Call-out quality monitoring and optimizing system based on big data

Abstract

The invention discloses an outbound call quality monitoring and optimizing system based on big data, and belongs to the technical field of outbound call quality optimization. Comprising a call-out voice acquisition unit, a voice grouping unit, a call-out quality evaluation unit, an optimized voice library generation unit and a call-out optimization unit, the call-out voice acquisition unit acquires all call-out voices in a time period from the current to a certain previous time node to obtain a voice data array # imgabs0 #, the voice grouping unit is connected to a voice analysis module, and the call-out quality evaluation unit is connected to the call-out optimization unit. Each element in the voice data array # imgabs1 # is grouped, and the outbound quality evaluation unit communicates with the voice grouping unit; the invention discloses an outbound call quality monitoring and optimizing system based on big data, the outbound call quality monitoring and optimizing system comprises an outbound call quality monitoring and optimizing unit, the outbound call quality monitoring and optimizing unit obtains an evaluation intersection # imgabs2, counts the number of # imgabs3 # and then calculates the outbound call quality ratio # imgabs5 #, and the optimization voice library generating unit conducts intersection calculation on # imgabs6 #, # imgabs7 # and # imgabs8 # to obtain an intersection # imgabs9 #. The outbound call quality monitoring and optimizing system based on the big data can monitor the outbound call quality in real time, optimizes outbound call behaviors according to the outbound call quality, and improves the outbound call quality.

Description

Technical Field

[0001] The present invention specifically relates to an outbound call quality monitoring and optimization system based on big data, and belongs to the technical field of outbound call quality optimization. Background Art

[0002] In the current intelligent dialogue system based on the large language model (LLM), one implementation method is to retrieve the vector library and return the alternative speech to the LLM. The LLM selects the best alternative speech based on the context. The current criterion for judging the best is to intervene in the speech selection ability of the LLM through expert experience. However, with the increase in business complexity, judging each dialogue through expert experience and manually intervening in the speech selection of the LLM is very inefficient and cannot even objectively reflect the customer's communication willingness. How to optimize the speech selection of the intelligent dialogue system and improve the efficiency of intelligent dialogue is an urgent problem to be solved. , Chinese Patent Publication No.: CN119088918A, discloses a speech optimization method, device, electronic device and storage medium. The method optimizes the corpus data by a preset sorting method, that is, according to the weight, new corpus is added or the corpus with a small weight value is deleted, so as to realize continuous optimization of the corpus according to different scenarios, and then generate a speech selection model that is more suitable for different scenarios. After receiving the dialogue request from the client terminal, it can more accurately reply to the target speech that meets the scenario, and improve the accuracy of speech selection; however, the optimization method cannot monitor and feedback the outbound call quality, and cannot optimize the outbound call quality in real time. Summary of the invention

[0003] To solve the above problems, the present invention proposes an outbound call quality monitoring and optimization system based on big data, which can monitor the outbound call quality in real time, optimize the outbound call behavior according to the outbound call quality, and improve the outbound call quality.

[0004] The outbound call quality monitoring and optimization system based on big data of the present invention comprises: An outbound voice acquisition unit acquires all outbound voices from the current time to a certain time point in the past to obtain a voice data array , ; The previous time node is a set time node or a time node when the optimization evaluation was last completed; A voice grouping unit, the voice grouping unit is connected to the voice analysis module, and the voice grouping unit obtains the voice data array , then the voice data array Each element in is grouped, and the grouping process is as follows: Divide the voice data array according to the call duration ,when The call duration of an element in When the element is included in the invalid communication data array ;when The call duration of an element in ,and When the element is included in the effective communication data array ;when The call duration of an element in When the element is included in the inefficient communication data array ; Divide the voice data array according to the number of statistical interactions ,when The number of call interactions of an element within is less than When the element is included in the invalid communication data array ;when The number of call interactions of an element within is greater than ,and When the element is included in the effective communication data array ;when The number of call interactions of an element within is greater than When the element is included in the inefficient communication data array ; Divide the voice data array according to whether the conversion is successful ,when After an element in the array is successfully converted, the element is included in the successfully converted data array. ;when If an element in the conversion fails, the element is included in the conversion failure data array. ; An outbound call quality evaluation unit, the outbound call quality evaluation unit communicates with the voice grouping unit; obtaining an evaluation intersection , = ; and statistics Number of , then calculate the outbound call quality ratio ,when When it is greater than the set threshold and greater than the set optimization cycle length, the optimization instruction is output; (1) in, For voice data array Total amount; The optimized voice library generation unit communicates with the voice grouping unit; the optimized voice library generation unit obtains , and ; and , and Perform intersection calculation to get the intersection , ; The outbound call optimization unit is connected to the outbound call quality evaluation unit and the optimized voice library generation unit in communication; when the outbound call optimization unit receives the optimization instruction, it Send to the outbound call optimization unit to obtain outbound call optimization data.

[0005] When working, the outbound voice acquisition unit acquires all outbound voices within a set time period, and sends all outbound voices to the voice grouping unit, which groups all outbound voices to complete the outbound voice grouping, and then sends the outbound voice groups to the outbound quality evaluation unit, which evaluates the grouped outbound voices to obtain the outbound quality ratio. , and based on the outbound call quality ratio and setting the optimization cycle duration to determine whether to output the optimization instruction; when the outbound call optimization unit obtains the optimization instruction, the outbound call optimization unit obtains the database generated by the optimized voice library generation unit, optimizes it through the outbound call optimization unit, and obtains the outbound call optimization data.

[0006] Furthermore, the The duration is 5-10 seconds. The duration is 3~5 minutes. The number of times is 2-3 times. The duration is 8-12 times. A complete call interaction includes an outbound call output and a voice segment of the answer; the successful conversion is a successful transaction of the outbound call, transfer to a dedicated business person, or a positive response from the called party.

[0007] Furthermore, the outbound voice acquisition unit acquires the intersection Then, extract the called party voice segment, the outgoing call party voice segment and the responding call party voice segment; the outgoing call party voice segment is the called party voice segment that occurs before the called party voice segment and is closest to the called party voice segment; the responding call party voice segment is the outgoing call voice segment that occurs after the called party voice segment and is closest to the called party voice segment; and extract the associated words from the called party voice segment, the associated words are generated by the outgoing call reply dictionary library, and for each called party voice segment with associated words, obtain one associated word or a combination of multiple associated words as an index item, and each index item is connected to a responding call party voice segment; complete the intersection Grouping, getting grouping ,in ∈ ; For voice data array No. Elements The voice segment of the responding caller corresponding to the index item; Each index item is connected to a voice segment of the calling party; the intersection is completed Grouping, getting grouping ,in ∈ ; For voice data array No. Elements The caller's voice segment corresponding to each index item.

[0008] Furthermore, the outbound call optimization unit includes a high-frequency word extraction unit, a test set generation module and a calculation module; the high-frequency word extraction unit works as follows: The high-frequency word extraction unit obtains the group , and group the response caller voice segments corresponding to the same index item to obtain N groups of groups, each group of , then Extract high-frequency words. For a certain collection The The group responds to the calling party voice segment, , For collection The total amount of data sets; get After that, Extract a single high-frequency word, where a single high-frequency word is defined as a phrase in The number of occurrences in reaches the set threshold, set the threshold The calculation is as follows: (2) in, for The number of times a phrase appears in Number of occurrences ,and ; When the phrase is greater than 4, the phrase is assumed to be a single high-frequency word; for Total number of phrases; According to the above calculation, we get the sequence of single high-frequency words. , , then, the logarithm series Perform secondary high-frequency word calculation to obtain secondary high-frequency word sequence ; is the last secondary high-frequency word in the secondary high-frequency word sequence; the secondary high-frequency word calculation process is as follows: Get the sequence Any single high-frequency word in the list, and count the single high-frequency words In the series Number of times it appears in ,when Will Feed ; (3) in, For the series The total number of elements in ; For the series Any element in The high-frequency word extraction unit obtains the group , and Extract high-frequency words, extraction process and grouping The extraction process is consistent, and the secondary high-frequency word series is output After the operation is completed, the high-frequency word extraction unit outputs and ; The test set generation module obtains and , and calculate the intersection and union ,A= ; = ; Get the first-level importance phrase and secondary importance phrases , the first-level importance phrase is the phrase corresponding to intersection A, the second-level importance phrase The phrase obtained by union B-intersection A; The calculation module obtains , And calculation series ; The calculation sequence =( )- ; Next, traverse the calculation sequence , and from the calculation series Get the content included The sequence of ;statistics Each set of data in contains The elements in get the corresponding data of this group Elements combined, completed All Elements are combined and the same Elements are combined and merged, and finally the same The number of element combinations is arranged in descending order to obtain an optimized phrase, which includes The corresponding phrases and the top 1 group Combinations or groups of elements Phrases corresponding to the intersection of element combinations; finally, outbound call data containing optimized phrases is generated through manual editing or automatic AI generation.

[0009] Furthermore, when the AI ​​automatically generates outbound call data, the number of words in the outbound call data needs to be set, and the number of words in the outbound call data is the average value of the voice segment of the responding caller corresponding to each index item.

[0010] Furthermore, the outbound call quality evaluation unit is connected to an evaluation output module, which generates daily, monthly, quarterly and annual evaluation reports; the outbound call quality evaluation unit calculates the outbound call quality ratio at each stage. , and output the outbound call quality ratio through the evaluation output module , and output the composition of the outbound call quality ratio at the same time, that is, output , , and % of.

[0011] Furthermore, the outbound call quality evaluation unit further includes an optimal time period outbound call evaluation module, which obtains the optimal time period outbound call evaluation module within each time period. , and The number of elements in the intersection and the number of elements in the union; get the efficiency of outbound calls , ; By counting the effective outbound call volume in each time period, which is composed of successful conversion volume and effective communication volume, and configuring corresponding proportional factors, the quality of outbound calls can be comprehensively evaluated.

[0012] Compared with the prior art, the outbound call quality monitoring and optimization system based on big data of the present invention can be applied to a combination of robots and humans. When the robot's outbound call is connected with the called party, the robot transfers it to the human. The outbound call data can be tested manually and the entire test outbound call can be recorded. At the same time, the outbound call quality is evaluated, and the evaluation results are optimized to obtain optimization results. The new optimization results are executed in a new round of outbound call process to improve the outbound call quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 It is a schematic diagram of the overall structure of the outbound call quality monitoring and optimization system based on big data of the present invention.

[0014] Figure 2 The figure is a schematic diagram of the work flow structure of the outbound call quality monitoring and optimization system of the present invention.

[0015] Figure 3 The figure is a schematic diagram of the structure of the process of refining and grouping all outbound voice calls according to the present invention. DETAILED DESCRIPTION

[0016] like Figures 1 to 3 The outbound call quality monitoring and optimization system based on big data shown includes: An outbound voice acquisition unit acquires all outbound voices from the current time to a previous time point to obtain a voice data array , ; The previous time node is a set time node or a time node when the optimization evaluation was last completed; A voice grouping unit, the voice grouping unit is connected to the voice analysis module, and the voice grouping unit obtains the voice data array , then the voice data array Each element in is grouped, and the grouping process is as follows: Divide the voice data array according to the call duration ,when The call duration of an element in When the element is included in the invalid communication data array ;when The call duration of an element in ,and When the element is included in the effective communication data array ;when The call duration of an element in When the element is included in the inefficient communication data array ; Divide the voice data array according to the number of statistical interactions ,when The number of call interactions of an element within is less than When the element is included in the invalid communication data array ;when The number of call interactions of an element within is greater than ,and When the element is included in the effective communication data array ;when The number of call interactions of an element within is greater than When the element is included in the inefficient communication data array ; Divide the voice data array according to whether the conversion is successful ,when After an element in the array is successfully converted, the element is included in the successfully converted data array. ;when If an element in the conversion fails, the element is included in the conversion failure data array. ; An outbound call quality evaluation unit, the outbound call quality evaluation unit communicates with the voice grouping unit; obtaining an evaluation intersection , = ; and statistics Number of , then calculate the outbound call quality ratio ,when When it is greater than the set threshold and greater than the set optimization cycle length, the optimization instruction is output; (1) in, For voice data array Total amount; The optimized voice library generation unit communicates with the voice grouping unit; the optimized voice library generation unit obtains , and ; and , and Perform intersection calculation to get the intersection , ; The outbound call optimization unit is connected to the outbound call quality evaluation unit and the optimized voice library generation unit in communication; when the outbound call optimization unit receives the optimization instruction, it Send to the outbound call optimization unit to obtain outbound call optimization data.

[0017] When working, the outbound voice acquisition unit acquires all outbound voices within a set time period, and sends all outbound voices to the voice grouping unit, which groups all outbound voices to complete the outbound voice grouping, and then sends the outbound voice groups to the outbound quality evaluation unit, which evaluates the grouped outbound voices to obtain the outbound quality ratio. , and based on the outbound call quality ratio and setting the optimization cycle duration to determine whether to output the optimization instruction; when the outbound call optimization unit obtains the optimization instruction, the outbound call optimization unit obtains the database generated by the optimized voice library generation unit, optimizes it through the outbound call optimization unit, and obtains the outbound call optimization data.

[0018] Said The duration is 5-10 seconds. The duration is 3~5 minutes. The number of times is 2-3 times. The duration is 8-12 times. A complete call interaction includes an outbound call output and a voice segment of the answer; the successful conversion is a successful transaction of the outbound call, transfer to a dedicated business person, or a positive response from the called party.

[0019] The outbound voice acquisition unit acquires the intersection Then, extract the called party voice segment, the outgoing call party voice segment and the responding call party voice segment; the outgoing call party voice segment is the called party voice segment that occurs before the called party voice segment and is closest to the called party voice segment; the responding call party voice segment is the outgoing call voice segment that occurs after the called party voice segment and is closest to the called party voice segment; and extract the associated words from the called party voice segment, the associated words are generated by the outgoing call reply dictionary library, and for each called party voice segment with associated words, obtain one associated word or a combination of multiple associated words as an index item, and each index item is connected to a responding call party voice segment; complete the intersection Grouping, getting grouping ,in ∈ ; For voice data array No. Elements The voice segment of the responding caller corresponding to the index item; Each index item is connected to a voice segment of the calling party; the intersection is completed Grouping, getting grouping ,in ∈ ; For voice data array No. Elements The caller's voice segment corresponding to each index item.

[0020] The outbound call optimization unit includes a high-frequency word extraction unit, a test set generation module and a calculation module; the high-frequency word extraction unit works as follows: The high-frequency word extraction unit obtains the group , and group the response caller voice segments corresponding to the same index item to obtain N groups of groups, each group of , then Extract high-frequency words. For a certain collection The The group responds to the calling party voice segment, , For collection The total amount of data sets; get After that, Extract a single high-frequency word, where a single high-frequency word is defined as a phrase in The number of occurrences in reaches the set threshold, set the threshold The calculation is as follows: (2) in, for The number of times a phrase appears in Number of occurrences ,and ; When the phrase is greater than 4, the phrase is assumed to be a single high-frequency word; for Total number of phrases; According to the above calculation, we get the sequence of single high-frequency words. , , then, the logarithm series Perform secondary high-frequency word calculation to obtain secondary high-frequency word sequence ; is the last secondary high-frequency word in the secondary high-frequency word sequence; the secondary high-frequency word calculation process is as follows: Get the sequence Any single high-frequency word in the list, and count the single high-frequency words In the series Number of times it appears in ,when Will Feed ; (3) in, For the series The total number of elements in ; For the series Any element in The high-frequency word extraction unit obtains the group , and Extract high-frequency words, extraction process and grouping The extraction process is consistent, and the secondary high-frequency word series is output After the operation is completed, the high-frequency word extraction unit outputs and ; The test set generation module obtains and , and calculate the intersection and union ,A= ; = ; Get the first-level importance phrase and secondary importance phrases , the first-level importance phrase is the phrase corresponding to intersection A, the second-level importance phrase The phrase obtained by union B-intersection A; The calculation module obtains , And calculation series ; The calculation sequence =( )- ; Next, traverse the calculation sequence , and from the calculation series Get the content included The sequence of ;statistics Each set of data in contains The elements in get the corresponding data of this group Elements combined, completed All Elements are combined and the same Elements are combined and merged, and finally the same The number of element combinations is arranged in descending order to obtain an optimized phrase, which includes The corresponding phrases and the top 1 group Combinations or groups of elements Phrases corresponding to the intersection of element combinations; finally, outbound call data containing optimized phrases is generated through manual editing or automatic AI generation.

[0021] When the AI ​​automatically generates outbound call data, the number of words in the outbound call data needs to be set, and the number of words in the outbound call data is the average value of the voice segment of the responding caller corresponding to each index item.

[0022] The outbound call quality evaluation unit is connected to an evaluation output module, which generates daily, monthly, quarterly and annual evaluation reports; the outbound call quality evaluation unit calculates the outbound call quality ratio at each stage. , and output the outbound call quality ratio through the evaluation output module , and output the composition of the outbound call quality ratio at the same time, that is, output , , and % of.

[0023] The outbound call quality evaluation unit further includes an optimal time period outbound call evaluation module, wherein the optimal time period outbound call evaluation module obtains the time intervals in each time period. , and The number of elements in the intersection and the number of elements in the union; get the efficiency of outbound calls , ; By counting the effective outbound call volume in each time period, which is composed of successful conversion volume and effective communication volume, and configuring corresponding proportional factors, the quality of outbound calls can be comprehensively evaluated.

[0024] The above embodiments are only preferred implementations of the present invention, so any equivalent changes or modifications made according to the structures, features and principles described in the scope of application of the present invention are included in the scope of application of the present invention.

Claims

1. An outbound call quality monitoring and optimization system based on big data, characterized by: include An outbound voice acquisition unit acquires all outbound voices from the current time to a certain time point in the past to obtain a voice data array , ; A voice grouping unit, the voice grouping unit is connected to the voice analysis module, and the voice grouping unit obtains the voice data array , then the voice data array Each element in is grouped, and the grouping process is as follows: Divide the voice data array according to the call duration ,when The call duration of an element in When the element is included in the invalid communication data array ;when The call duration of an element in ,and When the element is included in the effective communication data array ;when The call duration of an element in When the element is included in the inefficient communication data array ; Divide the voice data array according to the number of statistical interactions ,when The number of call interactions of an element within is less than When the element is included in the invalid communication data array ;when The number of call interactions of an element within is greater than ,and When the element is included in the effective communication data array ;when The number of call interactions of an element within is greater than When the element is included in the inefficient communication data array ; Divide the voice data array according to whether the conversion is successful ,when After an element in the array is successfully converted, the element is included in the successfully converted data array. ;when If an element in the conversion fails, the element is included in the conversion failure data array. ; An outbound call quality evaluation unit, the outbound call quality evaluation unit communicates with the voice grouping unit; obtaining an evaluation intersection , = ; and statistics Number of , then calculate the outbound call quality ratio ,when When it is greater than the set threshold and greater than the set optimization cycle length, the optimization instruction is output; （1） in, For voice data array Total amount; The optimized voice library generation unit communicates with the voice grouping unit; the optimized voice library generation unit obtains , and ; and , and Perform intersection calculation to get the intersection , ; The outbound call optimization unit is connected to the outbound call quality evaluation unit and the optimized voice library generation unit in communication; when the outbound call optimization unit receives the optimization instruction, it Send to the outbound call optimization unit to obtain outbound call optimization data.

2. The outbound call quality monitoring and optimization system based on big data according to claim 1 is characterized by: Said The duration is 5-10 seconds. The duration is 3~5 minutes. The number of times is 2-3 times. The duration is 8-12 times. A complete call interaction includes an outbound call output and a voice segment of the answer; the successful conversion is a successful transaction of the outbound call, transfer to a dedicated business person, or a positive response from the called party.

3. The outbound call quality monitoring and optimization system based on big data according to claim 1 is characterized by: The outbound voice acquisition unit acquires the intersection Then, extract the called party voice segment, the outgoing call party voice segment and the responding call party voice segment; the outgoing call party voice segment is the called party voice segment that occurs before the called party voice segment and is closest to the called party voice segment; the responding call party voice segment is the outgoing call voice segment that occurs after the called party voice segment and is closest to the called party voice segment; and extract the associated words from the called party voice segment, the associated words are generated by the outgoing call reply dictionary library, and for each called party voice segment with associated words, obtain one associated word or a combination of multiple associated words as an index item, and each index item is connected to a responding call party voice segment; complete the intersection Grouping, getting grouping ,in ∈ ; For voice data array No. Elements The voice segment of the responding caller corresponding to the index item; Each index item is connected to a voice segment of the calling party; the intersection is completed Grouping, getting grouping ,in ∈ ; For voice data array No. Elements The caller's voice segment corresponding to each index item.

4. The outbound call quality monitoring and optimization system based on big data according to claim 3 is characterized by: The outbound call optimization unit includes a high-frequency word extraction unit, a test set generation module and a calculation module; the high-frequency word extraction unit works as follows: The high-frequency word extraction unit obtains the group , and group the response caller voice segments corresponding to the same index item to obtain N groups of groups, each group of , then Extract high-frequency words. For a certain collection The The group responds to the calling party voice segment, , For collection The total number of data sets; Get After that, Extract a single high-frequency word, where a single high-frequency word is defined as a phrase in The number of occurrences in reaches the set threshold, set the threshold The calculation is as follows: （2） in, for The number of times a phrase appears in Number of occurrences ,and ; When the phrase is greater than 4, the phrase is assumed to be a single high-frequency word; for Total number of phrases; According to the above calculation, we get the sequence of single high-frequency words. , , then, the logarithm series Perform secondary high-frequency word calculation to obtain secondary high-frequency word sequence ; is the last secondary high-frequency word in the secondary high-frequency word sequence; the secondary high-frequency word calculation process is as follows: Get the sequence Any single high-frequency word in the list, and count the single high-frequency words In the series Number of times it appears in ,when When Feed ; （3） in, For the series The total number of elements in ; For the series Any element in The high-frequency word extraction unit obtains the group , and Extract high-frequency words, extraction process and grouping The extraction process is consistent, and the secondary high-frequency word series is output After the operation is completed, the high-frequency word extraction unit outputs and ; The test set generation module obtains and , and calculate the intersection and union ,A= ; = ; Get the first-level importance phrase and secondary importance phrases , the first-level importance phrase is the phrase corresponding to intersection A, the second-level importance phrase The phrase obtained by union B-intersection A; The calculation module obtains , And calculation series ; The calculation sequence =( )- ; Next, traverse the calculation sequence , and from the calculation series Get the content included The sequence of ;statistics Each set of data in contains The elements in get the corresponding data of this group Elements combined, completed All Elements are combined and the same Elements are combined and merged, and finally the same The number of element combinations is arranged in descending order to obtain an optimized phrase, which includes The corresponding phrases and the top 1 group Combinations or groups of elements Phrases corresponding to the intersection of element combinations; finally, outbound call data containing optimized phrases is generated through manual editing or automatic AI generation.

5. The outbound call quality monitoring and optimization system based on big data according to claim 4 is characterized by: When the AI ​​automatically generates outbound call data, the number of words in the outbound call data needs to be set, and the number of words in the outbound call data is the average value of the voice segment of the responding caller corresponding to each index item.

6. The outbound call quality monitoring and optimization system based on big data according to claim 1 is characterized by: The outbound call quality assessment unit is connected to an assessment output module, and the assessment output module generates daily, monthly, quarterly and annual assessment reports.

7. The outbound call quality monitoring and optimization system based on big data according to claim 1 is characterized by: The outbound call quality evaluation unit further includes an optimal time period outbound call evaluation module, wherein the optimal time period outbound call evaluation module obtains the time intervals in each time period. , and The number of elements in the intersection and the number of elements in the union; then, the outbound call efficiency is obtained , .

Images