Data processing method and device, electronic equipment and storage medium

Abstract

The embodiment of the invention provides a data processing method and device, electronic equipment and a storage medium, and the data processing method comprises the steps: obtaining seat recording data; if the seat recording data corresponds to at least two speakers and the at least two speakers comprise a seat user and a non-seat user, determining a target time period corresponding to the non-seat user in the seat recording data; acquiring client recording data; and according to the coincidence rate of the target time period and the speaking time period of the customer and the similarity between the voiceprint of the non-seat user and the voiceprint of the customer, determining whether audio data of the customer is serially connected into the seat recording data. Therefore, the detection accuracy can be improved when the crosstalk source in the seat recording data is detected.

Description

Technical Field

[0001] The present application relates to the technical field of voice data processing, and in particular to a data processing method, device, electronic device and storage medium. Background Art

[0002] In an outbound call scenario, the agent user can be a staff member of the business party, who is responsible for providing services to the business party's customers during the call. The business party needs to monitor the agent user's speech to ensure that what the agent user says meets the business party's requirements. Therefore, the business party often stores the agent user's recording data and the customer's recording data separately, and performs quality inspection on the agent user's recording data. However, the agent recording data may occasionally be mixed with voices other than the agent user himself, such as the customer's voice. Considering that the business party's requirements for the agent user's speech only constrain the agent user and the customer is not affected by it, if the customer's voice is strung into the agent recording data, it may cause a misjudgment of the quality inspection result.

[0003] Therefore, the business side has a need to accurately identify whether there is customer crosstalk in the agent recording data. If the accuracy of identifying customer crosstalk is not high, it will be detrimental to subsequent operation and maintenance, data analysis and business optimization. Summary of the invention

[0004] The embodiments of the present application provide a data processing method, device, electronic device and storage medium to improve the detection accuracy when detecting the source of noise in seat recording data.

[0005] In a first aspect, an embodiment of the present application provides a data processing method, including:

[0006] Acquire agent recording data; the agent recording data refers to the audio data of the agent user's speaking period when the agent user is communicating with the customer;

[0007] If the seat recording data corresponds to at least two speakers and the at least two speakers include the seat user and a non-seat user, determining a target time period corresponding to the non-seat user in the seat recording data;

[0008] Acquire customer recording data; the customer recording data refers to the audio data of the customer's speaking period when the agent user is communicating with the customer;

[0009] Whether the audio data of the customer is inserted into the seat recording data is determined according to the overlap rate between the target time period and the speaking time period of the customer, and the similarity between the voiceprint of the non-seat user and the voiceprint of the customer.

[0010] In a second aspect, an embodiment of the present application provides a data processing device, the device comprising:

[0011] An acquisition unit, used to acquire seat recording data; the seat recording data refers to the audio data of the seat user's speaking period when the seat user is communicating with the customer;

[0012] a determining unit, configured to determine a target time period corresponding to the non-seat user in the seat recording data if the seat recording data corresponds to at least two speakers and the at least two speakers include the seat user and a non-seat user;

[0013] The acquisition unit is further used to acquire customer recording data; the customer recording data refers to the audio data of the customer's speaking period when the seat user communicates with the customer;

[0014] The determination unit is further used to determine whether the customer's audio data is inserted into the seat recording data based on the overlap rate between the target time period and the customer's speaking time period, and the similarity between the voiceprint of the non-seat user and the voiceprint of the customer.

[0015] In a third aspect, an embodiment of the present application provides an electronic device, comprising: a processor; and a memory configured to store computer-executable instructions, wherein the computer-executable instructions, when executed, cause the processor to execute the data processing method as described in the first aspect.

[0016] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium for storing computer-executable instructions, which, when executed by a processor, implement the data processing method as described in the first aspect.

[0017] It can be seen that in the embodiment of the present application, first, the seat recording data is obtained; the seat recording data refers to the audio data of the seat user's speaking period when the seat user is communicating with the customer; if the seat recording data corresponds to at least two speakers and at least the two speakers include the seat user and the non-seat user, then the target time period corresponding to the non-seat user in the seat recording data is determined; then, the customer recording data is obtained; the customer recording data refers to the audio data of the customer's speaking period when the seat user is communicating with the customer; finally, based on the overlap rate between the target time period and the customer's speaking period, and the similarity between the voiceprint of the non-seat user and the voiceprint of the customer, it is determined whether the customer's audio data is inserted into the seat recording data. In this way, when it is determined that the agent recording data corresponds to at least two speakers, it can be determined that there is noise mixed in the agent recording data. The noise may come from the customer or from other agent users located next to the agent user. Different sources of noise may be caused by different abnormalities. Accurately identifying the source of noise is beneficial to subsequent operation and maintenance. In addition, the purpose of storing the agent recording data and the customer recording data separately includes quality inspection of the agent recording data to supervise the service quality of the agent user. Considering that the quality inspection result of the agent recording data is more likely to be affected when the customer is misjudged as an agent user, it is more important to determine whether the noise comes from the customer. By analyzing the overlap rate between the target time period corresponding to the non-agent user who emits the noise in the agent recording data and the customer's speaking time period, as well as the similarity between the voiceprint of the non-agent user and the voiceprint of the customer, it can be determined from multiple dimensions whether the customer's recording data is inserted into the agent recording data, thereby improving the accuracy of identifying the source of the noise. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this specification. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative labor.

[0019] Figure 1 A processing flow chart of a data processing method provided in an embodiment of the present application;

[0020] Figure 2 A flowchart of another data processing method provided in an embodiment of the present application;

[0021] Figure 3 A schematic diagram of a data processing device provided in an embodiment of the present application;

[0022] Figure 4 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0023] In order to enable those skilled in the art to better understand the technical solutions in the embodiments of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of this specification, not all of the embodiments. Based on the embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0024] In the outbound call scenario, the business party supervises the agent user by quality checking the recording data of the agent user during the call. The business party can record the entire conversation between the agent user and the customer through the call center, and save the voice of the agent user in one channel (i.e., the agent recording data), and save the voice of the customer (i.e., the customer recording data) in another channel. The purpose of separate storage is to facilitate the quality inspection of the agent user's agent recording data. Therefore, in the most ideal case, after the complete recording data is split into agent recording data and customer recording data, the customer recording data only includes the customer's voice, and the agent recording data only includes the agent user's own voice, and does not include the customer's voice or the voice of other agent users. In this most ideal case, the agent recording data only corresponds to one speaker, and there is no noise, so there is no need to analyze the source of the noise. As a result, the efficiency of quality inspection of the agent recording data is very high, and it is not easy to make misjudgments.

[0025] However, in actual applications, the voices of other people besides the agent user may appear in the agent recording data, and the voices of other people are noises to the agent recording data. The noises may be the voices of customers, the voices of other agent users near the agent user, the voices of publicly played audio files near the agent user, and so on.

[0026] On the one hand, if the noise involves words that do not meet the requirements of the business side, but is mistakenly judged as a mistake made by the agent user during the quality inspection of the agent recording data, it is obviously unfair to the agent user. On the other hand, if noise often appears in the agent recording data, it will increase the workload of data processing and reduce the efficiency of quality inspection, which is not what the business side wants to see. Therefore, when noise frequently appears in the agent recording data, the business side needs to promptly perform operation and maintenance on the call center so that as little noise as possible will appear in the agent recording data collected by the call center after operation and maintenance.

[0027] Noise from the customer side and noise from the agent user side may be caused by abnormalities in different parts of the call center. Therefore, if the source of the noise is not accurately identified, for example, the noise emitted by other agent users is misjudged as noise emitted by customers, it may mislead operation and maintenance personnel and reduce subsequent operation and maintenance efficiency. In order to solve the above problem, the embodiment of the present application provides a data processing method.

[0028] Figure 1 A processing flow chart of a data processing method provided in an embodiment of the present application. Figure 1 The data processing method can be executed by an electronic device, which can be a terminal device, such as a mobile phone, a laptop computer, an intelligent interactive device, etc.; or, the electronic device can also be a server, such as an independent physical server, a server cluster, or a cloud server capable of cloud computing. Figure 1 The data processing method provided in this embodiment specifically includes steps S102 to S108.

[0029] Step S102, obtaining agent recording data; the agent recording data refers to the audio data of the agent user's speaking period when the agent user is communicating with the customer.

[0030] In specific implementation, the agent user can communicate with the customer in the form of voice during the call. The agent user is the staff who provides services to the customer. The services that the agent user can provide include but are not limited to: consulting services, complaint services, service activation / cancellation services, recommendation services, etc.

[0031] For each agent user, when the agent user provides services to any customer through a call, the call center can record the entire conversation between the agent user and the customer from the time the call is connected to the time the call is ended. When recording, the call center can save the audio data of the agent user's speaking period in one of the left and right channels, and save the audio data of the customer's speaking period in the other of the left and right channels. That is, the call center saves the sound of the agent user in one channel and the sound of the customer in another channel. The two channels are independent of each other and do not affect each other.

[0032] Acquiring the agent recording data may be acquiring pre-stored audio data of the agent user's speaking period.

[0033] The speaking period of the agent user refers to the time period during which the agent user speaks in the entire communication process between the agent user and the customer. For example, from the start to the end of the call, which is T minutes in total, the call center records the entire communication process between the agent and the customer within these T minutes, obtaining an audio file with a duration of T minutes. Among them, the time periods [t1, t2] and [t5, t6] are the speaking periods of the agent user, and the time periods [t3, t4] and [t7, t8] are the speaking periods of the customer. Then, within the T minutes, there are 2 speaking periods of the agent user and 2 speaking periods of the customer. The agent recording data refers to the audio data of the time periods [t1, t2] and [t5, t6] in the audio file with a duration of T minutes. t1, t2, t3, t4, t5, t6, t7, and t8 are all real numbers greater than 0, and t1 < t2 < t3 < t4 < t5 < t6 < t7 < t8. T is a real number greater than 0.

[0034] It should be noted that although the agent recording data refers to the audio data of the speaking period of the agent user, malfunctions in the recording function of the call center or environmental factors may cause crosstalk. In theory, there should only be the voice of the agent user in the audio data of the speaking period of the agent user, but in fact, there may also be the voices of other people besides the agent user.

[0035] Step S104, if the agent recording data corresponds to at least two speakers and the at least two speakers include the agent user and a non-agent user, then determine the target time period corresponding to the non-agent user in the agent recording data.

[0036] It should be noted that the non-agent user refers to other users except this agent user, including the customer, or it can also be other agent users adjacent to this agent user, or it can also include the voices of the users in the environment where the agent is located. The non-agent user can be understood as the user to be identified. The user to be identified needs to be identified as whether it is the customer or other agent users, so as to determine what method to use to optimize the agent recording data with crosstalk noise.

[0037] In one case, when it is determined that the agent recording data corresponds to two speakers and the two speakers are the agent user and a non-agent user respectively, determine the target time period corresponding to the non-agent user in the agent recording data.

[0038] In another case, when it is determined that the agent recording data corresponds to more than two speakers and the more than two speakers are identified as the agent user and at least two non-agent users, determine the target time period corresponding to each non-agent user in the at least two non-agent users in the agent recording data.

[0039] The target time period corresponding to the non-seat user in the seat recording data, for example, the seat recording data is 20 seconds in total, among which the seat user is speaking from the 1st to the 12th seconds, the non-seat user is speaking from the 14th to the 16th seconds, and the seat user is speaking from the 17th to the 20th seconds. Then the 14th to the 16th seconds can be determined as the target time period corresponding to the non-seat user in the seat recording data.

[0040] In a specific implementation, after acquiring the seat recording data, it is determined that the non-seat user is before the target time period corresponding to the seat recording data. The embodiment of the present application determines that the seat recording data corresponds to at least two speakers and the at least two speakers include the seat user and the non-seat user in the following manner: performing speaker segmentation processing on the seat recording data to obtain a first segmentation result; the first segmentation result includes at least one speaker identifier; when the first segmentation result includes at least two speaker identifiers, voiceprint feature extraction is performed on the seat recording data according to the first segmentation result to determine the voiceprint feature corresponding to each speaker identifier; if it is determined according to the voiceprint feature of each speaker identifier and the pre-stored seat voiceprint feature that there is a speaker identifier that matches the pre-stored seat voiceprint feature among the at least two speaker identifiers, and there is a speaker identifier that does not match the pre-stored seat voiceprint feature, then it is determined that the seat recording data corresponds to at least two speakers and the at least two speakers include the seat user and the non-seat user.

[0041] The problem that Speaker Diarization solves is "who spoke and when". Given a speech containing multiple people speaking alternately, Speaker Diarization needs to determine the start and end time of each sentence and determine who said it.

[0042] Speaker segmentation processing can be to distinguish and mark the speaking time of different speakers in the collected voice signal stream.

[0043] Speaker segmentation is performed on the seat recording data to obtain a first segmentation result. The first segmentation result may include at least one speaker identifier and may also include a time period corresponding to each speaker identifier in the at least one speaker identifier in the seat recording data.

[0044] For example, in the seat recording data, the time period [t1, t2] corresponds to speaker 1, the time period [t3, t4] corresponds to speaker 2, the time period [t5, t6] corresponds to speaker 1, and the time period [t7, t8] corresponds to speaker 2. The first segmentation result of the seat recording data includes speaker identification 1 of speaker 1, and the time periods [t1, t2] and [t5, t6] when speaker 1 speaks in the seat recording data. The first segmentation result also includes speaker identification 2 of speaker 2, and the time periods [t3, t4] and [t7, t8] when speaker 2 speaks in the seat recording data. Among them, t1, t2, t3, t4, t5, t6, t7, t8 are all real numbers greater than 0, and t1 <t2<t3<t4<t5<t6<t7<t8。

[0045] When the first segmentation result includes only one speaker identifier, it means that only the agent user speaks in the agent recording data, there is no noise in the agent recording data, and thus there is no need to identify the source of the noise. Therefore, there is no need to execute steps S104-S108 in the data processing method provided in this embodiment.

[0046] When the first segmentation result includes at least two speaker identifiers, there are at least two possibilities: (1) noise other than the seat user appears in the seat recording data; (2) no noise appears in the seat recording data, but the seat user changes his or her own timbre or the environment in which the seat user is located changes, resulting in at least one of a number of reasons causing the seat user's voice to be mistakenly identified as the voices of two speakers.

[0047] Therefore, the fact that the first segmentation result of the seat recording data includes at least two speaker identifiers does not mean that the seat recording data corresponds to at least two speakers, and it is necessary to further determine whether noise appears in the seat recording data through voiceprint features.

[0048] In a specific implementation, when the first segmentation result includes at least two speaker identifiers, the recording segment corresponding to the speaker identifier can be determined in the seat recording data based on the time period corresponding to each speaker identifier among the at least two speaker identifiers in the seat recording data, and then the recording segment is processed by voiceprint recognition to obtain the voiceprint feature corresponding to the recording segment, and the voiceprint feature is used as the voiceprint feature corresponding to the speaker identifier.

[0049] The call center may pre-store the agent voiceprint features of the agent users. After obtaining the voiceprint features corresponding to each speaker identifier, similarity calculation is performed on the voiceprint features corresponding to each speaker identifier and the pre-stored agent voiceprint features to obtain the voiceprint similarity between the voiceprint features corresponding to each speaker identifier and the agent voiceprint features.

[0050] The call center may be pre-configured with a voiceprint similarity threshold for voiceprint matching and a preset similarity threshold for distinguishing different speakers. The voiceprint similarity threshold and the preset similarity threshold may have different numerical values.

[0051] For example, the voiceprint similarity threshold is 0.7, and the preset similarity threshold is 0.2. For each speaker identifier: if the voiceprint similarity between the voiceprint feature corresponding to the speaker identifier and the voiceprint feature of the seat is greater than or equal to the voiceprint similarity threshold, it is determined that the voiceprint feature corresponding to the speaker identifier matches the voiceprint feature of the seat; if the voiceprint similarity between the voiceprint feature corresponding to the speaker identifier and the voiceprint feature of the seat is less than the voiceprint similarity threshold, it is determined that the voiceprint feature corresponding to the speaker identifier does not match the voiceprint feature of the seat.

[0052] Taking the first segmentation result including x speaker identifiers as an example (x is a natural number greater than 1), after voiceprint recognition and similarity calculation, the following situations may occur:

[0053] (1) If, among the x speaker identifiers, the x speaker identifiers are all speaker identifiers whose voiceprint features match the pre-stored seat voiceprint features, it is determined that the seat recording data corresponds to one speaker and the speaker is a seat user.

[0054] (2) If, among the x speaker identifiers, y speaker identifiers are all speaker identifiers whose voiceprint features match the pre-stored seat voiceprint features, and the voiceprint similarity between the voiceprint features corresponding to each of the z speaker identifiers other than the y speaker identifiers and the pre-stored seat voiceprint features is less than a preset similarity threshold, it is determined that the seat recording data corresponds to at least two speakers and the at least two speakers include a seat user and a non-seat user. y is a natural number greater than or equal to 1, z is a natural number greater than or equal to 1, and x≥y+z.

[0055] (3) If, among the x speaker identifiers, y speaker identifiers are all speaker identifiers whose voiceprint features match the pre-stored seat voiceprint features, and the voiceprint similarity between the voiceprint features corresponding to each of the z speaker identifiers other than the y speaker identifiers and the pre-stored seat voiceprint features is greater than or equal to a preset similarity threshold, it is determined that the seat recording data corresponds to one speaker and the speaker is a seat user. y is a natural number greater than or equal to 1, z is a natural number greater than or equal to 1, and x≥y+z.

[0056] (4) If there is no speaker identifier whose voiceprint feature matches the pre-stored agent voiceprint feature among the x speaker identifiers, a notification message is generated, and the notification message is used to notify the call center to perform operation and maintenance.

[0057] The various situations listed above are merely exemplary. Among them, (3) may be caused by the following reasons: the seat user deliberately changed their voice during communication, or the environment where the seat user is located changed. (4) may be caused by the following reasons: there are significant changes in the voice of the current seat user compared to the voice of the seat user when the seat voiceprint feature was recorded in the call center. For example, the seat user's voice became hoarse due to illness. Or, the call center accidentally performed voiceprint matching on the seat recording data of Agent A with the seat voiceprint feature of Agent B.

[0058] In the case where the seat recording data corresponds to at least two speakers and at least two speakers include the seat user and a non-seat user, at least two speaker identifiers in the first segmentation result of the seat recording data include the first user identifier of the seat user and the second user identifier of the non-seat user.

[0059] The first segmentation result further includes the time period corresponding to each speaker identifier in the at least two speaker identifiers in the seat recording data.

[0060] Determining the target time period corresponding to the non-seat user in the seat recording data may be to determine the target time period corresponding to the non-seat user in the seat recording data according to the second user identifier and the time period corresponding to each speaker identifier in the seat recording data.

[0061] Step S106, obtaining customer recording data; the customer recording data refers to the audio data of the speaking period of the customer when the seat user communicates with the customer.

[0062] The customer recording data corresponds to the same call as the seat recording data. The speaking period of the customer refers to the time period during which the customer speaks during the entire communication process between the seat user and the customer.

[0063] For example, from the start to the end of the call, it is a total of T minutes. The call center records the entire communication process between the seat and the customer within these T minutes, obtaining an audio file with a time length of T minutes. Among them, the time periods [t1, t2] and [t5, t6] are the speaking time periods of the seat user, and the time periods [t3, t4] and [t7, t8] are the speaking time periods of the customer. Then, within the T minutes, there are 2 speaking time periods of the seat user and 2 speaking time periods of the customer. The customer recording data refers to the audio data of the time periods [t3, t4] and [t7, t8] in the audio file with a time length of T minutes. t1, t2, t3, t4, t5, t6, t7, t8 are all real numbers greater than 0, and t1 < t2 < t3 < t4 < t5 < t6 < t7 < t8. T is a real number greater than 0.

[0064] It should be noted that although customer recording data refers to the audio data during the customer's speaking period, a malfunction of the call center's recording function or environmental factors may cause crosstalk. In theory, the audio data during the customer's speaking period only contains the customer's voice, but in reality, there may also be voices of other people besides the customer.

[0065] In a specific implementation method, the method of obtaining agent recording data and obtaining customer recording data can be: obtaining call recording data; the call recording data is obtained by collecting audio data generated during the communication process through different channels when the agent user communicates with the customer; the call recording data is split according to the channel to obtain the recording data of the agent channel and the recording data of the customer channel; the recording data of the agent channel refers to the agent recording data, and the recording data of the customer channel refers to the customer recording data.

[0066] Call recording data is an audio file obtained by collecting audio data generated during the communication process between the agent user and the customer through different channels. The audio file includes both the agent user's voice and the customer's voice.

[0067] In specific implementation, the call recording data can be downloaded to the local computer through the recording management system of the call center, and the call recording file can be split into the recording file of the agent channel and the recording file of the agent channel using the channel splitting tool. The recording file of the agent channel includes the agent recording data, and the recording file of the customer channel includes the customer recording data.

[0068] Step S108, determining whether the customer's audio data is inserted into the agent recording data according to the overlap rate between the target time period and the customer's speaking time period, and the similarity between the voiceprint of the non-agent user and the customer's voiceprint.

[0069] If it is determined that the customer's audio data is intertwined in the agent recording data, then tag information of the customer's intertwined audio data may be added to the agent recording data to facilitate subsequent data analysis and / or operation and maintenance processing.

[0070] In addition, considering that the customer recording data may include only the customer's voice, or may include the customer's voice and other noises mixed into the customer recording data, before executing step S108, the customer recording data may be processed for speaker segmentation to obtain a second segmentation result.

[0071] If the second segmentation result includes only one speaker identifier and the time period corresponding to the speaker identifier in the customer recording data, the speaker identifier is determined as the third user identifier of the customer, and the customer's speaking time period is determined based on the customer recording data. The serial numbers such as "first", "second", and "third" appearing in this specification are only for distinguishing similar features and have no actual meaning, and will not be repeated below.

[0072] If the second segmentation result includes at least two speaker identifiers and the time period corresponding to each speaker identifier in the customer's recording data, the total duration of the time period corresponding to each speaker identifier can be sorted according to the numerical value, and the speaker identifier with the largest total duration value is determined as the customer's third user identifier. According to the third user identifier and the time period corresponding to each speaker identifier in the customer's recording data, the customer's speaking time period is determined, and the customer's speaking time period refers to the speaking time period that only includes the customer's voice. The customer's voiceprint can be determined in the following way:

[0073] When the second segmentation result only includes the third user identifier, voiceprint recognition processing is performed on the customer recording data to obtain the customer's voiceprint.

[0074] When the second segmentation result includes at least two speaker identifiers, the customer recording data is cropped and spliced ​​according to the third user identifier to obtain customer audio data excluding noise, and voiceprint recognition is performed on the customer audio data to obtain the customer's voiceprint.

[0075] The voiceprint of non-agent users can be determined in the following ways:

[0076] According to the target time period, the seat recording data is cropped to obtain the audio data to be identified that only includes the voices of non-seat users. The audio data to be identified is processed for voiceprint recognition to obtain the voiceprints of non-seat users.

[0077] In a specific implementation, before determining whether the customer's audio data is inserted into the seat recording data based on the overlap rate between the target time period and the customer's speaking time period, and the similarity between the voiceprint of the non-seat user and the voiceprint of the customer, the data processing method also includes: determining the start time point of the target time period as the first time point, and determining the end time point of the target time period as the second time point; determining the overlap time period in the customer's speaking time period based on the first time point and the second time point; and determining the overlap rate between the target time period and the customer's speaking time period based on the duration of the overlap time period and the duration of the target time period.

[0078] When the customer is speaking, even if there is crosstalk, it is impossible for the customer's recorded voice data not to capture the customer's voice while the voice of the customer is mixed into the agent's recorded voice data at the same time. Therefore, if there are noises made by someone other than the agent during the target time period, and the customer is not speaking at all during that target time period, or the overlapping part between the customer's speaking period and the target time period is very small, it can be shown that the noise is not made by the customer. For this reason, it is necessary to calculate the coincidence rate between the target time period and the customer's speaking period to rule out the situation where the audio data of the customer cannot be mixed into the agent's recorded voice data in part of the agent's recorded voice data.

[0079] In addition, considering the actual application, if crosstalk occurs, it is very likely that only a small amount of the voice of someone other than the agent is mixed into the agent's recorded voice data. Usually, the time length of the target time period is much shorter than the customer's speaking period. Therefore, it is not meaningful to directly calculate the ratio between the target time period and the customer's speaking period. It is necessary to first determine the overlapping part between the target time period and the customer's speaking period, that is, the overlapping time period, and then determine the coincidence rate according to the overlapping time period and the target time period.

[0080] For example, if the target time period is the time period [T1, T2], then the first time point is T1, the second time point is T2, the customer's speaking period includes the time period [T3, T4], and T3 < T1 < T2 < T4, where T1, T2, T3, and T4 are all real numbers greater than 0. According to the first time point T1 and the second time point T2, the overlapping time period can be determined as the time period [T1, T2] within the customer's speaking period.

[0081] Another example, if the target time period is the time period [T1, T2], then the first time point is T1, the second time point is T2, the customer's speaking period includes the time period [T3, T4], and T1 < T3 < T2 < T4, where T1, T2, T3, and T4 are all real numbers greater than 0. According to the first time point T1 and the second time point T2, the overlapping time period can be determined as [T3, T2] within the customer's speaking period.

[0082] Based on the duration of the overlapping time period and the duration of the target time period, determine the coincidence rate between the target time period and the customer's speaking period. It can be to calculate the ratio between the duration of the overlapping time period and the duration of the target time period to obtain the coincidence rate.

[0083] For example, the customer's speaking time period includes 1-20s, 25-35s, and 44-50s. The target time period is 25-30s, then the 25th second corresponds to the first time point, and the 30th second corresponds to the second time point. The duration of the target time period is 5s. According to the first time point and the second time point, the overlapping time period is determined to be 25-30s in the customer's speaking time period. The duration of the overlapping time period is 5s. According to the duration of the overlapping time period and the duration of the target time period, it is determined that the overlap rate between the target time period and the customer's speaking time period is 100%.

[0084] For another example, the client's speaking time periods include 1-20s, 25-35s, and 44-50s. The target time period is 23-26s, then the 23th second corresponds to the first time point, and the 26th second corresponds to the second time point. The duration of the target time period is 3s. According to the first time point and the second time point, the overlapping time period is determined to be 25-26s in the client's speaking time period. The duration of the overlapping time period is 1s.

[0085] According to the length of the overlapping time period and the length of the target time period, it is determined that the overlap rate between the target time period and the customer's speaking time period is 33.3%. In a specific implementation method, according to the overlap rate between the target time period and the customer's speaking time period, and the similarity between the voiceprint of the non-seat user and the customer's voiceprint, it is determined whether the customer's audio data is inserted into the agent recording data, including: comparing the overlap rate with the preset overlap rate threshold to obtain a first comparison result; comparing the similarity with the preset similarity threshold to obtain a second comparison result; when the first comparison result is used to indicate that the overlap rate is greater than the preset overlap rate threshold, and the second comparison result is used to indicate that the similarity is greater than the voiceprint similarity threshold, it is determined that the customer's audio data is inserted into the agent recording data.

[0086] The first comparison result may indicate that the overlap rate is greater than the preset overlap rate threshold, or may indicate that the overlap rate is less than or equal to the preset overlap rate threshold. The second comparison result may indicate that the similarity is greater than the voiceprint similarity threshold, or may indicate that the similarity is less than or equal to the voiceprint similarity threshold. Each threshold in this embodiment can be freely set as required.

[0087] In actual business scenarios, the customer's voiceprint is relatively easy to determine. The voiceprint of the customer that may be inserted into the agent's recording data during a call can only be the voiceprint of the customer who is communicating with the agent user. The voiceprints of other agents that may be inserted into the agent's recording data are relatively difficult to determine compared with the customer, because there may be multiple other agents sitting next to the agent user, and the main purpose of voiceprint comparison is to determine whether the noise comes from the customer side or the agent side. When the noise comes from the agent side, which agent the noise comes from does not affect subsequent data analysis or operation and maintenance, so there is no need to determine which agent user near the agent user has his voice inserted into the agent's recording data.

[0088] Therefore, in practical applications, in order to improve data processing efficiency, it can be assumed that the noise comes from either the customer or other seat users. When the seat recording data corresponds to at least two speakers, if the pre-configured judgment condition is met, it is determined that the customer's audio data is inserted into the seat recording data, otherwise it is determined that the audio data of other seat users is inserted into the seat recording data.

[0089] The recognition priority of customer crosstalk is higher than that of agent crosstalk because customers who are not bound by the business party are much more likely to say things that do not meet the business party's requirements during communication than other agent users who are bound by the business party. Therefore, customer crosstalk is more likely to trigger a misjudgment and cause the business party to punish the agent user. Therefore, in the event of customer crosstalk, the call center needs to operate and maintain as soon as possible.

[0090] When the first comparison result is used to indicate that the overlap rate is greater than the preset overlap rate threshold, and the second comparison result is used to indicate that the similarity is greater than the voiceprint similarity threshold, the pre-configured judgment condition is met, and it can be determined that the customer's audio data is inserted into the agent's recording data.

[0091] That is, the pre-configured judgment conditions include: the time period overlap rate is greater than the preset overlap rate threshold, and the similarity is greater than the voiceprint similarity threshold.

[0092] When the first comparison result is used to indicate that the overlap rate is less than or equal to the preset overlap rate threshold, and the second comparison result is used to indicate that the similarity is greater than the voiceprint similarity threshold, the pre-configured judgment condition is not met, and it can be determined that audio data of other seat users or audio data publicly played near the seat user is inserted into the seat recording data.

[0093] When the first comparison result is used to indicate that the overlap rate is greater than the preset overlap rate threshold, and the second comparison result is used to indicate that the similarity is less than or equal to the voiceprint similarity threshold, the pre-configured judgment condition is not met, and it can be determined that audio data of other seat users or audio data publicly played near the seat user is inserted into the seat recording data.

[0094] When the first comparison result is used to indicate that the overlap rate is less than or equal to the preset overlap rate threshold, and the second comparison result is used to indicate that the similarity is less than or equal to the voiceprint similarity threshold, the pre-configured judgment condition is not met, and it can be determined that audio data of other seat users or audio data publicly played near the seat user is inserted into the seat recording data.

[0095] By comparing voiceprints and calculating the overlap rate, it is possible to confirm whether a non-agent user is a customer from two different dimensions, thereby improving recognition accuracy when identifying the source of noise in agent recording data.

[0096] In a specific implementation, the data processing method also includes: if it is determined that the customer's audio data is inserted into the agent recording data, the customer's audio data in the agent recording data is deleted to obtain the agent recording data to be quality inspected; based on the quality inspection rules, the agent recording data to be quality inspected is quality inspected to obtain a quality inspection result, and the quality inspection result is used to indicate whether the communication between the agent user and the customer meets the quality requirements or does not meet the quality requirements; when it is determined according to the quality inspection result that the communication between the agent user and the customer does not meet the quality requirements, a reminder message is output to the agent user.

[0097] The customer's audio data in the agent recording data is deleted to obtain the agent recording data to be inspected. The agent recording data can be cropped and spliced ​​according to the target time period to obtain the agent recording data to be inspected. The agent recording data to be inspected does not include the audio data of the target time period in the agent recording data.

[0098] Based on the quality inspection rules, the quality inspection agent recording data is quality inspected to obtain the quality inspection results, which are used to indicate whether the communication between the agent user and the customer meets the quality requirements or does not meet the quality requirements. The quality inspection rules can reflect the requirements of the business party for the agent user in communication.

[0099] For example, each time it is detected that the agent user's speech includes uncivilized language, the agent user's evaluation score is negatively updated, and each time it is detected that the agent user's speech includes preset high-quality language, the agent user's evaluation score is positively updated. The quality inspection result can be the agent user's evaluation score.

[0100] After quality inspection is performed on the quality inspection agent recording data based on the quality inspection rules and the agent user's evaluation score is obtained, it can be determined whether the communication between the agent user and the customer meets the quality requirements or does not meet the quality requirements based on the numerical comparison result of the evaluation score and the preset score threshold.

[0101] Customers are not bound by the business party. If audio data of customers using uncivilized language is inserted into the agent recording data when the agent user communicates with the customer, it may cause misjudgment when quality inspection is performed based on the agent recording data. Therefore, it is necessary to first delete the audio data of the customer inserted into the agent recording data, and then perform quality inspection to obtain accurate quality inspection results, which objectively reflect whether the agent user's communication meets the quality requirements.

[0102] In a specific implementation, the data processing method also includes: if it is determined that the customer's audio data is interlaced in the agent's recording data, then outputting prompt information of customer-type crosstalk, which is used to prompt the use of an optimization method for customer-type crosstalk to optimize the recording process; if it is determined that the customer's audio data is not interlaced in the agent's recording data, then outputting prompt information of inter-agent crosstalk, which is used to prompt the use of an optimization method for inter-agent crosstalk to optimize the recording process.

[0103] When noise appears in the agent recording data, the source of the noise may include the customer side and the agent side. The noise from the customer side is the customer's audio data. For example, the agent user who is communicating with the customer is agent user 1, and the customer's voice is inserted into the agent recording data.

[0104] The noise from the seat side may be the audio data of other seat users near the seat user. For example, the seat user who is communicating with the customer is seat user 1. Seat user 2, seat user 3 and seat user 4 are sitting in front, behind and on the left and right of user 1. The voice of seat user 4 is inserted into the seat recording data. The noise from the seat side may also be the audio data played by the audio broadcasting device near the seat user.

[0105] For example, the agent user who is communicating with the customer is agent user 1. Within 10 meters of agent user 1, there is a speaker connected to a computer that is playing audio data of a short video. The speaking and laughing sounds in the short video are inserted into the agent recording data.

[0106] Within the preset time range, if the number of times the noise occurs exceeds the preset number threshold, or the duration of the noise is greater than the preset time threshold, it means that there is probably a problem with the recording function of the call center, and the operation and maintenance personnel are required to investigate the cause and perform operation and maintenance so that the call center can return to normal after operation and maintenance.

[0107] Noise coming from the customer side may be caused by an abnormality in a part of the call center's system or equipment, and noise coming from the agent side may be caused by an abnormality in another part of the call center's system or equipment. If the customer's audio data is inserted into the agent's recording data, it is mistakenly identified as noise on the agent's side, or if non-customer audio data is inserted into the agent's recording data, it is mistakenly identified as noise on the customer's side, which may mislead operation and maintenance personnel and reduce operation and maintenance efficiency. Therefore, after determining that the customer's audio data is inserted into the agent's recording data, a prompt message of customer-type crosstalk can be output, and the prompt message of customer-type crosstalk is used to prompt the optimization method of customer-type crosstalk to be used to optimize the recording process.

[0108] When it is determined that no customer audio data is interspersed in the agent recording data, that is, when it is determined that non-customer audio data is interspersed in the agent recording data, prompt information of customer-type crosstalk is output. The prompt information of customer-type crosstalk is used to prompt the optimization method of customer-type crosstalk to be adopted to optimize the recording process.

[0109] By outputting different types of prompt information, the operation and maintenance efficiency can be improved, allowing the operation and maintenance personnel to optimize the recording process as quickly as possible, thereby minimizing the noise in the agent recording data collected during the communication process between the agent user and the customer, and improving data processing efficiency.

[0110] In such Figure 1In the illustrated embodiment, first, the agent recording data is obtained; the agent recording data refers to the audio data of the agent user's speaking period when the agent user is communicating with the customer; if the agent recording data corresponds to at least two speakers and the at least two speakers include the agent user and the non-seat user, then the target time period corresponding to the non-seat user in the agent recording data is determined; then, the customer recording data is obtained; the customer recording data refers to the audio data of the customer's speaking period when the agent user is communicating with the customer; finally, based on the overlap rate between the target time period and the customer's speaking period, and the similarity between the voiceprint of the non-seat user and the voiceprint of the customer, it is determined whether the customer's audio data is inserted into the agent recording data. In this way, when it is determined that the agent recording data corresponds to at least two speakers, it can be determined that there is noise mixed in the agent recording data. The noise may come from the customer or from other agent users located next to the agent user. Different sources of noise may be caused by different abnormalities. Accurately identifying the source of noise is beneficial to subsequent operation and maintenance. In addition, the purpose of storing the agent recording data and the customer recording data separately includes quality inspection of the agent recording data to supervise the service quality of the agent user. Considering that the quality inspection result of the agent recording data is more likely to be affected when the customer is misjudged as an agent user, it is more important to determine whether the noise comes from the customer. By analyzing the overlap rate between the target time period corresponding to the non-agent user who emits the noise in the agent recording data and the customer's speaking time period, as well as the similarity between the voiceprint of the non-agent user and the voiceprint of the customer, it can be determined from multiple dimensions whether the customer's recording data is inserted into the agent recording data, thereby improving the accuracy of identifying the source of the noise.

[0111] Figure 2 A flow chart of another data processing method provided in an embodiment of the present application. Figure 2 As shown:

[0112] Step S202: The seat channel calls the speech recognition interface to start detection.

[0113] Seat channels can refer to Figure 1 The corresponding description part in the embodiments.

[0114] The speech recognition interface may be an ASR (Automatic Speech Recognition, speech recognition technology) interface. The purpose of speech recognition technology is to convert the vocabulary content in human speech into readable text.

[0115] In specific implementation, the recording management system can download the call recording data to the local computer, and use ffmpeg to split the call recording data into two recording files for the agent channel and the customer channel.

[0116] This step can refer to Figure 1The description of step S102 in the embodiment is not repeated here.

[0117] Step S204: Check whether two persons are obtained by segmentation.

[0118] In specific implementation, the recording management system can call the ASR system to identify the recording files of the seat channels and detect whether there are two people speaking.

[0119] If yes, then step S206 is executed. If no, it means that the audio data involved in the detection does not involve crosstalk.

[0120] Step S206: Call the voiceprint recognition service for each of the two voices to calculate the similarity.

[0121] When two people are detected talking, the voiceprint recognition system is called to obtain the voiceprint recognition result.

[0122] Step S208: the similarity is less than a preset similarity threshold.

[0123] Determine whether the similarity between the voices of two people is less than a preset similarity threshold.

[0124] The similarity threshold can be referred to Figure 1 The corresponding description part of the preset similarity threshold in the embodiment.

[0125] For example, compare the voiceprint similarity to see if it is less than 0.2. If the similarity is less than 0.2, it proves that there are two different people speaking.

[0126] If yes, then step S210 is executed. If no, it means that the two people segmented are misjudged, and actually the seat audio data involved in the detection only includes the voice of one seat user.

[0127] Step S210, recording the speaking time of the non-seat speaker.

[0128] This step can refer to Figure 1 Step S104 in the embodiment.

[0129] The non-seat speaker is the aforementioned non-seat user. The time when the non-seat speaker speaks is the aforementioned target time period.

[0130] Step S212: The client channel calls the speech recognition interface to start detection.

[0131] Customer voice can refer to Figure 1 The corresponding description part in the embodiments.

[0132] This step can refer to Figure 1 Step S106 in the embodiment.

[0133] like Figure 2As shown, after step S212 is executed, step S214 and step S216 may be executed simultaneously.

[0134] Step S214, calculating the temporal overlap rate according to the speaker segmentation result and the seat segmentation result.

[0135] Step S216, calculating the similarity of the voiceprint according to the speaker segmentation result and the seat segmentation result.

[0136] Step S218: the overlap rate is greater than 80%.

[0137] Determine whether the overlap rate is greater than 80%.

[0138] The recording management system can call the ASR system to identify the recording files of the customer's voice channel, compare the speaking time with the speaking time of the non-agent person identified by the agent's voice channel, and determine whether there is overlap in time. If the overlap rate is greater than 80%, it proves that there are two people other than the agent speaking at the same time.

[0139] Step S220: The similarity is greater than the voiceprint similarity threshold.

[0140] Determine whether the similarity is greater than the voiceprint similarity threshold.

[0141] For example, the recording management system can call the voiceprint recognition system to compare the recording of the customer's voice channel with the recording of the time when the agent's voice channel was recognized as the non-agent's voice, and determine whether the voiceprint similarity is greater than 0.8. If the similarity is greater than 0.8, it proves that it is the customer's voice.

[0142] When the judgment results of step S218 and step S220 are both yes, step S222 is executed.

[0143] Through multiple detection results, it is possible to determine whether the customer's voice has been crosstalked into the agent's channel, thereby greatly improving the accuracy of customer crosstalk detection in a variety of special scenarios.

[0144] Step S222: Mark the audio as customer crosstalk.

[0145] Since the technical concept is the same, the description in this embodiment is relatively simple, and the relevant parts may refer to the corresponding description of the method embodiment provided above.

[0146] In the above-mentioned embodiment, a data processing method is provided. Correspondingly, based on the same technical concept, the embodiment of the present application also provides a data processing device, which is described below in conjunction with the accompanying drawings.

[0147] Figure 3 Schematic diagram of a data processing device provided in an embodiment of the present application. This embodiment provides a data processing device 300, including:

[0148] The acquisition unit 302 is used to acquire the seat recording data; the seat recording data refers to the audio data of the seat user's speaking period when the seat user is communicating with the customer;

[0149] A determination unit 304 is configured to determine a target time period corresponding to the non-seat user in the seat recording data if the seat recording data corresponds to at least two speakers and the at least two speakers include the seat user and the non-seat user;

[0150] The acquisition unit 302 is further used to acquire customer recording data; the customer recording data refers to the audio data of the customer's speaking period when the agent user is communicating with the customer;

[0151] The determination unit 304 is further configured to determine whether the audio data of the customer is included in the agent recording data according to the overlap rate between the target time period and the speaking time period of the customer, and the similarity between the voiceprint of the non-agent user and the voiceprint of the customer.

[0152] Optionally, the data processing device 300 further includes:

[0153] A segmentation unit, configured to perform speaker segmentation processing on the seat recording data to obtain a first segmentation result; the first segmentation result includes at least one speaker identifier;

[0154] The determining unit 304 is further configured to extract voiceprint features from the seat recording data according to the first segmentation result to determine the voiceprint features corresponding to each speaker identifier when the first segmentation result includes at least two speaker identifiers;

[0155] If it is determined based on the voiceprint features of each speaker identifier and the pre-stored seat voiceprint features that there is a speaker identifier that matches the pre-stored seat voiceprint features among the at least two speaker identifiers, and there is a speaker identifier that does not match the pre-stored seat voiceprint features, then it is determined that the seat recording data corresponds to at least two speakers and the at least two speakers include a seat user and a non-seat user.

[0156] Optionally, the determining unit 304 is further configured to:

[0157] Comparing the overlap rate with a preset overlap rate threshold to obtain a first comparison result;

[0158] Comparing the similarity with a preset similarity threshold to obtain a second comparison result;

[0159] When the first comparison result is used to indicate that the overlap rate is greater than a preset overlap rate threshold, and the second comparison result is used to indicate that the similarity is greater than a voiceprint similarity threshold, it is determined that the audio data of the customer is inserted into the agent recording data.

[0160] Optionally, the determining unit 304 is further configured to:

[0161] Determine the start time point of the target time period as a first time point, and determine the end time point of the target time period as a second time point;

[0162] Determining an overlapping time period in the client's speaking period according to the first time point and the second time point;

[0163] The overlap rate between the target time period and the speaking time period of the client is determined according to the duration of the overlap time period and the duration of the target time period.

[0164] Optionally, the acquiring unit 302 is further configured to:

[0165] Acquire call recording data; the call recording data is obtained by collecting audio data generated during the communication process through different channels when the agent user communicates with the customer;

[0166] The data processing device 300 further includes:

[0167] The splitting unit is used to split the call recording data according to channels to obtain recording data of the seat channel and recording data of the customer channel; the recording data of the seat channel refers to the seat recording data, and the recording data of the customer channel refers to the customer recording data.

[0168] Optionally, the data processing device 300 further includes:

[0169] a deleting unit, configured to delete the audio data of the customer from the agent recording data if it is determined that the audio data of the customer is inserted into the agent recording data, so as to obtain the agent recording data to be inspected;

[0170] A quality inspection unit, configured to perform quality inspection on the seat recording data to be inspected based on quality inspection rules to obtain a quality inspection result, wherein the quality inspection result is used to indicate whether the communication between the seat user and the customer meets the quality requirements or does not meet the quality requirements;

[0171] When it is determined according to the quality inspection result that the communication between the agent user and the customer does not meet the quality requirements, a reminder message is output to the agent user.

[0172] Optionally, the data processing device 300 further includes:

[0173] An output unit, for outputting a prompt message of customer-type crosstalk if it is determined that the audio data of the customer is crosstalked into the agent recording data, for prompting to optimize the recording process by adopting an optimization method of customer-type crosstalk;

[0174] The output unit is further configured to output a prompt message of crosstalk between agents if it is determined that the audio data of the customer is not interlaced in the agent recording data, so as to prompt the recording process to be optimized by adopting an optimization method of crosstalk between agents.

[0175] The data processing device provided in the embodiment of the present application includes: an acquisition unit, which is used to acquire seat recording data; the seat recording data refers to the audio data of the seat user's speaking period when the seat user is communicating with the customer; a determination unit, which is used to determine the target time period corresponding to the non-seat user in the seat recording data if the seat recording data corresponds to at least two speakers and the at least two speakers include the seat user and the non-seat user; the acquisition unit is also used to acquire customer recording data; the customer recording data refers to the audio data of the customer's speaking period when the seat user is communicating with the customer; the determination unit is also used to determine whether the customer's audio data is inserted into the seat recording data based on the overlap rate between the target time period and the customer's speaking period, and the similarity between the voiceprint of the non-seat user and the voiceprint of the customer. In this way, when it is determined that the agent recording data corresponds to at least two speakers, it can be determined that there is noise mixed in the agent recording data. The noise may come from the customer or from other agent users located next to the agent user. Different sources of noise may be caused by different abnormalities. Accurately identifying the source of noise is beneficial to subsequent operation and maintenance. In addition, the purpose of storing the agent recording data and the customer recording data separately includes quality inspection of the agent recording data to supervise the service quality of the agent user. Considering that the quality inspection result of the agent recording data is more likely to be affected when the customer is misjudged as an agent user, it is more important to determine whether the noise comes from the customer. By analyzing the overlap rate between the target time period corresponding to the non-agent user who emits the noise in the agent recording data and the customer's speaking time period, as well as the similarity between the voiceprint of the non-agent user and the voiceprint of the customer, it can be determined from multiple dimensions whether the customer's recording data is inserted into the agent recording data, thereby improving the accuracy of identifying the source of the noise.

[0176] Corresponding to the data processing method described above, based on the same technical concept, an embodiment of the present application further provides an electronic device, which is used to execute the data processing method provided above. Figure 4 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application.

[0177] like Figure 4As shown, the electronic device may have relatively large differences due to different configurations or performances, and may include one or more processors 401 and memory 402, and the memory 402 may store one or more storage applications or data. Among them, the memory 402 can be a temporary storage or a permanent storage. The application stored in the memory 402 may include one or more modules (not shown in the figure), and each module may include a series of computer executable instructions in the electronic device. Furthermore, the processor 401 can be configured to communicate with the memory 402 to execute a series of computer executable instructions in the memory 402 on the electronic device. The electronic device may also include one or more power supplies 403, one or more wired or wireless network interfaces 404, one or more input / output interfaces 405, one or more keyboards 406, etc.

[0178] In a specific embodiment, the electronic device includes a memory and one or more programs, wherein the one or more programs are stored in the memory, and the one or more programs may include one or more modules, and each module may include a series of computer executable instructions for the electronic device, and the one or more programs configured to be executed by one or more processors include the following computer executable instructions:

[0179] Acquire agent recording data; the agent recording data refers to the audio data of the agent user's speaking period when the agent user is communicating with the customer;

[0180] If the seat recording data corresponds to at least two speakers and the at least two speakers include the seat user and a non-seat user, determining a target time period corresponding to the non-seat user in the seat recording data;

[0181] Acquire customer recording data; the customer recording data refers to the audio data of the customer's speaking period when the agent user is communicating with the customer;

[0182] Whether the audio data of the customer is inserted into the seat recording data is determined according to the overlap rate between the target time period and the speaking time period of the customer, and the similarity between the voiceprint of the non-seat user and the voiceprint of the customer.

[0183] An embodiment of a computer-readable storage medium provided in this specification is as follows:

[0184] Corresponding to the data processing method described above, based on the same technical concept, an embodiment of the present application also provides a computer-readable storage medium.

[0185] The computer-readable storage medium provided in this embodiment is used to store computer-executable instructions. When the computer-executable instructions are executed by a processor, the following process is implemented:

[0186] Acquire agent recording data; the agent recording data refers to the audio data of the agent user's speaking period when the agent user is communicating with the customer;

[0187] If the seat recording data corresponds to at least two speakers and the at least two speakers include the seat user and a non-seat user, determining a target time period corresponding to the non-seat user in the seat recording data;

[0188] Acquire customer recording data; the customer recording data refers to the audio data of the customer's speaking period when the agent user is communicating with the customer;

[0189] Whether the audio data of the customer is inserted into the seat recording data is determined according to the overlap rate between the target time period and the speaking time period of the customer, and the similarity between the voiceprint of the non-seat user and the voiceprint of the customer.

[0190] It should be noted that the embodiments of the computer-readable storage medium in this specification and the embodiments of the data processing method in this specification are based on the same inventive concept, so the specific implementation of this embodiment can refer to the implementation of the corresponding method mentioned above, and the repeated parts will not be repeated.

[0191] The above is a description of a specific embodiment of the specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0192] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems or computer program products. Therefore, the embodiments of the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, this specification may adopt the form of a computer program product implemented on one or more computer-readable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0193] This specification is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of this specification. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable device to generate a machine, so that the instructions executed by the processor of the computer or other programmable device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0194] These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable device to work in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0195] These computer program instructions may also be loaded onto a computer or other programmable device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for executing on the computer or other programmable device to implement the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0196] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0197] The memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0198] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0199] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0200] Embodiments of the present application may be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. One or more embodiments of the present specification may also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communications network. In a distributed computing environment, program modules may be located in local and remote computer storage media, including storage devices.

[0201] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0202] The above description is only an embodiment of this document and is not intended to limit this document. For those skilled in the art, this document may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of this document should be included in the scope of the claims of this document.

Claims

1. A data processing method, characterized in that: include: Get the agent recording data; The agent recording data refers to the audio data of the agent user's speaking period when the agent user is communicating with the customer; If the seat recording data corresponds to at least two speakers and the at least two speakers include the seat user and a non-seat user, determining a target time period corresponding to the non-seat user in the seat recording data; Acquire customer recording data; the customer recording data refers to the audio data of the customer's speaking period when the agent user is communicating with the customer; Whether the audio data of the customer is inserted into the seat recording data is determined according to the overlap rate between the target time period and the speaking time period of the customer, and the similarity between the voiceprint of the non-seat user and the voiceprint of the customer.

2. The method according to claim 1, characterized in that After acquiring the seat recording data, before determining the target time period corresponding to the non-seat user in the seat recording data, the method further includes: Perform speaker segmentation processing on the seat recording data to obtain a first segmentation result; the first segmentation result includes at least one speaker identifier; When the first segmentation result includes at least two speaker identifiers, performing voiceprint feature extraction on the seat recording data according to the first segmentation result to determine the voiceprint feature of each speaker identifier; If it is determined based on the voiceprint features of each speaker identifier and the pre-stored seat voiceprint features that there is a speaker identifier that matches the pre-stored seat voiceprint features among the at least two speaker identifiers, and there is a speaker identifier that does not match the pre-stored seat voiceprint features, then it is determined that the seat recording data corresponds to at least two speakers and the at least two speakers include a seat user and a non-seat user.

3. The method according to claim 1, characterized in that The determining whether the audio data of the customer is inserted into the seat recording data according to the overlap rate between the target time period and the speaking time period of the customer and the similarity between the voiceprint of the non-seat user and the voiceprint of the customer comprises: Comparing the overlap rate with a preset overlap rate threshold to obtain a first comparison result; Comparing the similarity with a preset similarity threshold to obtain a second comparison result; When the first comparison result indicates that the overlap rate is greater than a preset overlap rate threshold, and the second comparison result indicates that the similarity is greater than a voiceprint similarity threshold, it is determined that the audio data of the customer is inserted into the agent recording data.

4. The method according to claim 1, characterized in that: Before determining whether the seat recording data includes the audio data of the customer based on the overlap rate between the target time period and the speaking time period of the customer and the similarity between the voiceprint of the non-seat user and the voiceprint of the customer, the method further includes: Determine the start time point of the target time period as a first time point, and determine the end time point of the target time period as a second time point; Determining an overlapping time period in the client's speaking period according to the first time point and the second time point; The overlap rate between the target time period and the speaking time period of the client is determined according to the duration of the overlap time period and the duration of the target time period.

5. The method according to claim 1, characterized in that Before obtaining the seat recording data, the method further includes: Acquire call recording data; the call recording data is obtained by collecting audio data generated during the communication process through different channels when the agent user communicates with the customer; The call recording data is split according to channels to obtain recording data of an agent channel and recording data of a customer channel; the recording data of the agent channel refers to the agent recording data, and the recording data of the customer channel refers to the customer recording data.

6. The method according to claim 1, characterized in that The method further comprises: If it is determined that the audio data of the customer is inserted into the agent recording data, the audio data of the customer in the agent recording data is deleted to obtain the agent recording data to be inspected; Performing quality inspection on the seat recording data to be inspected based on the quality inspection rules to obtain a quality inspection result, wherein the quality inspection result is used to indicate whether the communication between the seat user and the customer meets the quality requirements or does not meet the quality requirements; When it is determined according to the quality inspection result that the communication between the agent user and the customer does not meet the quality requirements, a reminder message is output to the agent user.

7. The method according to claim 1, characterized in that The method further comprises: If it is determined that the audio data of the customer is cross-talked into the agent recording data, outputting prompt information of customer-type crosstalk to prompt the recording process to be optimized by adopting an optimization method of customer-type crosstalk; If it is determined that the audio data of the customer is not intertwined in the agent recording data, a prompt message of inter-agent crosstalk is output to prompt the recording process to be optimized by using an optimization method of inter-agent crosstalk.

8. A data processing device, characterized in that: The device comprises: An acquisition unit, used to acquire seat recording data; the seat recording data refers to the audio data of the seat user's speaking period when the seat user is communicating with the customer; a determining unit, configured to determine a target time period corresponding to the non-seat user in the seat recording data if the seat recording data corresponds to at least two speakers and the at least two speakers include the seat user and a non-seat user; The acquisition unit is further used to acquire customer recording data; the customer recording data refers to the audio data of the customer's speaking period when the seat user communicates with the customer; The determination unit is further used to determine whether the customer's audio data is inserted into the seat recording data based on the overlap rate between the target time period and the customer's speaking time period, and the similarity between the voiceprint of the non-seat user and the voiceprint of the customer.

9. An electronic device, characterized in that: The device comprises: A processor; and a memory configured to store computer executable instructions, which, when executed, cause the processor to perform the data processing method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium is used to store computer-executable instructions, and the computer-executable instructions, when executed by a processor, implement the data processing method according to any one of claims 1 to 7.

Images