply drag and drop the recording data package to be inspected into the data package upload area, then enter the corresponding quality inspection rules in the quality inspection rule entry area. Once the quality inspection rules are completed, press the "Upload" button, and the recording quality inspection data package and corresponding quality inspection rules will be uploaded to the recording quality inspection system. If the call center has no other recording quality inspection data packages to upload, simply press the "End Recording Upload" button. After that, the call center only needs to wait for the recording quality inspection system to feedback the recording quality inspection results.

[0086] In this embodiment, step S2 specifically includes:

[0087] Step S21: Receive a signal to start uploading the recording to be inspected.

[0088] When an operator at a call center (assuming this is user A in this example) presses the "Start Uploading Recordings" button on the recording quality inspection system, the recording quality inspection system server receives a signal from user A to start uploading the recording to be inspected. At this point, the recording quality inspection system server allocates storage space to receive and store the recording to be inspected and the inspection rules uploaded by user A.

[0089] Step S22. Create a quality inspection folder to store the recording data package and quality inspection rules uploaded by the user, and one quality inspection folder stores one recording data package and the quality inspection rules corresponding to the recording data package. One recording data package includes multiple recordings to be quality inspected, and the quality inspection rules are stored through the quality inspection rule text.

[0090] When user A uploads a recording data package, the recording quality inspection system's server creates a quality inspection folder in the corresponding storage space to store the recording data package and the corresponding quality inspection rules. The quality inspection folder can be named with the user name, upload time, and upload batch. For example, a quality inspection folder might be named "XX Co., Ltd. XX Year XX Month XX Day XX Hour XX Minute Batch 1."

[0091] In this embodiment, assume that user A uploaded two recording data packages: recording data package 1 and recording data package 2. The recording quality inspection system server stores recording data package 1 in quality inspection folder 1, titled "User A, XX year, XX month, XX day, XX hour, 21 minute, first batch," and also stores the quality inspection rule text 1 corresponding to recording data package 1 in quality inspection folder 1. Recording data package 2 is stored in quality inspection folder 2, titled "User A, XX year, XX month, XX day, XX hour, 23 minute, second batch," and also stores the quality inspection rule text 2 corresponding to recording data package 2 in quality inspection folder 2.

[0092] Step S23: Receive a signal indicating that uploading of the recording to be inspected is complete.

[0093] When user A has finished uploading all the recording data packets and presses the "end of recording upload" button on the recording quality inspection system, the server of the recording quality inspection system will receive a signal from user A indicating that the upload of the recording to be inspected has ended.

[0094] Step S24: Create a quality inspection task bar according to the quality inspection folder. One quality inspection task bar is created for each quality inspection folder. The task bar name of the quality inspection task bar is the same as the folder name of the quality inspection folder.

[0095] When the recording quality inspection system server receives a signal from user A that the recording to be inspected has finished uploading, it creates a quality inspection task bar for the inspection folder. For example, it creates quality inspection task bar #1 for inspection folder #1, with the same name as the inspection folder: "User A, XX year, XX month, XX day, XX hour, 21 minute, first batch." It also creates quality inspection task bar #2 for inspection folder #2, with the same name as the inspection folder: "User A, XX year, XX month, XX day, XX hour, 23 minute, second batch."

[0096] Created quality inspection task bars can be sorted in the recording quality inspection system in chronological order, with earlier created quality inspection task bars being placed at the top. The recording quality inspection system can also display the sorted quality inspection task bars on the task bar display interface, making it easier for the recording quality inspection system staff to understand the general status of the recording quality inspection.

[0097] Step S25: Set the task status of the created quality inspection task bar to the first task status.

[0098] Once the quality inspection folder has created the corresponding quality inspection task bar, the server of the recording quality inspection system will immediately set the task status of the quality inspection task bar to the first task status. For example, in this embodiment, the task status of quality inspection task bar No. 1, named "User A XX year XX month XX day XX hour 21 minute first batch", will be set to the first task status. The task status of quality inspection task bar No. 2, named "User A XX year XX month XX day XX hour 23 minute second batch", will also be set to the first task status.

[0099] like Figure 1 As shown, the recording quality inspection method of this embodiment further includes:

[0100] Step S4. Search for a quality inspection task bar with a task status of the first task status, convert the recording to be inspected corresponding to the found quality inspection task bar into a text to be inspected, and set the task status of the quality inspection task bar to the second task status.

[0101] The recording quality inspection system will automatically search for the quality inspection task bar at regular intervals, find the quality inspection task bar with the task status of the first task status, and convert the recording to be quality inspected in the recording data package in the quality inspection folder corresponding to the found quality inspection task bar into text to be quality inspected.

[0102] In this embodiment, step S4 specifically includes:

[0103] Step S41: Search for a quality inspection task bar whose task status is the first task status.

[0104] The quality inspection task bar is traversed in order to find the first quality inspection task bar whose task status is the first task status. For example, the quality inspection task bar found in this embodiment may be the quality inspection task bar No. 1 named "User A XX year XX month XX day XX hour 21 minute first batch".

[0105] Step S42. Determine the number of times the quality inspection task bar is found when the task state is the first task state. If the number is 1, trigger step S43 and return to step S41; otherwise, directly return to step S41.

[0106] Since the quality inspection task bar No. 1 is found for the first time when the task status is the first task status, it is necessary to return to step S41 after triggering step S43.

[0107] At this point, two steps are carried out simultaneously:

[0108] First aspect:

[0109] Step S43: Acquire a quality inspection folder whose folder name is the same as the task bar name of the quality inspection task bar.

[0110] The name of the quality inspection task bar No. 1 is "User A XX year XX month XX day XX hour 21 minute first batch", so the quality inspection folder No. 1 named "User A XX year XX month XX day XX hour 21 minute first batch" can be found through the name of the quality inspection task bar.

[0111] Step S44: Convert the recordings to be inspected in the inspection folder into texts to be inspected that correspond one to one with the recordings to be inspected.

[0112] Assuming that there are 88 recordings to be inspected in the recording data package of quality inspection folder No. 1, the recording quality inspection system will convert the 88 recordings to be inspected into texts to be inspected in turn. That is, quality inspection folder No. 1 will eventually get 88 texts to be inspected, and the text content in the texts to be inspected is the recording content of the recordings to be inspected.

[0113] It takes a certain amount of time to convert the recordings to be checked into texts, and the specific time depends on the number of recordings to be checked and the length of the recordings to be checked. If there are a large number of recordings to be checked or the length of the recordings to be checked is long, the conversion process will be longer. If there are a small number of recordings to be checked or the length of the recordings to be checked is short, the conversion process will be faster.

[0114] Specifically, step S44 includes:

[0115] S441. Start the recording to be inspected, and identify the first voiceprint and the second voiceprint through the recording to be inspected.

[0116] Assume that a recording to be reviewed contains a conversation between only two people. Since each person has a different voiceprint, two different voiceprints can be extracted from the recording to be reviewed. Furthermore, when extracting voiceprints, the recording to be reviewed does not need to be read in its entirety; only a portion of it is needed. Therefore, the first and second voiceprints can be obtained quickly after the recording to be reviewed begins.

[0117] The specific method of extracting the first and second voiceprints is as follows:

[0118] First, obtain the first recording segment (from the start of the quality inspection recording to the first pause), and calculate the fundamental frequency, formant frequency, intonation, and vocal tract characteristics of the first recording segment (i.e., the first voiceprint feature). Then obtain the second recording segment (from the first recording segment to the first pause), and calculate the fundamental frequency, formant frequency, intonation, and vocal tract characteristics of the second recording segment (i.e., the second voiceprint feature). Next, calculate the similarity value of the two voiceprint features. If the first and second recording segments are produced by two different people, the similarity value of the two voiceprint features will be less than the set threshold. If the first and second recording segments are made by the same person, the similarity between the two voiceprint features will be greater than the set threshold. At this point, it is necessary to continue acquiring the third recording segment (the recording to be inspected, from the second recording segment to the first pause) and calculate the fundamental frequency, formant frequency, intonation, and vocal tract characteristics of the third recording segment (i.e., the third voiceprint feature). The similarity between the third voiceprint feature and the first (or second) voiceprint feature is then calculated. If the third and first (or second) recording segments are made by different people, the similarity between the two voiceprint features will be less than the set threshold. Otherwise, continue acquiring the fourth recording segment.

[0119] This embodiment assumes that the similarity between the first and second voiceprint features is less than a set threshold, indicating that the first and second recording segments were produced by two different individuals. It then determines which of the first and second voiceprint features differs most significantly in terms of fundamental frequency, formant frequency, intonation, or vocal tract characteristics. This embodiment assumes that the fundamental frequency differs most significantly, and thus uses the fundamental frequency value of the first voiceprint feature as the first voiceprint in this step, and the fundamental frequency value of the second voiceprint feature as the second voiceprint in this step.

[0120] S442. Restart the recording to be inspected, convert the recording to be inspected into text, and when the recording to be inspected changes from the first voiceprint to the second voiceprint or from the second voiceprint to the first voiceprint, wrap the text.

[0121] After the first and second voiceprints are extracted, this step will reread the recording to be inspected to obtain the text corresponding to the voice data. The setting of the first and second voiceprints separates the originally uninterrupted text, and each line of text corresponds to a person's speech. Suppose a recording to be inspected is: first, person A speaks (abbreviated as speech a1), then person B speaks (abbreviated as speech b1), then person A speaks (abbreviated as speech a2), and finally person B speaks (abbreviated as speech b2). Then, speech a1 is recorded as a speech by person A in the first line of text; speech b1 is recorded as a speech by person B in the second line of text; speech a2 is recorded as a speech by person A in the third line of text; and speech b2 is recorded as a speech by person B in the fourth line of text. Separating the text lines facilitates the subsequent extraction of text phrases.

[0122] Specifically, assume that the first and second voiceprints in this embodiment are fundamental frequency values. Then, first obtain the first recording segment (from the start of the quality inspection recording to the first pause), then convert the first recording segment into text and store it in the first line of the text. Next, calculate the fundamental frequency value 1 of the first recording segment and compare this fundamental frequency value 1 with the fundamental frequency values of the first and second voiceprints. If the fundamental frequency value 1 is closer to the fundamental frequency value of the first voiceprint, the first recording segment is determined to belong to the first voiceprint; otherwise, the first recording segment is determined to belong to the second voiceprint. This embodiment assumes that the first recording segment belongs to the first voiceprint.

[0123] Next, the second recording segment is obtained (the recording to be inspected, from the first recording segment to the first pause). The fundamental frequency value 2 of the second recording segment is calculated and compared with the fundamental frequencies of the first and second voiceprints. If the fundamental frequency value 2 is closer to that of the first voiceprint, the second recording segment is determined to belong to the first voiceprint; otherwise, the second recording segment is determined to belong to the second voiceprint. If the second recording segment belongs to the second voiceprint, it is a different voiceprint from the first recording segment (i.e., it is produced by two different people). The second recording segment is converted to text and stored in the second line of the text. If the second recording segment belongs to the first voiceprint, it is the same voiceprint as the first recording segment (i.e., it is produced by the same person). The second recording segment is converted to text and stored in the first line of the text. This example assumes that the second recording segment belongs to the second voiceprint. Therefore, the second recording segment is converted to text and stored in the second line of the text.

[0124] Next, the third recording segment is obtained (the recording to be inspected, from the second recording segment to the first pause). The fundamental frequency value 3 of the third recording segment is calculated and compared with the fundamental frequencies of the first and second voiceprints. If the fundamental frequency value 3 is closer to that of the first voiceprint, the third recording segment is determined to belong to the first voiceprint; otherwise, the third recording segment is determined to belong to the second voiceprint. If the third recording segment belongs to the first voiceprint, it is a different voiceprint from the second recording segment (i.e., it is produced by two different people). The third recording segment is converted to text and stored in the third line of the text. If the third recording segment belongs to the second voiceprint, it is the same voiceprint as the second recording segment (i.e., it is produced by the same person). The third recording segment is converted to text and stored in the second line of the text. This example assumes that the third recording segment belongs to the second voiceprint. Therefore, the third recording segment is converted to text and stored in the second line of the text.

[0125] Continue to repeat this step until all the recording segments are converted into text and stored in the text.

[0126] S443. Input a line of text into the machine learning model to obtain a text phrase corresponding to the line of text. Repeat step S443 until all lines of text have corresponding text phrases.

[0127] The machine learning model is not specifically limited and can, for example, be a large natural language model. Before being used, the machine learning model has been trained on a large amount of text with text phrase tags. This machine learning model can understand the semantics of the corresponding line of text to a certain extent, thereby more comprehensively and accurately extracting text phrases.

[0128] S444. Store all text phrases in the text to obtain the text to be quality checked.

[0129] This step integrates all the text phrases and stores them in the text, thus forming the text to be inspected. The subsequent step S641 requires obtaining the phrases to be inspected from the text to be inspected. The text phrases in the text to be inspected formed in this step can be directly used as the phrases to be inspected required for step S641. That is, the text to be inspected formed in this step makes the subsequent step S641 easier to operate.

[0130] Step S45: Set the task status of the quality inspection task bar to the second task status.

[0131] When all the recordings to be inspected in the recording data package of the quality inspection folder No. 1 are converted into texts to be inspected, the recording quality inspection system will set the task status of the quality inspection task bar No. 1 to the second task status.

[0132] Second aspect:

[0133] Continue to step S41. Search for a quality inspection task bar whose task status is the first task status.

[0134] Based on the quality inspection task bar No. 1, the quality inspection task bar is traversed sequentially to find the second quality inspection task bar with the task status of the first task state. For example, the quality inspection task bar found in this embodiment may be the quality inspection task bar No. 2 named "User A XX year XX month XX day XX hour 23 second batch".

[0135] Then continue to step S42... If quality inspection task bar No. 2 is the first time it is found in the first task state, then after triggering step S43, return to step S41; if quality inspection task bar No. 2 is not the first time it is found in the first task state, then directly return to step S41. In this embodiment, since quality inspection task bar No. 2 is also the first time it is found in the first task state, it is also necessary to trigger step S43 before returning to step S41.

[0136] Since there are only quality inspection task bars 1 and 2 in this embodiment, when the process returns to step S41, the entire traversal process of the quality inspection task bars is completed. It is necessary to wait for a certain period of time (for example, 5 minutes) before continuing the traversal process of the quality inspection task bar with the first task status.

[0137] Assume that after a certain period of time, the task status of quality inspection task bar No. 1 is still the first task status (because the recording data package of its corresponding quality inspection folder No. 1 contains more recordings to be inspected, and the conversion time is longer), and the task status of quality inspection task bar No. 2 changes to the second task status (because the recording data package of its corresponding quality inspection folder No. 2 contains fewer recordings to be inspected, and the conversion time is faster). In addition, quality inspection task bar No. 3 is added (the task status of quality inspection task bar No. 3 is the first task status).

[0138] The specific process of re-performing step S4 is as follows:

[0139] Because the task status of quality inspection task bar No. 1 is still the first task status, quality inspection task bar No. 1 is found first. However, because quality inspection task bar No. 1 is not the first time it is found in the first task status, there is no need to trigger the quality inspection text conversion step again. Instead, the search continues for the next quality inspection task bar. Since the task status of quality inspection task bar No. 2 is the second task status, it is skipped and will not be found. Because the task status of quality inspection task bar No. 3 is the first task status, quality inspection task bar No. 3 will be found next. Since quality inspection task bar No. 3 is the first time it is found in the first task status, it is necessary to trigger the quality inspection text conversion step before continuing to search for the next quality inspection task bar. Since there are only three quality inspection task bars at present, the entire traversal process for the current quality inspection task bar is completed. It is necessary to wait for a certain period of time before continuing the traversal and search steps for quality inspection task bars with the first task status.

[0140] like Figure 1 As shown, the recording quality inspection method of this embodiment further includes:

[0141] Step S6. Search for a quality inspection task bar with a task status of the second task status, and perform quality inspection on the text to be quality inspected corresponding to the found quality inspection task bar using quality inspection rules to obtain a quality inspection result, and set the task status of the quality inspection task bar to the third task status.

[0142] The recording quality inspection system automatically searches for quality inspection task bars at regular intervals, finds quality inspection task bars with a task status of the second task status, and then performs quality inspection on the text to be inspected in the quality inspection folder corresponding to the found quality inspection task bar according to the quality inspection rules to obtain a quality inspection result. Step S6 can be performed simultaneously with step S4. Step S4 searches for quality inspection task bars with a task status of the first task status, and step S6 searches for quality inspection task bars with a task status of the second task status.

[0143] In this embodiment, step S6 specifically includes:

[0144] Step S61: Search for a quality inspection task bar whose task status is the second task status.

[0145] The quality inspection task bar is traversed in order to find the first quality inspection task bar whose task status is the second task status. For example, the quality inspection task bar found in this embodiment may be the quality inspection task bar No. 2 named "User A XX year XX month XX day XX hour 23 minute second batch".

[0146] Step S62. Determine the number of times the quality inspection task bar is found when the task state is the second task state. If the number is 1, trigger step S63 and return to step S61; otherwise, directly return to step S61.

[0147] Since the quality inspection task bar No. 2 is found for the first time when the task status is the second task status, it is necessary to return to step S61 after triggering step S63.

[0148] At this point, two steps are carried out simultaneously:

[0149] First aspect:

[0150] Step S63: Acquire a quality inspection folder whose folder name is the same as the task bar name of the quality inspection task bar.

[0151] The name of the quality inspection task bar No. 2 is "User A's 2nd batch at XX / XX / XX / XX:23", so the quality inspection folder No. 2 named "User A's 2nd batch at XX / XX / XX / XX:23" can be found through the name of the quality inspection task bar.

[0152] Step S64: The documents to be inspected in the quality inspection folder are inspected in turn according to the quality inspection rules to obtain quality inspection results corresponding to the documents to be inspected, and all quality inspection results are stored in the same quality inspection result document.

[0153] Assuming that there are 6 texts to be quality inspected in quality inspection folder No. 2, the recording quality inspection system will inspect the 6 texts to be quality inspected in turn to obtain quality inspection results, that is, quality inspection folder No. 2 will eventually obtain 1 quality inspection result text, which stores 6 quality inspection results.

[0154] The quality inspection process takes a certain amount of time, and the specific time is determined by the number of documents to be inspected and the content of the documents to be inspected. If the number of documents to be inspected is large or the content is large, the quality inspection process will be longer. If the number of documents to be inspected is small or the content is small, the quality inspection process will be faster.

[0155] Step S65: Set the task status of the quality inspection task bar to the third task status.

[0156] When all the texts to be inspected in the quality inspection folder No. 2 have been inspected, the recording quality inspection system will set the task status of the quality inspection task bar No. 2 to the third task status.

[0157] Second aspect:

[0158] Continue to step S61. Search for a quality inspection task bar whose task status is the second task status.

[0159] Based on quality inspection task bar No. 2, the process continues to traverse the quality inspection task bars in order until a second quality inspection task bar with the second task status is found. In this embodiment, there are only three quality inspection task bars, and only quality inspection task bar No. 2 has the second task status. Therefore, when the process returns to step S61, the entire traversal of the quality inspection task bars is complete. It is necessary to wait for a certain period of time (e.g., 8 minutes) before continuing the traversal and searching for the quality inspection task bar with the second task status.

[0160] Assume that after a certain period of time, the task status of quality inspection task bar No. 1 changes to the second task status, the task status of quality inspection task bar No. 2 is still the second task status, and the task status of quality inspection task bar No. 3 is still the first task status.

[0161] The specific process of re-performing step S6 is as follows:

[0162] Because the task status of quality inspection task bar No. 1 is the second task status, quality inspection task bar No. 1 is found first. And because quality inspection task bar No. 1 is the first time it is found in the second task status, it is necessary to trigger the quality inspection step before continuing to search for the next quality inspection task bar. Since the task status of quality inspection task bar No. 2 is still the second task status, quality inspection task bar No. 2 will be found next. However, because quality inspection task bar No. 2 is not the first time it is found in the second task status, there is no need to trigger the quality inspection step again. Instead, the search for the next quality inspection task bar will continue. Since the task status of quality inspection task bar No. 3 is the first task status, it will be skipped directly and will not be found. Since there are only three quality inspection task bars at present, the entire traversal process for the current quality inspection task bar has ended. It is necessary to wait for a certain period of time before continuing the traversal and search steps for the quality inspection task bar with the second task status.

[0163] Furthermore, step S64 of this embodiment specifically includes:

[0164] Step S641: Obtain a text to be inspected, and obtain phrases to be inspected in the text to be inspected.

[0165] Taking QA task No. 2 as an example, the recording QA system first obtains the first QA text in QA folder No. 2, and then obtains the phrases to be QA from the first QA text. Assume that the phrases to be QA obtained are X1, X2, X3, X4, X5, X6, X7, X8, X9, and X10.

[0166] Step S642: Obtain a quality inspection rule in the quality inspection rule text, where the quality inspection rule includes a quality inspection phrase and a quality inspection tag.

[0167] The recording quality inspection system then obtains the first quality inspection rule from the quality inspection rule text in quality inspection folder No. 2. Assume that the quality inspection tag of the first quality inspection rule is AA, and the quality inspection phrases are A1, A2, A3, A4, and A5.

[0168] Step S643. Match the phrase to be inspected with the quality inspection phrases in the quality inspection rules. When the phrase to be inspected matches at least one quality inspection phrase, the text to be inspected is determined to be abnormal and the quality inspection label of the quality inspection rule is assigned to the text to be inspected. Otherwise, the text to be inspected is determined to be normal.

[0169] The phrases to be inspected are matched against the inspection phrases one by one. For example, first, the phrase to be inspected X1 is matched against the inspection phrase. If there is a match between the phrase to be inspected X1 and the inspection phrase, then the subsequent phrases to be inspected do not need to be matched, and the matching result of the first inspection rule is obtained. If there is no match between the phrase to be inspected X1 and the inspection phrase, then the matching continues with the phrase to be inspected X2...

[0170] Assuming that in this embodiment, the phrase X2 to be inspected matches the phrase A3, then the result of the first text to be inspected relative to the first quality inspection rule is that the text to be inspected is abnormal, and the quality inspection label of the text to be inspected is AA.

[0171] Step S644: Determine whether there is any quality inspection rule in the quality inspection rule text that does not match the text to be quality inspected. If there is any quality inspection rule that does not match, return to step S642; otherwise, go to step S645.

[0172] Assuming that there is a second quality inspection rule in the quality inspection rule text of quality inspection folder No. 2, then return to step S642 to obtain the second quality inspection rule in the quality inspection rule text. Assuming that the quality inspection label of the second quality inspection rule is BB, and the quality inspection phrases are B1, B2, B3, B4 and B5. Then match the phrases to be quality inspected X1, X2, X3, X4, X5, X6, X7, X8, X9, X10 with the quality inspection phrases B1, B2, B3, B4, B5. When the phrases to be quality inspected X1 and the quality inspection phrases B2 match, it is determined that the result of the first text to be quality inspected relative to the second quality inspection rule is that the text to be quality inspected is abnormal, and the quality inspection label of the text to be quality inspected is BB. Assuming that there is no next quality inspection rule in the quality inspection rule text of quality inspection folder No. 2, then enter step S645.

[0173] Step S645. When the text to be inspected is determined to be normal each time, the quality inspection result of the text to be inspected is determined to be normal; when the text to be inspected is determined to be abnormal at least once, the quality inspection result of the text to be inspected is determined to be abnormal, and a quality inspection label is assigned to the quality inspection result.

[0174] In this embodiment, since the first document to be inspected is judged to be abnormal twice and the inspection labels are AA and BB respectively, the inspection result of the first document to be inspected is abnormal and the inspection labels of the inspection result are AA and BB.

[0175] Then, steps S641 to S645 are repeated to perform quality inspection on the second document to be inspected in the quality inspection folder No. 2. Assuming that the quality inspection result of the second document to be inspected is normal with respect to the first quality inspection rule and the quality inspection result with respect to the second quality inspection rule is also normal, then the quality inspection result of the second document to be inspected is normal.

[0176] Repeat steps S641 to S645 to perform a quality inspection on the third document in the quality inspection folder No. 2. Assume that the quality inspection result of the third document is normal with respect to the first quality inspection rule, but abnormal with respect to the second quality inspection rule, and the quality inspection label is BB. Then, the quality inspection result of the third document is abnormal, and the quality inspection label of the quality inspection result is BB.

[0177] Repeat steps S641 to S645 until the quality inspection results for all six documents in the quality inspection folder No. 2 are obtained. Assume that the quality inspection result of the fourth document is abnormal, and the quality inspection label of the quality inspection result is AA; the quality inspection result of the fifth document is normal; and the quality inspection result of the sixth document is normal.

[0178] At this point, the quality inspection results of all the texts to be inspected in the quality inspection folder No. 2 have been obtained, and these quality inspection results are stored in the quality inspection result text of the quality inspection folder No. 2 in chronological order.

[0179] like Figure 1 As shown, the recording quality inspection method of this embodiment further includes:

[0180] Step S8. Search for a quality inspection task bar whose task status is the third task status, send the quality inspection result corresponding to the found quality inspection task bar to the user, and set the task status of the quality inspection task bar to the fourth task status.

[0181] The recording quality control system automatically searches for quality control task bars at regular intervals, finds any that are in the third task state, and sends the quality control results from the quality control folder corresponding to the found task bar to the user. Once the quality control results are sent, the recording quality control system sets the task state of the corresponding task bar to the fourth task state.

[0182] Step S8 can be performed synchronously with step S4 and step S6. Step S4 is to search for the quality inspection task bar with the task status of the first task status, step S6 is to search for the quality inspection task bar with the task status of the second task status, and step S8 is to search for the quality inspection task bar with the task status of the third task status.

[0183] The staff of the recording quality inspection system can see the quality inspection task bars of different task states on the display interface of the recording quality inspection system. When the task state of the quality inspection task bar is the fourth task state, the staff can manually delete the quality inspection task bar and the quality inspection folder corresponding to the quality inspection task bar.

[0184] The recording quality inspection method of this embodiment can help call centers perform recording quality inspection. When there is a need for recording quality inspection, the call center only needs to upload the recording to be inspected and the quality inspection rules to the recording quality inspection system. After receiving the uploaded recording to be inspected and the quality inspection rules, the recording quality inspection system can create corresponding quality inspection folders and quality inspection task bars for the recording to be inspected and the quality inspection rules, and first convert the recording to be inspected into a text to be inspected through the quality inspection folder and the quality inspection task bar, and then perform quality inspection on the text to be inspected to obtain the quality inspection results, and finally feed back the quality inspection results to the call center. No matter how many recordings to be inspected the call center has, the recording quality inspection system can comprehensively and relatively quickly feed back the recording quality inspection results.

[0185] In addition, this embodiment creates a quality inspection task bar for each quality inspection task (i.e., a quality inspection folder that stores the recorded data package to be quality inspected). The quality inspection task bar includes a task name that is the same as the quality inspection folder name and a task status corresponding to the quality inspection status of the recorded data to be quality inspected in the quality inspection folder.

[0186] When a quality inspection folder (i.e., a quality inspection task bar) is just created, the task status is the first state. The recording quality inspection system automatically scans the task status of the quality inspection task bar. When a quality inspection task bar enters the first state for the first time, the recording quality inspection system automatically determines that the recording data to be inspected in the quality inspection folder with the same task name as the quality inspection task bar needs to be converted from the recording to be inspected to text to be inspected. Then, the recording quality inspection system automatically triggers the "convert the recording to be inspected to text to be inspected" quality inspection process for the corresponding quality inspection folder.

[0187] When the audio recordings to be inspected in a quality inspection folder are converted to text to be inspected, the task status of the quality inspection task bar corresponding to the quality inspection folder will automatically change to the second status. The audio quality inspection system will automatically scan the task status of the quality inspection task bar. When a quality inspection task bar enters the second status for the first time, the audio quality inspection system will automatically determine that the text to be inspected in the quality inspection folder with the same task name as the quality inspection task bar needs to be inspected according to the quality inspection rules. Then, the audio quality inspection system will automatically trigger the quality inspection procedure of "performing quality inspection according to the quality inspection rules to obtain quality inspection results" for the corresponding quality inspection folder.

[0188] When a quality inspection folder receives a quality inspection result, the task status of the quality inspection task bar corresponding to the quality inspection folder will automatically change to the third status. The recording quality inspection system will automatically scan the task status of the quality inspection task bar. When a quality inspection task bar enters the third status for the first time, the recording quality inspection system will automatically determine that the quality inspection folder with the same task name as the quality inspection task bar has received a quality inspection result and needs to be sent to the user. Then, the recording quality inspection system will automatically trigger the "quality inspection result sent to user" quality inspection procedure for the corresponding quality inspection folder.

[0189] When a quality inspection folder has sent its quality inspection results to the user, the task status of the quality inspection task bar corresponding to the quality inspection folder will automatically change to the fourth status. The recording quality inspection system will automatically scan the task status of the quality inspection task bar. When a quality inspection task bar reaches the fourth status, the recording quality inspection system will automatically determine that the quality inspection folder with the same task name as the quality inspection task bar has completed the recording quality inspection operation.

[0190] This embodiment sets a quality inspection task bar so that the recording quality inspection system can automatically and efficiently track each quality inspection task, and then enables the recording quality inspection system to process the relevant procedures for each quality inspection task in a timely manner, thereby improving the efficiency of recording quality inspection.

[0191] Example 2:

[0192] A recording quality inspection system, such as Figure 2 As shown, it includes: a quality inspection task bar creation module, a quality inspection text conversion module, a quality inspection result acquisition module and a quality inspection result sending module.

[0193] The quality inspection task bar creation module is used to receive the recordings and quality inspection rules uploaded by the user, and create a quality inspection task bar for the received recordings and quality inspection rules. The quality inspection task bar includes the task name and task status, and the task status of the quality inspection task bar is set to the first task status.

[0194] The text conversion module for quality inspection is used to search for a quality inspection task bar with a task status of the first task status, convert the quality inspection recording corresponding to the found quality inspection task bar into a text for quality inspection, and set the task status of the quality inspection task bar to the second task status.

[0195] The quality inspection result acquisition module is used to find the quality inspection task bar with the task status of the second task status, and perform quality inspection on the text to be quality inspected corresponding to the found quality inspection task bar through quality inspection rules to obtain the quality inspection result, and set the task status of the quality inspection task bar to the third task status.

[0196] The quality inspection result sending module is used to search for a quality inspection task bar with a task status of the third task status, send the quality inspection result corresponding to the found quality inspection task bar to the user, and set the task status of the quality inspection task bar to the fourth task status.

[0197] Among them, the quality inspection task bar creation module includes: a recording upload start signal receiving unit, a recording and rule storage unit, a recording upload end signal receiving unit, a quality inspection task bar creation unit and a first task status setting unit. The recording upload start signal receiving unit is used to receive the recording upload start signal to be quality inspected. The recording and rule storage unit is used to create a quality inspection folder to store the recording data package and quality inspection rules uploaded by the user, and one quality inspection folder stores one recording data package and the quality inspection rules corresponding to the recording data package. One recording data includes multiple recordings to be quality inspected, and the quality inspection rules are stored through the quality inspection rule text. The recording upload end signal receiving unit is used to receive the recording upload end signal to be quality inspected. The quality inspection task bar creation unit is used to create a quality inspection task bar according to the quality inspection folder. One quality inspection task bar is created for each quality inspection folder, and the task bar name of the quality inspection task bar is the same as the folder name of the quality inspection folder. The first task status setting unit is used to set the task state of the created quality inspection task bar to the first task state.

[0198] The module for converting text to be inspected includes: a first task bar search unit, a first task bar search times determination unit, a first quality inspection folder acquisition unit, a text conversion unit, and a second task status setting unit. The first task bar search unit is used to search for a quality inspection task bar with a task status of the first task status. The first task bar search times determination unit is used to determine the number of times a quality inspection task bar has been found when the task status is the first task status. The first quality inspection folder acquisition unit is used to obtain a quality inspection folder with a folder name that is the same as the task bar name of the quality inspection task bar. The text conversion unit is used to convert the recordings to be inspected in the quality inspection folder into text to be inspected that corresponds one-to-one to the recordings to be inspected. The second task status setting unit is used to set the task status of the quality inspection task bar to the second task status.

[0199] The text conversion unit includes: a voiceprint recognition subunit, a text character acquisition subunit, a text phrase acquisition subunit and a text to be inspected formation subunit.

[0200] The voiceprint recognition subunit is used to start the recording to be inspected and identify the first and second voiceprints through the recording to be inspected. The text acquisition subunit is used to restart the recording to be inspected, convert the recording to be inspected into text, and perform line wrapping on the text when the recording to be inspected changes from the first voiceprint to the second voiceprint or from the second voiceprint to the first voiceprint. The text phrase acquisition subunit is used to input a line of text into the machine learning model to obtain the text phrase corresponding to the line of text. The text to be inspected formation subunit is used to store all text phrases in the text to obtain the text to be inspected.

[0201] The quality inspection result acquisition module includes: a second task bar search unit, a second task bar search times determination unit, a second quality inspection folder acquisition unit, a quality inspection result acquisition unit and a third task status setting unit. The second task bar search unit is used to search for a quality inspection task bar with a task status of the second task status. The second task bar search times determination unit is used to determine the number of times the quality inspection task bar is found when the task status is the second task status. The second quality inspection folder acquisition unit is used to acquire a quality inspection folder with the same folder name as the task bar name of the quality inspection task bar. The quality inspection result acquisition unit is used to perform quality inspection on the text to be quality inspected in the quality inspection folder in turn through the quality inspection rules to obtain quality inspection results corresponding one-to-one to the text to be quality inspected, and all quality inspection results are stored in the same quality inspection result text. The third task status setting unit is used to set the task status of the quality inspection task bar to the third task status.

[0202] The quality inspection result acquisition unit includes: a phrase acquisition subunit for quality inspection, a quality inspection rule acquisition subunit, a text determination subunit for quality inspection, a quality inspection rule text determination subunit, and a quality inspection result determination subunit. The phrase acquisition subunit for quality inspection is used to acquire a text to be quality inspected and acquire the phrase to be quality inspected in the text to be quality inspected. The quality inspection rule acquisition subunit is used to acquire a quality inspection rule in the quality inspection rule text, and the quality inspection rule includes a quality inspection phrase and a quality inspection label. The text determination subunit for quality inspection is used to match the phrase to be quality inspected with the quality inspection phrases in the quality inspection rule. When the phrase to be quality inspected matches at least one quality inspection phrase, the text to be quality inspected is determined to be abnormal and the quality inspection label of the quality inspection rule is assigned to the text to be quality inspected. Otherwise, the text to be quality inspected is determined to be normal. The quality inspection rule text determination subunit is used to determine whether there is a quality inspection rule in the quality inspection rule text that does not match the text to be quality inspected. The quality inspection result determination subunit is used to determine the quality inspection result of the text to be quality inspected as normal when the text to be quality inspected is determined to be normal each time; when the text to be quality inspected is determined to be abnormal at least once, the quality inspection result of the text to be quality inspected is determined to be abnormal, and a quality inspection label of the text to be quality inspected is assigned to the quality inspection result.

[0203] The recording quality inspection system of this embodiment can help call centers perform recording quality inspection. When there is a need for recording quality inspection, the call center only needs to upload the recording to be inspected and the quality inspection rules to the recording quality inspection system. After receiving the uploaded recording to be inspected and the quality inspection rules, the recording quality inspection system can create corresponding quality inspection folders and quality inspection task bars for the recording to be inspected and the quality inspection rules, and first convert the recording to be inspected into a text to be inspected through the quality inspection folder and the quality inspection task bar, and then perform quality inspection on the text to be inspected to obtain the quality inspection results, and finally feedback the quality inspection results to the call center. No matter how many recordings to be inspected the call center has, the recording quality inspection system can comprehensively and relatively quickly feedback the recording quality inspection results.

[0204] Although certain embodiments of the present specification are shown in the accompanying drawings, it should be understood that the present specification can be implemented in various forms and should not be construed as limited to the embodiments described herein. Instead, these embodiments are provided to provide a more thorough and complete understanding of the present specification. It should be understood that the drawings and embodiments of the present specification are for illustrative purposes only and are not intended to limit the scope of protection of the present specification.

Claims

1. A recording quality inspection method, characterized in that: include: Step S2. Receive the recording to be quality checked and the quality inspection rules uploaded by the user, and create a quality inspection task bar for the received recording to be quality checked and the quality inspection rules, the quality inspection task bar includes a task name and a task status, and the task status of the quality inspection task bar is set to the first task status; Step S4. Find the quality inspection task bar with the first task status, convert the quality inspection recording corresponding to the found quality inspection task bar into the quality inspection text, and set the task status of the quality inspection task bar to the second task status; The step S4 specifically includes: Step S41. Search for a quality inspection task bar with a task status of the first task state; Step S42. Determine the number of times the quality inspection task bar is found in the task status of the first task state. If the number is 1, trigger Step S43 and return to Step S41; otherwise, directly return to Step S41; Step S43. Obtain a quality inspection folder with a folder name identical to the task bar name of the quality inspection task bar; Step S44. Convert the recordings to be inspected in the quality inspection folder into texts to be inspected corresponding to the recordings to be inspected; Step S45. Set the task status of the quality inspection task bar to the second task state; The step S44 specifically includes: S441. Start the recording to be inspected, and identify the first voiceprint and the second voiceprint through the recording to be inspected; The S441 includes: S4411. obtaining a first recording segment and calculating the fundamental frequency value, formant frequency value, intonation value and vocal tract feature value of the first recording segment to obtain a first voiceprint feature; S4412. obtaining a second recording segment and calculating the fundamental frequency value, formant frequency value, intonation value and vocal tract feature value of the second recording segment to obtain a second voiceprint feature; S4413. calculating the similarity value between the first voiceprint feature and the second voiceprint feature. When the similarity value between the first voiceprint feature and the second voiceprint feature is less than a set threshold, the first voiceprint and the second voiceprint recognition is completed and S4415 is executed; otherwise, S4414 is executed; S4414. obtaining a third recording segment and using it as the second recording segment, and S4412 is executed; S4415. determining which of the fundamental frequency value, formant frequency value, intonation value and vocal tract feature value of the first voiceprint feature and the second voiceprint feature has the largest difference and determining the feature value with the largest difference as the first voiceprint and the second voiceprint; S442. Restart the recording to be inspected, convert the recording to be inspected into text, and when the recording to be inspected changes from the first voiceprint to the second voiceprint or from the second voiceprint to the first voiceprint, wrap the text; S443. Inputting a line of text into the machine learning model to obtain a text phrase corresponding to the line of text; repeating step S443 until all lines of text have obtained corresponding text phrases; S444. All the text phrases are stored in the text to obtain the text to be inspected; Step S6. Find a quality inspection task bar with a task status of the second task status, perform quality inspection on the to-be-inspected text corresponding to the found quality inspection task bar according to the quality inspection rules to obtain a quality inspection result, and set the task status of the quality inspection task bar to the third task status; Step S8. Search for a quality inspection task bar whose task status is the third task status, send the quality inspection result corresponding to the found quality inspection task bar to the user, and set the task status of the quality inspection task bar to the fourth task status.

2. The recording quality inspection method according to claim 1, characterized in that: The step S2 specifically includes: Step S21. Receive a signal to start uploading the recording to be inspected; Step S22. Create a quality inspection folder to store the recording data package and quality inspection rules uploaded by the user, and one of the quality inspection folders stores a recording data package and the quality inspection rules corresponding to the recording data package, one of the recording data includes multiple recordings to be quality inspected, and the quality inspection rules are stored through the quality inspection rule text; Step S23: receiving a signal indicating that uploading of the recording to be inspected has ended; Step S24. Create a quality inspection task bar according to the quality inspection folder. One quality inspection task bar is created for each quality inspection folder. The task bar name of the quality inspection task bar is the same as the folder name of the quality inspection folder. Step S25: Set the task status of the created quality inspection task bar to the first task status.

3. The recording quality inspection method according to claim 1, characterized in that: The step S6 comprises: Step S61: Find a quality inspection task bar whose task status is the second task status; Step S62. Determine the number of times the quality inspection task bar is found in the second task state. If the number is 1, step S63 is triggered and the process returns to step S61. Otherwise, the process returns to step S61. Step S63. Acquire a quality inspection folder whose folder name is the same as the task bar name of the quality inspection task bar; Step S64: The documents to be inspected in the quality inspection folder are inspected in turn according to the quality inspection rules to obtain quality inspection results corresponding to the documents to be inspected, and all the quality inspection results are stored in the same quality inspection result file; Step S65. Setting the task status of the quality inspection task bar to the third task status; Wherein, the step S64 specifically includes: Step S641: Obtain a text to be inspected, and obtain a phrase to be inspected in the text to be inspected; Step S642: Obtain a quality inspection rule from the quality inspection rule text, wherein the quality inspection rule includes a quality inspection phrase and a quality inspection tag; Step S643. Match the phrase to be inspected with the inspection phrases in the inspection rules. If the phrase to be inspected matches at least one of the inspection phrases, determine that the text to be inspected is abnormal and assign the inspection label of the inspection rule to the text to be inspected. Otherwise, determine that the text to be inspected is normal. Step S644: Determine whether there is any quality inspection rule in the quality inspection rule text that does not match the text to be inspected. If there is any quality inspection rule that does not match, return to step S642; otherwise, proceed to step S645. Step S645. When the text to be quality inspected is determined to be normal each time, the quality inspection result of the text to be quality inspected is determined to be normal; when the text to be quality inspected is determined to be abnormal at least once, the quality inspection result of the text to be quality inspected is determined to be abnormal, and the quality inspection label of the text to be quality inspected is assigned to the quality inspection result.

4. A recording quality inspection system, characterized in that: include: A quality inspection task bar creation module is used to receive the recording to be quality inspected and the quality inspection rules uploaded by the user, and create a quality inspection task bar for the received recording to be quality inspected and the quality inspection rules, wherein the quality inspection task bar includes a task name and a task status, and the task status of the quality inspection task bar is set to the first task status; a text-to-be-inspected conversion module, configured to search for a quality inspection task bar whose task status is a first task status, convert the quality inspection recording corresponding to the found quality inspection task bar into a text-to-be-inspected, and set the task status of the quality inspection task bar to a second task status; The text conversion module to be inspected includes: a first task bar search unit for searching a quality inspection task bar whose task status is the first task status; a first task bar search times determination unit for determining the number of times the quality inspection task bar is found when the task status is the first task status; a first quality inspection folder acquisition unit for acquiring a quality inspection folder having the same folder name as the task bar name of the quality inspection task bar; a text conversion unit for sequentially converting the recordings to be quality inspected in the quality inspection folder into texts to be quality inspected that correspond one-to-one to the recordings to be quality inspected; and a second task status setting unit for setting the task status of the quality inspection task bar to the second task status; The text conversion unit includes: The voiceprint recognition subunit is configured to perform the following steps: S441. Start the recording to be inspected, and identify the first voiceprint and the second voiceprint through the recording to be inspected; S441 includes: S4411. Obtain the first recording segment and calculate the fundamental frequency value, formant frequency value, intonation value and vocal tract characteristic value of the first recording segment to obtain the first voiceprint feature; S4412. Obtain the second recording segment and calculate the fundamental frequency value, formant frequency value, intonation value and vocal tract characteristic value of the second recording segment to obtain the second voiceprint feature; S4413. Calculate Calculate the similarity value between the first voiceprint feature and the second voiceprint feature. When the similarity value between the first voiceprint feature and the second voiceprint feature is less than the set threshold, the recognition of the first voiceprint and the second voiceprint is completed and S4415 is executed; otherwise, S4414 is executed; S4414. The third recording segment is obtained and used as the second recording segment, and S4412 is executed; S4415. Determine which of the fundamental frequency value, formant frequency value, intonation value and vocal tract feature value of the first voiceprint feature and the second voiceprint feature has the largest difference, and determine the feature value with the largest difference as the first voiceprint and the second voiceprint; The text acquisition subunit is used to restart the recording to be inspected, convert the recording to be inspected into text, and perform line wrapping on the text when the recording to be inspected changes from the first voiceprint to the second voiceprint or from the second voiceprint to the first voiceprint; a text phrase acquisition subunit, configured to input a line of text characters into a machine learning model to obtain a text phrase corresponding to the line of text characters; The text to be inspected forms a subunit, which is used to store all the text phrases in the text to obtain the text to be inspected; a quality inspection result acquisition module, configured to search for a quality inspection task bar whose task status is the second task status, perform quality inspection on the to-be-inspected text corresponding to the found quality inspection task bar according to the quality inspection rules to obtain a quality inspection result, and set the task status of the quality inspection task bar to the third task status; The quality inspection result sending module is used to search for a quality inspection task bar with a task status of the third task status, send the quality inspection result corresponding to the found quality inspection task bar to the user, and set the task status of the quality inspection task bar to the fourth task status.

5. The recording quality inspection system according to claim 4, characterized in that: The quality inspection task bar creation module includes: A recording upload start signal receiving unit, used to receive a recording upload start signal to be inspected; A recording and rule storage unit is used to create a quality inspection folder to store recording data packets and quality inspection rules uploaded by users, and one of the quality inspection folders stores one recording data packet and the quality inspection rules corresponding to the recording data packet, one of the recording data packets includes multiple recordings to be quality inspected, and the quality inspection rules are stored in the form of quality inspection rule text; A recording upload completion signal receiving unit, used to receive a signal indicating the completion of uploading the recording to be inspected; A quality inspection task bar creation unit is used to create a quality inspection task bar according to the quality inspection folder, one quality inspection task bar is created for each quality inspection folder, and the task bar name of the quality inspection task bar is the same as the folder name of the quality inspection folder; The first task state setting unit is used to set the task state of the created quality inspection task bar to the first task state.

6. The recording quality inspection system according to claim 4, characterized in that: The quality inspection result acquisition module includes: A second task bar searching unit, configured to search for a quality inspection task bar whose task status is the second task status; A second task bar search times determination unit, configured to determine the number of times the quality inspection task bar has been found when the task state is the second task state; A second quality inspection folder obtaining unit is used to obtain a quality inspection folder whose folder name is the same as the task bar name of the quality inspection task bar; A quality inspection result acquisition unit is used to perform quality inspection on the documents to be inspected in the quality inspection folder in turn according to the quality inspection rules to obtain quality inspection results corresponding to the documents to be inspected, and all the quality inspection results are stored in the same quality inspection result text; A third task state setting unit, configured to set the task state of the quality inspection task bar to a third task state; Wherein, the quality inspection result acquisition unit includes: The phrase to be inspected obtaining subunit is used to obtain a text to be inspected and obtain the phrase to be inspected in the text to be inspected; A quality inspection rule acquisition subunit, configured to acquire a quality inspection rule from the quality inspection rule text, wherein the quality inspection rule includes a quality inspection phrase and a quality inspection tag; a text-to-be-checked determination subunit, configured to match the phrase to be checked with the quality inspection phrases in the quality inspection rules; when the phrase to be checked matches at least one of the quality inspection phrases, determine that the text to be checked is abnormal and assign the quality inspection label of the quality inspection rule to the text to be checked; otherwise, determine that the text to be checked is normal; A quality inspection rule text determination subunit is used to determine whether there is a quality inspection rule in the quality inspection rule text that does not match the text to be quality inspected; The quality inspection result determination subunit is used to determine the quality inspection result of the text to be quality inspected as normal when the text to be quality inspected is determined to be normal each time; when the text to be quality inspected is determined to be abnormal at least once, determine the quality inspection result of the text to be quality inspected as abnormal, and assign the quality inspection label of the text to be quality inspected to the quality inspection result.

Images