A Consistency Audit System and Method for Survey Data Based on Customer Interaction Services

Abstract

This invention relates to the field of voice interaction survey data processing technology, and discloses a survey data consistency auditing system and method based on customer interaction services. The system acquires at least one audit item data, including basic data and several voice samples of interaction between the visitor and the corresponding surveyee; parses each voice sample to form corresponding interaction data; performs consistency auditing on the process of parsing voice samples to form interaction data, including basic record consistency auditing and voice emotion consistency auditing; creates a task library, matches and / or updates audit tasks in the task library based on the basic data of the audit item, obtaining multiple target audit tasks for basic record consistency auditing; constructs an emotion recognition model to identify the emotions of the visitor and the surveyee in the voice samples, and performs voice emotion consistency auditing; marks voice segments with abnormalities in the voice sample consistency auditing as anomalies and performs a second review; thus improving the efficiency and accuracy of voice interaction survey auditing.

Description

Technical Field

[0001] This invention relates to the field of voice interaction survey data processing technology, specifically to a survey data consistency verification system and method based on customer interaction services. Background Technology

[0002] Market research, as a crucial tool for business decision-making, centers on the systematic collection, analysis, and interpretation of data to help companies better understand the market environment, consumer behavior, and competitive landscape. Effective market research not only reveals potential opportunities and threats in the market but also provides companies with the basis for formulating or adjusting marketing strategies, thereby optimizing products and services and enhancing competitiveness.

[0003] When conducting market research, researchers can choose from various methods to obtain information, including but not limited to questionnaires, telephone interviews, face-to-face interviews, and online surveys. Each method has its characteristics and applicable scenarios. For example, in the service industry, questionnaire-based telephone interviews are widely used because they allow direct contact with the target audience, immediate answers to questions, and a high response rate, making the collected research data highly valuable.

[0004] In-depth analysis and review of survey data are essential for unlocking its full value. However, as businesses expand and their services increase, traditional review methods face challenges. Traditional methods of manually listening to voice samples obtained through surveys are time-consuming, labor-intensive, and susceptible to subjective factors such as the reviewer's professional competence and concentration, which can affect the final quality assessment. Furthermore, with massive amounts of voice samples, full review becomes impractical, while random sampling cannot guarantee coverage of all key issues, potentially missing important feedback and negatively impacting the accuracy and reliability of the survey results. Summary of the Invention

[0005] This invention aims to provide a data consistency auditing system and method based on customer interaction services. It adopts a dual auditing approach that combines intelligent auditing and secondary review to reduce the probability of mis-auditing and omissions, overcomes the limitations of traditional manual monitoring auditing methods, greatly improves auditing efficiency, and reduces labor costs. At the same time, based on the data characteristics of the audited project itself, the created task library, and the emotion recognition model, it can not only achieve data consistency auditing, but also dynamically adjust the auditing strategy to improve the adaptability and timeliness of the auditing work.

[0006] The basic solution provided by this invention is: a method for verifying the consistency of survey data based on customer interaction services, the method comprising:

[0007] S100, Obtain at least one audit item data, which includes basic data and several voice samples obtained from the interaction between several visitors and the corresponding survey subjects.

[0008] S200 parses each speech sample to form corresponding interactive data, which includes multiple questions and their answers;

[0009] S210, to conduct consistency audit on the process of parsing voice samples to form interactive data, including basic record consistency audit and voice emotion consistency audit;

[0010] S211, Create a task library to represent the mapping relationship between several review tasks and review items; Based on the basic data of the review items, match and / or update the review tasks in the task library to obtain multiple target review tasks; Based on the multiple target review tasks, perform basic record consistency review of voice samples and corresponding interaction data.

[0011] S212, Construct an emotion recognition model to identify the emotions of visitors and survey subjects in the voice samples respectively, and conduct voice emotion consistency verification between the voice samples and corresponding interaction data.

[0012] S300 marks the speech segments with abnormalities in the consistency audit of the speech samples as abnormal, and sends the abnormal speech segments to several review terminals according to the preset distribution method. The review terminals perform secondary review and corresponding review processing based on the review standards.

[0013] This invention provides a method for verifying the consistency of survey data based on customer interaction services, and also provides a system for verifying the consistency of survey data based on customer interaction services. The system includes a voice sample acquisition module, a parsing module, a consistency verification module, a secondary verification module, and a verification terminal.

[0014] The voice sample acquisition module is used to acquire basic data of at least one audit item and several voice samples obtained from the interaction between several visitors and the corresponding survey subjects.

[0015] The parsing module is used to parse speech samples to form interactive data, which includes multiple questions and their answers.

[0016] The consistency audit module is used to perform consistency audits on the process of parsing voice samples to form interactive data, including basic record consistency audits and voice emotion consistency audits.

[0017] The consistency audit module includes a task library module, which creates a task library to represent the mapping relationship between several audit tasks and audit projects; a basic record consistency audit submodule, which matches and / or updates audit tasks in the task library based on the basic data of the audit projects to obtain multiple target audit tasks, and performs basic record consistency audit of voice samples and corresponding interaction data based on multiple target audit tasks; and a voice emotion consistency audit submodule, which builds an emotion recognition model to identify the emotions of visitors and survey subjects in voice samples, and performs voice emotion consistency audit of voice samples and corresponding interaction data.

[0018] The secondary review module is used to mark the speech segments with abnormalities in the consistency review of the speech samples as abnormal, and send the abnormal speech segments to several review terminals according to the preset distribution method.

[0019] The review end is used for secondary review and corresponding review processing based on the review standards.

[0020] The working principle and advantages of this invention are as follows:

[0021] Compared to existing technologies, this solution introduces advanced information technology and combines intelligent review with secondary verification to overcome the limitations of traditional manual monitoring and review methods, significantly improving review efficiency and reducing labor costs. The introduction of intelligent review increases the total survey sample size, reaching more respondents and improving survey outreach efficiency. It also collects more questionnaire samples to support data analysis, further enhancing the survey's effectiveness and providing stronger data support for corporate strategic planning. To ensure the accuracy of intelligent review and balance the number of secondary verifications, the solution proposes consistency checks and dynamic adjustments to the review strategy. Based on the data characteristics of the review projects and a constructed task library, the matching and updating of review projects and tasks allows for adjustments to the review strategy to adapt to changing requirements, improving the adaptability and timeliness of the review work. Based on an emotion recognition model, it can gain a deeper understanding of the respondents' and visitors' true intentions and attitudes by analyzing non-verbal information such as tone, emotion, and intonation retained in voice samples, thus improving the accuracy of the review. Attached Figure Description

[0022] Figure 1 This is a flowchart illustrating the survey data consistency verification method based on customer interaction services provided in Embodiment 1 of the present invention.

[0023] Figure 2 This is a schematic diagram of the structure of the survey data consistency audit system based on customer interaction services provided in Embodiment 1 of the present invention;

[0024] Figure 3 This is a schematic diagram of the interface of the survey data consistency verification system based on customer interaction services provided in Embodiment 1 of the present invention. Figure 1 ;

[0025] Figure 4 This is a schematic diagram of the interface of the survey data consistency verification system based on customer interaction services provided in Embodiment 1 of the present invention. Figure 2 ;

[0026] Figure 5 This is a schematic diagram of the interface of the survey data consistency verification system based on customer interaction services provided in Embodiment 1 of the present invention. Figure 3 ;

[0027] Figure 6 This is a schematic diagram of the interface of the survey data consistency verification system based on customer interaction services provided in Embodiment 1 of the present invention. Figure 4 ;

[0028] Figure 7 This is a schematic diagram of the interface of the survey data consistency verification system based on customer interaction services provided in Embodiment 1 of the present invention. Figure 5 ;

[0029] Figure 8 This is a schematic diagram of the interface of the survey data consistency verification system based on customer interaction services provided in Embodiment 1 of the present invention. Figure 6 ;

[0030] Figure 9 This is a schematic diagram of the interface of the survey data consistency verification system based on customer interaction services provided in Embodiment 1 of the present invention. Figure 7 ;

[0031] Figure 10 This is a schematic diagram of the interface of the survey data consistency verification system based on customer interaction services provided in Embodiment 1 of the present invention. Figure 8 . Detailed Implementation

[0032] The following detailed explanation illustrates the specific implementation methods:

[0033] Example 1

[0034] The basics are as follows: Figure 1 The following is an example of a method for verifying the consistency of survey data based on customer interaction services. The method includes:

[0035] S100, acquire basic data of at least one audit item and several voice samples obtained from the interaction between several visitors and the corresponding survey subjects.

[0036] Specifically, audit projects, such as satisfaction surveys in the telecommunications and service industries, employ questionnaire-based voice interaction surveys, typically conducted by telephone, but can also utilize any other channel capable of voice interaction. The basic data for audit projects includes user information, questionnaire content, interview criteria, and quality control standards. User information includes company name, industry, etc.; questionnaire content includes multiple questions and multiple pre-set answers; interview criteria define the requirements for the interviewer during voice interaction with the survey respondents, such as standard introductory and closing remarks; and quality control standards define the consistency requirements for the voice samples obtained from the voice interaction, including audit indicators and their corresponding standard conditions.

[0037] S200 parses speech samples to form interactive data, which includes multiple questions and their answers.

[0038] Specifically, natural language processing, text vectorization, semantic analysis, question answering, and large language modeling techniques are employed, including but not limited to speech-to-text conversion, text segmentation, and question-and-answer matching and extraction. Ultimately, the interaction data of all questions and corresponding answers in the corresponding voice samples of the survey respondents are obtained. There may be situations where the survey respondents answer all the questions in the questionnaire, answer only some of the questions, or ask some new questions.

[0039] S210, a consistency audit is performed on the process of parsing speech samples to form interactive data, including a basic record consistency audit and a speech emotion consistency audit. In this embodiment, if all audits are normal and there are no abnormal markers, the audit is considered passed and no secondary audit is required. If there are abnormal markers, the audit is considered failed and a secondary audit is required.

[0040] Specifically, conducting in-depth analysis of the text and emotional content within voice samples for review can improve the accuracy of the correspondence between voice samples and interaction data.

[0041] S211, Create a task library to represent the mapping relationship between several audit tasks and audit projects.

[0042] Specifically, a task library is established. Based on the audit indicators and corresponding standard conditions in the regular audit quality inspection standards, each audit indicator is transformed into a specific audit task. For example, if an audit indicator requires complete recording, an audit task is created to check whether the voice sample includes all questions and corresponding answers; if an audit indicator requires the interviewer to inform the survey recipient that "the call will be recorded," or the interviewer to inform the survey recipient that "you can receive calls from others locally for free, so receiving calls from us is also free," or the interviewer to inform the recipient using a third-party identity that "we are an independent survey company, collecting your opinions as an independent third party," then an audit task is created accordingly to check whether the interviewer accurately conveyed the corresponding content in the voice sample; if an audit indicator requires auditing outbound call time, an audit task is created to check whether the outbound call time in the voice sample meets the corresponding duration requirement; specific audit tasks can be initially set based on the audit indicators in the regular audit quality inspection standards and dynamically adjusted later.

[0043] During the review process, based on the basic data of the review project, review tasks are matched and / or updated in the task library to obtain multiple target review tasks; based on the multiple target review tasks, the consistency of the basic records of voice samples and corresponding interaction data is reviewed.

[0044] Specifically, the basic data for the audit project includes quality inspection standards. All audit indicator features are extracted from the quality inspection standards and matched against all audit tasks in the current task library. If a match is found, it becomes the target audit task; otherwise, the audit task is updated with the corresponding quality inspection standard content and becomes the target audit task. The matching method can use a coding model to encode the quality inspection standard content, and then use this coding for clustering. Tasks within the same cluster are assigned the same audit task.

[0045] Updating audit tasks includes classifying and rearranging their levels. Specifically, when a new audit indicator is added, it is matched with existing audit tasks, and the tasks are rearranged. If a new audit task contains multiple subtasks, all related subtasks are re-instantiated and then combined into a larger task in sequence. The order of subtasks within a new audit task matters, but it does not affect the existing tasks. Furthermore, if a new audit indicator has the highest matching degree with one of the existing audit tasks, it is updated as a subtask under that existing audit task. Multiple audit tasks are formed in the task library, each with multiple sequential subtasks. Each audit project matches the corresponding target audit task from the task library based on quality inspection standards.

[0046] The consistency of the basic records of voice samples and corresponding interaction data is verified based on multiple target verification tasks. The verification time depends on the longest target verification task, rather than all verification tasks being used for each verification project. For example, if project A uses 8 verification tasks and project B uses 10 verification tasks, the verification time for project A depends on the longest one among its 8 verification tasks, and the verification time for project B depends on the longest one among its 10 verification tasks.

[0047] The results are then summarized: when all sub-tasks under the corresponding audit project are completed and pass the test, the consistency audit of the basic record of the voice sample is passed; otherwise, the consistency audit of the basic record of the voice sample fails, and the failed tasks and their error information are marked, that is, the corresponding voice segments are marked as abnormal, and text annotations are made in the interactive data, such as Q1 has an anomaly, Q3 has an anomaly, incomplete recording, etc. All sub-tasks can be processed independently and in parallel, improving data processing efficiency.

[0048] After updating the audit tasks, check the audit tasks against the audit indicators. If any tasks are found to be missing, add them to the task library in a timely manner to fill in the missing tasks and ensure that the audit tasks fully cover the quality inspection standards of the audit project.

[0049] S212, Construct an emotion recognition model to identify the emotions of visitors and survey subjects in the voice samples, and conduct a voice emotion consistency check between the voice samples and the corresponding interaction data.

[0050] Specifically, voice samples retain non-verbal information such as tone, emotion, and intonation, which helps to understand the true intentions, attitudes, and other emotional information of the survey subjects and visitors more deeply. Therefore, by making consistency judgments based on voice intonation, interview time, response time, and other interactive data during the interview process, and conducting relevant analysis and testing, we can analyze the accuracy of the data more deeply.

[0051] Using a voiceprint recognition model, different voices are assigned to different speakers (interviewers, respondents). Then, an emotion recognition model is used to identify the emotions of the interviewer and the respondents in the voice samples. The interviewer's emotion can be represented as positive or negative, and the respondents' emotion can be represented as whether there is a possibility that the interviewer is inducing the respondents' emotions. This is used to determine whether the consistency passes the test, as detailed below:

[0052] In S212, the emotion recognition model includes a visitor emotion recognition sub-model, which is used to judge the positive and negative of the visitor's speech segment. If the judgment is negative, the speech emotion consistency check is abnormal, and the corresponding speech segment judged as negative is marked as abnormal.

[0053] In this embodiment, the emotion2vec model is used to label some data in the interactive scenario of this solution (using annotations), and the visitor emotion recognition sub-model is fine-tuned and trained according to the emotion2vec model to obtain the model. Specifically, based on the emotion2vec model, a neural network layer is extended, and the training labels are positive and negative, with overfitting to negative, thus forming a binary classification model. The visitor's voice and tone are detected, and negative voice samples are judged as abnormal samples; if the number of samples to be reviewed is small, positive voice samples can be sampled for review.

[0054] The timing patterns of most interactions and most interviews are relatively fixed; therefore, the response time and total interview time are usually predictable.

[0055] Therefore, in S212, the emotion recognition model includes a sub-model for recognizing the emotion of the surveyed subject. This sub-model calculates the duration difference between the response duration of each question in the surveyed subject's speech segment and the average response duration for that question. If the duration difference does not meet the threshold requirement, the speech emotion consistency check is abnormal, and the corresponding speech segment for that question and its response is marked as abnormal. The duration is used to determine whether the interviewer may be inducing the surveyed subject.

[0056] In this embodiment, the average response time for the question is the average of the response times for the same question across all respondents corresponding to the same interviewer, or the average of the response times for the same question across all respondents. The threshold requirement can be that if the response time exceeds the average by 30% to 50%, the access time for that question is considered abnormal. In cases involving a large number of responses, only responses that do not meet the threshold requirement and their corresponding audio segments can be considered abnormal.

[0057] S300 marks the speech segments with abnormalities in the consistency audit of the speech samples as abnormal, and sends the abnormal speech segments to several review terminals according to the preset distribution method. The review terminals perform secondary review and corresponding review processing based on the review standards.

[0058] Specifically, the preset distribution method includes evenly distributing abnormal audio segments to the currently active review terminals, ensuring that each abnormal audio segment can only be processed by one review terminal, and balancing the duration during the distribution. For example, if there are a total of 100 abnormal audio segments, totaling 242 minutes, and 3 review terminals, the distribution is generally based on 80-85 minutes of abnormal audio segments per review terminal. However, it is necessary to ensure that each abnormal audio segment can only be processed by one review terminal, and to distribute the tasks evenly by minimizing the difference in the duration of each review terminal, so that the review tasks are fairly distributed to each review terminal, thereby improving review efficiency.

[0059] The review criteria integrate at least quality inspection criteria, review tasks, and sentiment determination criteria, which can cover all review requirements for anomalies and provide review processing methods based on the criteria, such as updating the answers to questions in the interaction data.

[0060] S211 also includes, for the same speech sample, using a large language model to extract all answers for each question, repeatedly extracting answers several times, judging whether the extracted answers are consistent each time, comparing the number of consistent and inconsistent occurrences, and determining whether the basic record consistency audit passes. If the number of consistent occurrences is greater than the number of inconsistent occurrences, the basic record consistency audit passes; otherwise, the question and its corresponding speech segment are marked as abnormal. In this embodiment, the same question is judged 3 or 5 times. This number of judgments is a relatively balanced choice, which can balance system performance and judgment effect.

[0061] It also includes S500, which displays the statistical results of the survey process, including the number of samples reviewed by the system, the number of samples reviewed, the number of samples that failed the review, the rejection rate, the number of people who participated in the review, the analysis of sample data sources, the analysis of review samples, the data analysis of the interviewers, and the analysis of rejection types; it also includes a list of abnormal voice segments, which are associated with information such as reviewers, task progress, task status, review time, and review results.

[0062] like Figure 2 As shown, this solution provides a survey data consistency audit system based on customer interaction services to implement the above-mentioned survey data consistency audit method based on customer interaction services.

[0063] The system includes a voice sample acquisition module, a parsing module, a consistency verification module, a secondary verification module, and a verification terminal;

[0064] The voice sample acquisition module is used to acquire basic data of at least one audit item and several voice samples obtained from the interaction between several visitors and the corresponding survey subjects.

[0065] The parsing module is used to parse speech samples to form interactive data, which includes multiple questions and their answers.

[0066] The consistency audit module is used to perform consistency audits on the process of parsing voice samples to form interactive data, including basic record consistency audits and voice emotion consistency audits.

[0067] The consistency audit module includes a task library module, which creates a task library to represent the mapping relationship between several audit tasks and audit projects; a basic record consistency audit submodule, which matches and / or updates audit tasks in the task library based on the basic data of the audit projects to obtain multiple target audit tasks, and performs basic record consistency audit of voice samples and corresponding interaction data based on multiple target audit tasks; and a voice emotion consistency audit submodule, which builds an emotion recognition model to identify the emotions of visitors and survey subjects in voice samples, and performs voice emotion consistency audit of voice samples and corresponding interaction data.

[0068] The secondary review module is used to mark the speech segments with abnormalities in the consistency review of the speech samples as abnormal, and send the abnormal speech segments to several review terminals according to the preset distribution method.

[0069] The review end is used for secondary review and corresponding review processing based on the review standards.

[0070] It also includes a display panel for statistical presentation of survey results. Specifically, as follows... Figure 3 , Figure 4 , Figure 5 , Figure 6 , Figure 7 , Figure 8 , Figure 9 and Figure 10 As shown.

[0071] It is understandable that the above system can fully execute the above methods, and the specific process will not be described in detail.

[0072] This embodiment provides a data consistency auditing system and method based on customer interaction services. By introducing advanced information technology and combining intelligent auditing with secondary review, it overcomes the limitations of traditional manual monitoring auditing methods, significantly improving auditing efficiency and reducing labor costs. The introduction of intelligent auditing increases the total sample size, reaching more survey respondents and improving survey promotion efficiency. It also collects more questionnaire samples to support data analysis, further enhancing the survey's effectiveness and providing stronger data support for corporate strategic planning. To ensure the accuracy of intelligent auditing and balance the number of secondary reviews, the system proposes consistency auditing and dynamic adjustment of auditing strategies. Based on a constructed task library, the system adjusts auditing strategies to adapt to constantly changing auditing requirements by matching and updating auditing projects and tasks, improving the timeliness and adaptability of the auditing work. Based on an emotion recognition model, it can gain a deeper understanding of the true intentions and attitudes of survey respondents and visitors by analyzing non-verbal information such as tone, emotion, and intonation retained in voice samples, thus improving the accuracy of the audit.

[0073] Example 2

[0074] Unlike Implementation Example 1, each audit task is classified into importance levels, and the importance level of each audit task and / or the audit pass standard of the audit task are dynamically adjusted according to the audit project data or standard differences, that is, the audit degree is dynamically adjusted.

[0075] Specifically, the importance level can be divided into three categories: "important", "average", and "not important".

[0076] The approval criteria for review tasks at each importance level are dynamically adjusted based on the review project data. The total number of speech segments corresponding to "unimportant" review tasks is obtained from all speech samples in the review project data. After performing a consistency review on all speech samples according to all target review tasks, if the proportion of the number of abnormally marked speech segments detected in the "unimportant" review task to the total number of speech segments corresponding to the "unimportant" review task does not exceed a threshold proportion, or does not exceed the threshold number, the review is allowed. The abnormal markings are only retained as a record for this review, but no secondary review or statistical analysis is performed.

[0077] The approval criteria for review tasks at each importance level are dynamically adjusted based on standard differences. Since there is typically a time lag between the actual interview and the actual review (e.g., the actual interview is completed in April, and the actual review in May), there may be discrepancies between the interview standards and the quality control standards at the time of review. For example, the interview standards may not require interviewers to inform respondents that "the call will be recorded," while the quality control standards require them to do so. This situation can lead to anomalies in almost all voice samples, further resulting in inaccurate reviews.

[0078] Therefore, to avoid inaccurate audits due to standard differences and an excessive number of anomalies requiring secondary audits, the standard conditions for corresponding audit tasks are dynamically adjusted based on standard differences. This relaxes requirements, making it easier for samples to pass audits. Specifically, a comparison is made between the quality inspection standard and the access standard. Audit indicators in the quality inspection standard are marked with standard differences. The target audit task, while carrying an importance level, may also have a standard difference mark. After auditing using the target audit task, abnormal audio segments detected by the audit task with the standard difference mark are sorted according to the importance level of the audit task. They can be allowed to pass using the aforementioned less important audit task detection method, or the number of valid review terminals or audit pressure can be limited to allow passing audits. Data from such allowed audits is still marked as anomalies, but these anomaly marks are only retained as records for this audit and are not used for secondary review statistics.

[0079] Correspondingly, the task library module is also used to classify audit tasks into importance levels and dynamically adjust the audit pass standards for each importance level based on audit project data and standard differences.

[0080] The survey data consistency audit system and method based on customer interaction services provided in this embodiment can still accurately audit anomalies. However, it marks anomalies of unimportant level and anomalies caused by standard differences as anomalies but does not push them for secondary review. This reduces the amount of auditing that needs to be pushed for secondary review, thereby balancing the manual cost of secondary review with the accuracy of auditing, and improving the adaptability of auditing work in the process of dynamic changes in auditing standards.

[0081] Example 3

[0082] Unlike Examples 1 and 2, the consistency audit order and audit level are dynamically adjusted based on changes in external conditions.

[0083] Specifically, changes in external conditions, such as an increase or decrease in the number of projects to be reviewed, a shortening of review time, or a decrease in the number of reviewers, necessitate adaptive adjustments to the review methods to ensure that the review work can be carried out effectively despite changes in external conditions.

[0084] Set up a whitelist of customers and their levels. When there are multiple items to be reviewed, sort them according to the customer level. When conducting a second review, review them according to the review order.

[0085] Further refine the review process and its service quality level. Based on the number of review items, divide the review process into multiple groups according to the service quality level. Each group handles one review item. The number of review processes and the service quality level of each group are matched according to the customer level. This ensures that the review work can cover multiple review items at the same time, and that the review work of multiple customers is carried out in an orderly manner, avoiding excessive waiting time for any customer.

[0086] When the review time is shortened and the number of reviewers is reduced, the review level can be adjusted while the review order is determined. The review level can be adjusted in accordance with the adjustment method in Example 2.

[0087] The survey data consistency audit system and method based on customer interaction services provided in this embodiment can not only ensure the efficient and timely completion of audit tasks according to established standards without affecting other audit results, but also react quickly when external conditions change, dynamically adapt to constantly changing situations, flexibly and automatically adjust audit strategies, maintain the timeliness of audit processing, and enhance the applicability and flexibility of the system. This provides enterprises with more reliable data support and decision-making basis, and promotes continuous improvement of service quality and optimization of service experience.

[0088] The above descriptions are merely embodiments of the present invention. Commonly known structures and characteristics of the solutions are not described in detail here. Those skilled in the art are aware of all common technical knowledge in the field prior to the application date or priority date, are aware of all existing technologies in that field, and have the ability to apply conventional experimental methods prior to that date. Those skilled in the art can, under the guidance of this application, improve and implement this solution in combination with their own capabilities. Some typical known structures or methods should not be obstacles for those skilled in the art to implement this application. It should be noted that those skilled in the art can make several modifications and improvements without departing from the structure of the present invention. These should also be considered within the scope of protection of the present invention, and will not affect the effectiveness of the implementation of the present invention or the practicality of the patent.

Claims

1. A method for verifying the consistency of survey data based on customer interaction services, characterized in that: S100, Obtain at least one audit item data, which includes basic data and several voice samples obtained from the interaction between several visitors and the corresponding survey subjects. S200 parses each speech sample to form corresponding interactive data, which includes multiple questions and their answers; S210, to conduct consistency audit on the process of parsing voice samples to form interactive data, including basic record consistency audit and voice emotion consistency audit; S211, Create a task library to represent the mapping relationship between several review tasks and review items; Based on the basic data of the review items, match and / or update the review tasks in the task library to obtain multiple target review tasks; Based on the multiple target review tasks, perform basic record consistency review of voice samples and corresponding interaction data. S212, Construct an emotion recognition model to identify the emotions of visitors and survey subjects in the voice samples respectively, and conduct voice emotion consistency verification between the voice samples and corresponding interaction data. S300 marks the speech segments with abnormalities in the consistency audit of the speech samples as abnormal, and sends the abnormal speech segments to several review terminals according to the preset distribution method. The review terminals perform secondary review and corresponding review processing based on the review standards. By matching and updating audit projects and tasks, audit strategies can be adjusted to adapt to ever-changing audit requirements. Classify each audit task by importance level; dynamically adjust the importance level of each audit task and / or the audit pass criteria based on audit project data or standard differences; The visitor's emotions are expressed as positive or negative, and the surveyee's emotions are expressed as if there is a possibility that the visitor may have induced the surveyee's emotions, in order to determine whether the consistency is passed.

2. The method for verifying the consistency of survey data based on customer interaction services according to claim 1, characterized in that, The basic data for auditing projects includes quality inspection standards. All audit indicator features in the quality inspection standards are extracted and matched with all audit tasks in the current task library. If a match is found, it is used as the target audit task. If no match is found, the audit task is updated with the corresponding quality inspection standard content and used as the target audit task.

3. The method for verifying the consistency of survey data based on customer interaction services according to claim 1, characterized in that, The update of audit tasks includes the attribution level of audit tasks and their combination arrangement; multiple audit tasks are executed independently to verify the consistency of basic records.

4. The method for verifying the consistency of survey data based on customer interaction services according to claim 1, characterized in that, In S212, the emotion recognition model includes a visitor emotion recognition sub-model, which is used to judge the positive and negative of the visitor's speech segment. If the judgment is negative, the speech emotion consistency check is abnormal, and the corresponding speech segment judged as negative is marked as abnormal.

5. The method for verifying the consistency of survey data based on customer interaction services according to claim 1, characterized in that, In S212, the emotion recognition model includes a survey subject emotion recognition sub-model, which is used to calculate the duration difference between the response duration of each question in the speech segment to which the survey subject belongs and the average response duration of that question. If the duration difference does not meet the threshold requirement, the speech emotion consistency review is abnormal, and the speech segment corresponding to the question and its response is marked as abnormal.

6. The method for verifying the consistency of survey data based on customer interaction services according to claim 5, characterized in that, The average response time for this question is the average of the response times for this question to all respondents corresponding to the same interviewer, or the average of the response times for this question to all respondents.

7. The method for verifying the consistency of survey data based on customer interaction services according to claim 1, characterized in that, S211 includes the following steps: for the same speech sample, use a large language model to extract all answers for each question, repeatedly extract answers several times, determine whether the answers extracted each time are consistent, compare the number of times consistency and inconsistency occur, and determine whether the consistency audit of the basic record has passed.

8. The method for verifying the consistency of survey data based on customer interaction services according to claim 7, characterized in that, If the number of consistent occurrences is greater than the number of non-consistent occurrences, the consistency audit of the basic record will pass; otherwise, the question and its corresponding audio segment will be marked as abnormal.

9. A survey data consistency verification system based on customer interaction services, characterized in that: The system implements the data consistency verification method for surveys based on customer interaction services as described in any one of claims 1-8; the system includes a voice sample acquisition module, a parsing module, a consistency verification module, a secondary verification module, and a verification terminal; The voice sample acquisition module is used to acquire basic data of at least one audit item and several voice samples obtained from the interaction between several visitors and the corresponding survey subjects. The parsing module is used to parse speech samples to form interactive data, which includes multiple questions and their answers. The consistency audit module is used to perform consistency audits on the process of parsing voice samples to form interactive data, including basic record consistency audits and voice emotion consistency audits. The consistency audit module includes a task library module, which creates a task library to represent the mapping relationship between several audit tasks and audit projects; a basic record consistency audit submodule, which matches and / or updates audit tasks in the task library based on the basic data of the audit projects to obtain multiple target audit tasks, and performs basic record consistency audit of voice samples and corresponding interaction data based on multiple target audit tasks; and a voice emotion consistency audit submodule, which builds an emotion recognition model to identify the emotions of visitors and survey subjects in voice samples, and performs voice emotion consistency audit of voice samples and corresponding interaction data. The secondary review module is used to mark the speech segments with abnormalities in the consistency review of the speech samples as abnormal, and send the abnormal speech segments to several review terminals according to the preset distribution method. The review end is used for secondary review and corresponding review processing based on the review standards.

Images