Complaint handling method, handling system, computer equipment and storage medium

By identifying the appeal scenario and using processing strategies and deep neural network models to judge the authenticity of the appeal information of online ride-hailing drivers, the problem of false appeal identification in the existing technology is solved, the recognition efficiency and accuracy are improved, and the cost of manual review is reduced.

CN111861501BActive Publication Date: 2024-11-26BEIJING DIDI INFINITY TECH & DEV CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN201911290485.1
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2019-12-16
Publication Date
2024-11-26
Estimated Expiration
2040-11-26

AI Technical Summary

Technical Problem

The prior art is difficult to accurately identify whether the complaints of online ride-hailing drivers are false, resulting in unfair treatment of rewards and punishments, and the cost of manual verification is high and the efficiency is low.

Method used

By obtaining order information and appeal information, identifying the appeal scenario, using the processing strategies and pre-store rules or deep neural network models corresponding to the appeal scenario to judge the authenticity of the appeal information, and comparing it with voice or video information to determine whether the appeal information is true.

Benefits of technology

It realizes accurate judgment of appeal information, saves labor costs and time, improves the effectiveness of appeal information identification, and avoids misjudgment of false appeals.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN111861501B_ABST
    Figure CN111861501B_ABST
Patent Text Reader

Abstract

The present invention proposes a complaint handling method, a handling system, a computer device and a storage medium. The complaint handling method includes: obtaining order information and complaint information; identifying the complaint scenario according to the complaint information, and obtaining a handling strategy corresponding to the complaint scenario; determining whether the complaint information is valid according to the handling strategy and order information. The technical solution of the present invention can accurately judge the falsity of the complaint information in combination with the complaint scenario, without the need for manual customer service to review the complaint information, saving labor costs and time, and improving the effectiveness of identifying the complaint information.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a complaint handling method, a complaint handling system, a computer device, and a computer-readable storage medium. Background Art

[0002] In order to create a good online car-hailing travel experience environment, online car-hailing platforms will formulate a series of rules and reward and punishment measures, such as: when matching drivers and passengers to form an order, if the driver cancels the order, the driver's service points will be deducted. If the driver does not agree with a certain reward or punishment, he can file a complaint with the platform. If the driver's complaint is in line with the actual situation, the reward or punishment can be exempted. However, the driver's complaint may not be in line with the actual situation, that is, there may be false complaints. When the driver's complaint is false, the platform should not be exempted from reward and punishment. It is important to identify whether the driver's complaint is false.

[0003] The related technology includes the following two identification methods:

[0004] (1) A comprehensive credit rating method based on the driver's historical file information evaluates the driver's credit rating by the number of completed orders, number of canceled orders, number of complaints, number of negative reviews from passengers, etc. If the driver's credit score is too low, the driver's complaint is considered false. The driver's credit reflects the possibility of the driver engaging in improper behavior. It is determined by historical behavior information, but it is impossible to determine the driver's behavior for a single order. This leads to two major problems: there is a lack of evidence to evaluate the user's future behavior based on the driver's history, and the driver will not recognize the rewards and punishments he or she has received in a single order; it is impossible to evaluate drivers without historical information (new drivers).

[0005] (2) Manual verification method based on order information viewing: the platform establishes a tool to view order information, which is checked by manual customer service to determine whether the driver has described false information. This method has the following two main problems: since manual verification requires thinking and judging of various information, it is difficult to consider multiple issues, such as the driver inducing passengers to cancel orders by phone. If the manual verification only checks the basic information of the order without listening to the call recording, key information will be missed; the cost is high. Due to the long time of video and call, manual verification takes a long time. Therefore, when the order volume is large, the labor cost required is high. Summary of the invention

[0006] The present invention aims to solve at least one of the technical problems existing in the prior art or related art.

[0007] To this end, one aspect of the present invention is to propose a complaint handling method.

[0008] Another aspect of the present invention is to provide a complaint handling system.

[0009] Another aspect of the present invention provides a computer device.

[0010] Yet another aspect of the present invention provides a computer-readable storage medium.

[0011] In view of this, according to one aspect of the present invention, a complaint handling method is proposed, including: obtaining order information and complaint information; identifying the complaint scenario based on the complaint information, and obtaining the processing strategy corresponding to the complaint scenario; determining whether the complaint information is valid based on the processing strategy and order information.

[0012] The complaint handling method provided by the present invention obtains the complaint information selected and filled in by the driver, identifies the complaint scenario according to the complaint information, and different complaint scenarios correspond to different handling strategies. The handling strategy corresponding to the complaint scenario is used to determine the consistency between the complaint information and the actual facts of the order information to determine whether the complaint information is established. The technical solution of the present invention can accurately determine the falsity of the complaint information in combination with the complaint scenario, without the need for manual customer service to review the complaint information, saving labor costs and time, and improving the effectiveness of identifying the complaint information.

[0013] The above-mentioned complaint handling method according to the present invention may also have the following technical features:

[0014] The above technical solution also includes: when it is determined that the appeal information is not valid, the reason why the appeal information is not valid is prompted.

[0015] In this technical solution, when it is determined that the complaint information is not valid, the reason why the complaint information is not valid will be prompted so that the driver and manual customer service can know the criteria for judgment.

[0016] In any of the above technical solutions, the step of determining whether the complaint information is valid based on the processing strategy and order information specifically includes: obtaining voice or video information in the order information; and using a processing strategy of comparing the complaint information with the voice or video information to determine whether the complaint information is valid.

[0017] In this technical solution, in some scenarios, voice or video information can be used to identify the authenticity of complaint information. For example, in a complaint scenario where the driver complains that the passenger is overloaded and therefore cannot complete the service, by identifying whether there are keywords in the voice in the order information, the passenger mentions that there are x people traveling together (x is greater than the maximum number of people n), or identifying whether there are more than n people in the video in the order information. If all are not met, the driver's complaint is identified as false. The reason why the complaint information is not established can be explained as follows: the voice or video at a certain time, minute, and second shows that the content of the driver's complaint is not established.

[0018] In any of the above technical solutions, the step of determining whether the complaint information is valid based on the processing strategy and order information specifically includes: using the processing strategy to determine whether the order information complies with pre-stored rules to determine whether the complaint information is valid.

[0019] In this technical solution, in some scenarios, the truth of the complaint information can be judged by using pre-stored rules, where the pre-stored rules are rules summarized based on expert experience, and the pre-stored rules may include that the distance cannot exceed the distance threshold, the time cannot exceed the time threshold, etc. For example, in the complaint scenario where the driver complains that the dispatch distance (i.e. the straight-line distance or road distance from the driver's order pick-up point to the passenger boarding point) is too far and the service cannot be completed, it is determined whether the dispatch distance exceeds y kilometers (excluding special cases such as elevated roads and intersections). If it does not exceed y kilometers, the driver's complaint is identified as false. The reason why the complaint information is not established can be explained as follows: according to a certain rule, the content of the driver's complaint is not established.

[0020] In any of the above technical solutions, the step of determining whether the complaint information is valid based on the processing strategy and order information specifically includes: determining whether the complaint information is valid based on the order information using a complaint processing model.

[0021] In this technical solution, for other types of scenarios that mainly have the following problems, a pre-trained complaint handling model can be used to determine whether the complaint information is valid. The problems include: the scenario is complex, and it is difficult to identify false information through certain voice and video recognition or formulating rules; it is difficult to explain the reasons for identifying false information in complex scenarios; in addition to considering whether the complaint is false, it is also necessary to determine whether the driver's complaint can be approved. Whether it can be approved refers to whether the driver can be exempted from punishment based on the complaint information. False complaints will definitely not be approved, and real complaints may not be approved. In other words, for complex scenarios, the complaint handling model can be used to determine whether the complaint information is valid, thereby improving the effectiveness of false complaint information identification. The reason why the complaint information is not valid can be the reason item with the highest probability of determining that the driver's complaint will not be approved.

[0022] In any of the above technical solutions, it also includes: obtaining sampling order information and sampling complaint information corresponding to the sampling order information; parsing the sampling complaint information to obtain the sampling complaint label and sampling complaint additional information; using the sampling order information, sampling complaint label and sampling complaint additional information to establish a complaint handling model.

[0023] In this technical solution, an accurate complaint handling model is established, and the complaint handling model can be a deep neural network model for multi-task learning. In the feature part of multi-task learning, it mainly includes order features and complaint features. For order features, it mainly includes time, distance, coordinates, trajectory, congestion, navigation and other features to characterize the true situation of this order. The complaint features include complaint labels and complaint filling information (complaint filling information includes text information and / or image information). The above features are deeply trained at the same time, and the main tasks are whether the complaint is passed, whether the complaint item is established (that is, whether it is false), and the reason for the driver's judgment.

[0024] In any of the above technical solutions, the step of identifying the complaint scenario based on the complaint information specifically includes: parsing the complaint information to obtain the complaint label and the complaint additional information; and identifying the complaint scenario using a scenario recognition model based on the complaint label and the complaint additional information.

[0025] In this technical solution, the complaint information includes a complaint label and complaint additional information. The complaint label is the complaint option selected by the driver by clicking in the terminal application. The complaint additional information is the information additionally filled in by the driver, which may include text information and / or image information. Based on the complaint label and complaint additional information, the scene recognition model is used to identify the complaint scene, thereby providing a processing basis for identifying the authenticity of the complaint information.

[0026] In any of the above technical solutions, it also includes: obtaining sampling appeal information; parsing the sampling appeal information to obtain sampling appeal labels and sampling appeal additional information; annotating the sampling appeal labels with scenes; and establishing a scene recognition model using the sampling appeal additional information and the sampling appeal labels with scene annotations.

[0027] In this technical solution, the establishment of a scene recognition model mainly includes two steps: the first step is to map the scene corresponding to the complaint label selected by the driver, specifically including: (1) mapping the same type of labels to the same scene; (2) correcting the labels selected incorrectly by the driver; (3) for the part where the driver selects the label "other", mapping it to the corresponding scene according to the complaint reason added by the driver. The second step is text classification or image classification, and the complaint reason added by the driver can be text information or image information. Using the complaint label as a marker, using a text classification algorithm, the text information added by the driver is preprocessed, feature represented, model trained, and model evaluated, so as to train a multi-classification text classification model, namely, a scene recognition model. Among them, the text classification algorithm can adopt an end-to-end deep learning model, such as TextCNN (text convolutional neural network), TestRNN (test recurrent neural network), RCNN (candidate region convolutional neural network), Bert (semantic representation model), etc. Or through image recognition, image classification is performed, using the complaint label as a marker, and a multi-classification image classification model, namely, a scene recognition model, is obtained through an image classification algorithm. Among them, the image classification algorithm can adopt GoogleNet (a deep learning structure), ResNet (Resigual Neural Network), DenseNet (Dense Neural Network), etc.

[0028] According to another aspect of the present invention, a complaint handling system is proposed, including: an information acquisition unit, used to obtain order information and complaint information; a scene recognition unit, used to identify the complaint scene according to the complaint information, and obtain the processing strategy corresponding to the complaint scene; a judgment unit, used to determine whether the complaint information is valid according to the processing strategy and order information.

[0029] In the complaint handling system provided by the present invention, the information acquisition unit acquires the complaint information selected and filled in by the driver, the scene recognition unit recognizes the complaint scene according to the complaint information, different complaint scenes correspond to different processing strategies, and the judgment unit uses the processing strategy corresponding to the complaint scene to determine the consistency between the complaint information and the actual facts of the order information to determine whether the complaint information is established. The technical solution of the present invention can accurately judge the falsity of the complaint information in combination with the complaint scene, without the need for manual customer service to review the complaint information, saving labor costs and time, and improving the effectiveness of complaint information recognition.

[0030] The above complaint handling system according to the present invention may also have the following technical features:

[0031] The above technical solution also includes: a prompting unit, which is used to prompt the reason why the appeal information is not established when it is determined that the appeal information is not established.

[0032] In this technical solution, when it is determined that the complaint information is not valid, the prompt unit will prompt the reason why the complaint information is not valid, so that the driver and the manual customer service can know the criteria for judgment.

[0033] In any of the above technical solutions, the judgment unit is specifically used to obtain voice or video information in the order information, and to use a processing strategy of comparing the complaint information with the voice or video information to determine whether the complaint information is valid.

[0034] In this technical solution, in some scenarios, voice or video information can be used to identify the authenticity of complaint information. For example, in a complaint scenario where the driver complains that the passenger is overloaded and therefore cannot complete the service, by identifying whether there are keywords in the voice in the order information, the passenger mentions that there are x people traveling together (x is greater than the maximum number of people n), or identifying whether there are more than n people in the video in the order information. If all are not met, the driver's complaint is identified as false. The reason why the complaint information is not established can be explained as follows: the voice or video at a certain time, minute, and second shows that the content of the driver's complaint is not established.

[0035] In any of the above technical solutions, the judgment unit is specifically used to determine whether the complaint information is valid by using a processing strategy to determine whether the order information complies with pre-stored rules.

[0036] In this technical solution, in some scenarios, the truth of the complaint information can be judged by using pre-stored rules, where the pre-stored rules are rules summarized based on expert experience, and the pre-stored rules may include that the distance cannot exceed the distance threshold, the time cannot exceed the time threshold, etc. For example, in the complaint scenario where the driver complains that the dispatch distance (i.e. the straight-line distance or road distance from the driver's order pick-up point to the passenger boarding point) is too far and the service cannot be completed, it is determined whether the dispatch distance exceeds y kilometers (excluding special cases such as elevated roads and intersections). If it does not exceed y kilometers, the driver's complaint is identified as false. The reason why the complaint information is not established can be explained as follows: according to a certain rule, the content of the driver's complaint is not established.

[0037] In any of the above technical solutions, the judgment unit is specifically used to determine whether the complaint information is valid based on the order information and using the complaint handling model.

[0038] In this technical solution, for other types of scenarios that mainly have the following problems, a pre-trained complaint handling model can be used to determine whether the complaint information is valid. The problems include: the scenario is complex, and it is difficult to identify false information through certain voice and video recognition or formulating rules; it is difficult to explain the reasons for identifying false information in complex scenarios; in addition to considering whether the complaint is false, it is also necessary to determine whether the driver's complaint can be approved. Whether it can be approved refers to whether the driver can be exempted from punishment based on the complaint information. False complaints will definitely not be approved, and real complaints may not be approved. In other words, for complex scenarios, the complaint handling model can be used to determine whether the complaint information is valid, thereby improving the effectiveness of false complaint information identification. The reason why the complaint information is not valid can be the reason item with the highest probability of determining that the driver's complaint will not be approved.

[0039] In any of the above technical solutions, it also includes: a first model building unit, used to obtain sampling order information and sampling complaint information corresponding to the sampling order information; parse the sampling complaint information to obtain the sampling complaint label and sampling complaint additional information; use the sampling order information, sampling complaint label and sampling complaint additional information to establish a complaint handling model.

[0040] In this technical solution, the first model building unit is used to build an accurate complaint handling model, and the complaint handling model can be a deep neural network model of multi-task learning. In the feature part of multi-task learning, it mainly includes order features and complaint features. For order features, it mainly includes time, distance, coordinates, trajectory, congestion, navigation and other features to characterize the true situation of this order. The complaint features include complaint labels and complaint filling information (complaint filling information includes text information and / or image information). The above features are deeply trained at the same time, and the main tasks are whether the complaint is passed, whether the complaint item is established (that is, whether it is false), and the reason for the driver's judgment.

[0041] In any of the above technical solutions, the scene recognition unit is specifically used to parse the complaint information, obtain the complaint label and the complaint supplementary information, and use the scene recognition model to identify the complaint scene based on the complaint label and the complaint supplementary information.

[0042] In this technical solution, the complaint information includes a complaint label and complaint additional information. The complaint label is the complaint option selected by the driver by clicking in the terminal application. The complaint additional information is the information additionally filled in by the driver, which may include text information and / or image information. Based on the complaint label and complaint additional information, the scene recognition model is used to identify the complaint scene, thereby providing a processing basis for identifying the authenticity of the complaint information.

[0043] In any of the above technical solutions, it also includes: a second model building unit, which is used to obtain sampling appeal information; parse the sampling appeal information to obtain sampling appeal labels and sampling appeal additional information; perform scene annotation on the sampling appeal labels; and establish a scene recognition model using the sampling appeal additional information and the sampling appeal labels after scene annotation.

[0044] In this technical solution, a scene recognition model is established by using a second model building unit, which mainly includes two steps: the first step is to map scenes corresponding to the complaint labels selected by the driver, specifically including: (1) mapping labels of the same type to the same scene; (2) correcting labels selected incorrectly by the driver; (3) for the part where the driver selects the label "other", mapping it to the corresponding scene according to the complaint reasons added by the driver. The second step is text classification or image classification, and the complaint reasons added by the driver can be text information or image information. Using the complaint labels as tags, using text classification algorithms, the text information added by the driver is preprocessed, feature represented, model trained, and model evaluated, so as to train a multi-classification text classification model, namely, a scene recognition model. Among them, the text classification algorithm can adopt an end-to-end deep learning model, such as TextCNN, TestRNN, RCNN, Bert, etc. Or by image recognition, image classification is performed, using the complaint labels as tags, and a multi-classification image classification model, namely, a scene recognition model, is obtained by an image classification algorithm. Among them, the image classification algorithm can adopt GoogleNet, ResNet, DenseNet, etc.

[0045] According to another aspect of the present invention, a computer device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements a complaint handling method as described in any of the above technical solutions when executing the computer program.

[0046] The computer device provided by the present invention implements the steps of the complaint handling method of any of the above technical solutions when the processor executes the computer program, so the computer device includes all the beneficial effects of the complaint handling method of any of the above technical solutions.

[0047] According to another aspect of the present invention, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the complaint handling method of any of the above technical solutions is implemented.

[0048] The computer-readable storage medium provided by the present invention implements the steps of the complaint handling method of any of the above-mentioned technical solutions when the computer program is executed by the processor. Therefore, the computer-readable storage medium includes all the beneficial effects of the complaint handling method of any of the above-mentioned technical solutions.

[0049] Additional aspects and advantages of the present invention will become apparent from the following description or may be learned by practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] The above and / or additional aspects and advantages of the present invention will become apparent and easily understood from the description of the embodiments in conjunction with the following drawings, in which:

[0051] Figure 1 A schematic diagram showing a flow chart of a complaint handling method according to a first embodiment of the present invention;

[0052] Figure 2 A schematic flow chart showing a method for handling complaints according to a second embodiment of the present invention;

[0053] Figure 3 A schematic diagram showing a flow chart of a complaint handling method according to a third embodiment of the present invention;

[0054] Figure 4 A schematic diagram showing a flow chart of a complaint handling method according to a fourth embodiment of the present invention;

[0055] Figure 5 A schematic flow chart of a false recognition method of expression feedback according to a fifth embodiment of the present invention is shown;

[0056] Figure 6 A schematic diagram showing false identification categories according to an embodiment of the present invention;

[0057] Figure 7 A structural diagram of a deep neural network model for multi-task learning according to an embodiment of the present invention is shown;

[0058] Figure 8 A schematic block diagram showing a complaint handling system according to a first embodiment of the present invention;

[0059] Fig. 9 A schematic block diagram showing a complaint handling system according to a second embodiment of the present invention;

[0060] Fig.10 A schematic block diagram showing a complaint handling system according to a third embodiment of the present invention;

[0061] Fig.11 A schematic block diagram of a complaint handling system according to a fourth embodiment of the present invention is shown. DETAILED DESCRIPTION

[0062] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present invention and the features in the embodiments can be combined with each other without conflict.

[0063] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below.

[0064] The embodiment of the first aspect of the present invention provides a complaint handling method, which is described in detail through the following embodiments.

[0065] Embodiment 1, Figure 1 The flowchart of the complaint handling method of the first embodiment of the present invention is shown. The method includes:

[0066] Step 102, obtaining order information and complaint information;

[0067] Step 104, identifying the complaint scenario according to the complaint information, and obtaining a processing strategy corresponding to the complaint scenario;

[0068] Step 106, determining whether the complaint information is valid based on the processing strategy and order information.

[0069] The complaint handling method provided by the present invention obtains the complaint information selected and filled in by the driver, identifies the complaint scenario according to the complaint information, and different complaint scenarios correspond to different handling strategies. The handling strategy corresponding to the complaint scenario is used to determine the consistency between the complaint information and the actual facts of the order information to determine whether the complaint information is established. The technical solution of the present invention can accurately determine the falsity of the complaint information in combination with the complaint scenario, without the need for manual customer service to review the complaint information, saving labor costs and time, and improving the effectiveness of identifying the complaint information.

[0070] In any of the above technical solutions, step 106, based on the processing strategy and order information, determines whether the complaint information is valid, specifically including: obtaining the voice or video information in the order information; and using a processing strategy of comparing the complaint information with the voice or video information to determine whether the complaint information is valid.

[0071] In this technical solution, in some scenarios, voice or video information can be used to identify the authenticity of complaint information. For example, in a complaint scenario where the driver complains that the passenger is overloaded and therefore cannot complete the service, by identifying whether there are keywords in the voice in the order information, the passenger mentions that there are x people traveling together (x is greater than the maximum number of people n), or identifying whether there are more than n people in the video in the order information. If all are not met, the driver's complaint is identified as false. The reason why the complaint information is not established can be explained as follows: the voice or video at a certain time, minute, and second shows that the content of the driver's complaint is not established.

[0072] In any of the above technical solutions, step 106, determining whether the complaint information is valid based on the processing strategy and order information, specifically includes: using the processing strategy of judging whether the order information complies with the pre-stored rules to determine whether the complaint information is valid.

[0073] In this technical solution, in some scenarios, the truth of the complaint information can be judged by using pre-stored rules, where the pre-stored rules are rules summarized based on expert experience, and the pre-stored rules may include that the distance cannot exceed the distance threshold, the time cannot exceed the time threshold, etc. For example, in the complaint scenario where the driver complains that the dispatch distance (i.e. the straight-line distance or road distance from the driver's order pick-up point to the passenger boarding point) is too far and the service cannot be completed, it is determined whether the dispatch distance exceeds y kilometers (excluding special cases such as elevated roads and intersections). If it does not exceed y kilometers, the driver's complaint is identified as false. The reason why the complaint information is not established can be explained as follows: according to a certain rule, the content of the driver's complaint is not established.

[0074] In any of the above technical solutions, step 106, determining whether the complaint information is valid based on the processing strategy and order information, specifically includes: determining whether the complaint information is valid based on the order information using a complaint processing model.

[0075] In this technical solution, for other types of scenarios that mainly have the following problems, a pre-trained complaint handling model can be used to determine whether the complaint information is valid. The problems include: the scenario is complex, and it is difficult to identify false information through certain voice and video recognition or formulating rules; it is difficult to explain the reasons for identifying false information in complex scenarios; in addition to considering whether the complaint is false, it is also necessary to determine whether the driver's complaint can be approved. Whether it can be approved refers to whether the driver can be exempted from punishment based on the complaint information. False complaints will definitely not be approved, and real complaints may not be approved. In other words, for complex scenarios, the complaint handling model can be used to determine whether the complaint information is valid, thereby improving the effectiveness of false complaint information identification. The reason why the complaint information is not valid can be the reason item with the highest probability of determining that the driver's complaint will not be approved.

[0076] Embodiment 2: Figure 2 The flowchart of the complaint handling method of the second embodiment of the present invention is shown. The method includes:

[0077] Step 202, obtaining order information and complaint information;

[0078] Step 204, identifying the complaint scenario according to the complaint information, and obtaining a processing strategy corresponding to the complaint scenario;

[0079] Step 206, determining whether the appeal information is valid based on the processing strategy and order information, and if it is determined that the appeal information is not valid, indicating the reason why the appeal information is not valid.

[0080] In this technical solution, when it is determined that the complaint information is not valid, the reason why the complaint information is not valid will be prompted so that the driver and manual customer service can know the criteria for judgment.

[0081] In any of the above technical solutions, in step 206, it is determined whether the complaint information is valid based on the processing strategy and order information, specifically including: obtaining voice or video information in the order information; and using a processing strategy that compares the complaint information with the voice or video information to determine whether the complaint information is valid.

[0082] In any of the above technical solutions, in step 206, whether the complaint information is valid is determined based on the processing strategy and order information, specifically including: using the processing strategy of judging whether the order information complies with pre-stored rules to determine whether the complaint information is valid.

[0083] In any of the above technical solutions, in step 206, determining whether the complaint information is valid based on the processing strategy and order information specifically includes: determining whether the complaint information is valid based on the order information using a complaint processing model.

[0084] Embodiment 3, Figure 3 The flowchart of the complaint handling method of the third embodiment of the present invention is shown. The method includes:

[0085] Step 302, obtaining sampling order information and sampling complaint information corresponding to the sampling order information; parsing the sampling complaint information, obtaining a sampling complaint label and sampling complaint supplementary information; using the sampling order information, the sampling complaint label and the sampling complaint supplementary information, establishing a complaint processing model;

[0086] Step 304, obtaining order information and complaint information;

[0087] Step 306, identifying the complaint scenario according to the complaint information, and obtaining a processing strategy corresponding to the complaint scenario;

[0088] Step 308, based on the order information, using the complaint processing model to determine whether the complaint information is valid, and if it is determined that the complaint information is not valid, the reason why the complaint information is not valid is prompted.

[0089] In this technical solution, an accurate complaint handling model is established, and the complaint handling model can be a deep neural network model for multi-task learning. In the feature part of multi-task learning, it mainly includes order features and complaint features. For order features, it mainly includes time, distance, coordinates, trajectory, congestion, navigation and other features to characterize the true situation of this order. The complaint features include complaint labels and complaint filling information (complaint filling information includes text information and / or image information). The above features are deeply trained at the same time, and the main tasks are whether the complaint is passed, whether the complaint item is established (that is, whether it is false), and the reason for the driver's judgment.

[0090] Embodiment 4: Figure 4 The flowchart of the complaint handling method according to the fourth embodiment of the present invention is shown. The method includes:

[0091] Step 402, obtaining sampling appeal information; parsing the sampling appeal information, obtaining sampling appeal labels and sampling appeal supplementary information; annotating the sampling appeal labels with scenes; and establishing a scene recognition model using the sampling appeal supplementary information and the sampling appeal labels with scene annotations;

[0092] Step 404, obtaining order information and complaint information;

[0093] Step 406, parsing the complaint information, obtaining the complaint label and the complaint supplementary information; and identifying the complaint scene using the scene recognition model according to the complaint label and the complaint supplementary information;

[0094] Step 408, obtaining a processing strategy corresponding to the complaint scenario;

[0095] Step 410, determining whether the complaint information is valid based on the processing strategy and order information, and if it is determined that the complaint information is not valid, indicating the reason why the complaint information is not valid.

[0096] In this technical solution, the establishment of a scene recognition model mainly includes two steps: the first step is to map the scene corresponding to the complaint label selected by the driver, specifically including: (1) mapping the same type of labels to the same scene; (2) correcting the labels selected incorrectly by the driver; (3) for the part where the driver selects the label "other", it is mapped to the corresponding scene according to the complaint reason added by the driver. The second step is text classification or image classification, and the complaint reason added by the driver can be text information or image information. Using the complaint label as a marker, using a text classification algorithm, the text information added by the driver is preprocessed, feature represented, model trained, and model evaluated, so as to train a multi-classification text classification model, namely, a scene recognition model. Among them, the text classification algorithm can adopt an end-to-end deep learning model, such as TextCNN, TestRNN, RCNN, Bert, etc. Or through image recognition, image classification is performed, using the complaint label as a marker, and a multi-classification image classification model, namely, a scene recognition model, is obtained through an image classification algorithm. Among them, the image classification algorithm can adopt GoogleNet, ResNet, DenseNet, etc.

[0097] When identifying a complaint scenario, the complaint information includes a complaint label and complaint additional information. The complaint label is the complaint option that the driver clicks and selects in the terminal application. The complaint additional information is the information additionally filled in by the driver, which may include text information and / or image information. Based on the complaint label and complaint additional information, the scene recognition model is used to identify the complaint scenario, thereby providing a processing basis for identifying the authenticity of the complaint information.

[0098] In any of the above technical solutions, in step 410, it is determined whether the complaint information is valid based on the processing strategy and order information, specifically including: obtaining voice or video information in the order information; and using a processing strategy that compares the complaint information with the voice or video information to determine whether the complaint information is valid.

[0099] In any of the above technical solutions, in step 410, it is determined whether the complaint information is valid based on the processing strategy and order information, specifically including: using the processing strategy of judging whether the order information complies with the pre-stored rules to determine whether the complaint information is valid.

[0100] In any of the above technical solutions, in step 410, determining whether the complaint information is valid based on the processing strategy and order information specifically includes: determining whether the complaint information is valid based on the order information using a complaint processing model.

[0101] It should be noted that when using the complaint handling model to determine whether the complaint information is valid, it is necessary to establish the complaint handling model in advance. The steps of establishing the complaint handling model include: obtaining sampling order information and sampling complaint information corresponding to the sampling order information; parsing the sampling complaint information, obtaining the sampling complaint label and sampling complaint additional information; using the sampling order information, sampling complaint label and sampling complaint additional information to establish the complaint handling model.

[0102] Embodiment 5, Figure 5 A schematic flow chart of a false recognition method for expression feedback according to a fifth embodiment of the present invention is shown. The method comprises:

[0103] 1. Expression feedback scene recognition

[0104] The driver's complaint data within a preset time period is selected from the database. The complaint data mainly includes various detailed information: the driver's ID (Identity Document, unique code), the order ID related to the complaint, the complaint time, the complaint label selected by the driver, and the complaint reason manually filled in by the driver. In the expression feedback scene recognition, there are mainly two sub-steps: label relationship mapping and text classification.

[0105] (1) Label relationship mapping

[0106] Mapping is performed according to the labels selected by the driver. This mainly includes: mapping labels of the same type to the same reason; correcting the parts where the driver selected the wrong label; and mapping the parts where the driver selected the label "other" to the correct label according to the complaint reason manually filled in by the driver.

[0107] (2) Text Classification

[0108] Using the complaint label as the row marker, the text classification algorithm is used to preprocess, represent features, train the model, and evaluate the reasons for the complaint manually filled in by the driver, so as to train a multi-classification text classification model. Among them, the text classification algorithm can use an end-to-end deep learning model, such as TextCNN, TestRNN, RCNN, Bert, etc.

[0109] In another embodiment, the complaint data includes picture evidence provided by the driver. Image classification is mainly performed by image recognition. Similarly, using the complaint label as a marker, a multi-classification image classification model is obtained by an image classification algorithm. Among them, the image classification algorithm can use GoogleNet, ResNet, DenseNet, etc.

[0110] 2. Offline training of false identification model

[0111] According to the expression feedback scenario obtained by the above recognition, the basic order information of the corresponding order is obtained from the database, mainly including: order ID, driver ID, passenger ID, order creation time, order acceptance time, pick-up distance, driver's entire journey coordinate point / time, passenger's entire journey coordinate point / time, driver and passenger call information, driver and passenger IM information, driver and passenger trip recording / video, etc. Figure 6 As shown, according to the expression feedback scenario, it is divided into three false recognition categories: speech and video recognition judgment, rule judgment, and multi-task learning model judgment.

[0112] (1) Voice / video recognition

[0113] For voice and video recognition, it is mainly done through voice recognition keywords and key image recognition in the video. For example, if the driver complains that the passenger is overloaded and therefore cannot complete the service, in this scenario, by identifying whether there are keywords in the voice, the passenger mentions that there are x people in the group (x is greater than the maximum number of people n), or identifying whether there are more than n people in the video, if all are not met, then the driver's feedback is identified as false.

[0114] For the recognition of this part, the explanation output reason is: the voice / video display at a certain time, minute and second shows that the content complained is not valid.

[0115] (2) Rule Identification

[0116] For the scenarios in the rule part, a rule-based method for identifying false feedback is used based on the summary of expert experience. For example, if a driver complains that the dispatch distance (i.e., the straight-line distance and road distance from the driver's pickup point to the passenger's boarding point) is too far, the rule is used to determine whether the dispatch distance exceeds y kilometers (excluding special cases such as elevated roads and intersections). If the driver does not hit the rule, it is considered that the driver has made false feedback.

[0117] For the identification of this part, the reason for the explanation output is: according to certain rules, the content complained is not valid.

[0118] (3) Multi-task learning model recognition

[0119] For other types of scenarios, the following problems exist: the scenarios are complex, and it is difficult to identify falsehoods through certain voice and video recognition or formulating rules; it is difficult to explain the reasons for identifying falsehoods in complex scenarios; in addition to considering whether the complaint is false, it is also necessary to determine whether the driver's complaint can be approved (considering the cognitive differences of drivers, there are cases where drivers use the reasons for complaints incorrectly, but they should not be punished). Therefore, a deep neural network model for multi-task learning is constructed, and the network structure of the model is as follows: Figure 7As shown in the figure: In the feature part of multi-task learning, the main features include order features and complaint features. For order features, they mainly include time, distance, coordinates, trajectory, congestion, navigation and other features to describe the true situation of this order. Complaint features include complaint tags and complaint texts. After embedding, the Attention layer is added. The above features are deeply trained at the same time. The main task is whether the complaint is passed, and whether the complaint item is established (that is, whether it is false) and the driver's judgment reason are introduced as auxiliary tasks.

[0120] For this part of the identification, the prediction result of the driver's decision reason (ie, the reason item with the highest probability that the driver's appeal will not be approved) is output.

[0121] 3. Model launch

[0122] Deploy the trained model online, that is, deploy the scene recognition model and the false recognition model. When the driver initiates a complaint to express feedback, the scene recognition model is first used to identify the complaint scene based on the driver's selection reason and input text; then the scene is classified and different false recognition models are selected for different scenes. The online system will make subsequent judgments in real time based on the output results of the model.

[0123] 4. Result output

[0124] The output of the false identification model gives a judgment on whether the driver made a false complaint. At the same time, the output of the false identification model includes the corresponding explanation and is output to the customer service, who then transparently transmits the reason why the driver was judged as false.

[0125] The false feedback expression identification method proposed in the present invention utilizes the complaint label and the complaint filled-in text to identify the complaint scene through natural language processing technology, and divides it into different categories according to different scenes, and uses voice / video recognition, rules, and multi-task learning deep neural network methods to identify the authenticity of user expression feedback.

[0126] The embodiment of the second aspect of the present invention provides a complaint handling system, which is described in detail through the following embodiments.

[0127] Embodiment 1, Figure 8 FIG. 8 is a schematic block diagram of a complaint handling system 800 according to a first embodiment of the present invention. The complaint handling system 800 includes:

[0128] Information acquisition unit 802, used to acquire order information and complaint information;

[0129] A scenario identification unit 804 is used to identify the complaint scenario according to the complaint information and obtain a processing strategy corresponding to the complaint scenario;

[0130] The judgment unit 806 is used to determine whether the complaint information is valid according to the processing strategy and order information.

[0131] In the complaint handling system provided by the present invention, the information acquisition unit 802 acquires the complaint information selected and filled in by the driver, the scene recognition unit 804 recognizes the complaint scene according to the complaint information, different complaint scenes correspond to different processing strategies, and the judgment unit 806 uses the processing strategy corresponding to the complaint scene to determine the consistency between the complaint information and the actual facts of the order information to determine whether the complaint information is established. The technical solution of the present invention can accurately judge the falsity of the complaint information in combination with the complaint scene, without the need for manual customer service to review the complaint information, saving labor costs and time, and improving the effectiveness of complaint information recognition.

[0132] In any of the above technical solutions, the judgment unit 806 is specifically used to obtain the voice or video information in the order information, and use a processing strategy of comparing the complaint information with the voice or video information to determine whether the complaint information is valid.

[0133] In this technical solution, in some scenarios, voice or video information can be used to identify the authenticity of complaint information. For example, in a complaint scenario where the driver complains that the passenger is overloaded and therefore cannot complete the service, by identifying whether there are keywords in the voice in the order information, the passenger mentions that there are x people traveling together (x is greater than the maximum number of people n), or identifying whether there are more than n people in the video in the order information. If all are not met, the driver's complaint is identified as false. The reason why the complaint information is not established can be explained as follows: the voice or video at a certain time, minute, and second shows that the content of the driver's complaint is not established.

[0134] In any of the above technical solutions, the judgment unit 806 is specifically used to determine whether the complaint information is valid by using a processing strategy for judging whether the order information complies with pre-stored rules.

[0135] In this technical solution, in some scenarios, the truth of the complaint information can be judged by using pre-stored rules, where the pre-stored rules are rules summarized based on expert experience, and the pre-stored rules may include that the distance cannot exceed the distance threshold, the time cannot exceed the time threshold, etc. For example, in the complaint scenario where the driver complains that the dispatch distance (i.e. the straight-line distance or road distance from the driver's order pick-up point to the passenger boarding point) is too far and the service cannot be completed, it is determined whether the dispatch distance exceeds y kilometers (excluding special cases such as elevated roads and intersections). If it does not exceed y kilometers, the driver's complaint is identified as false. The reason why the complaint information is not established can be explained as follows: according to a certain rule, the content of the driver's complaint is not established.

[0136] In any of the above technical solutions, the judgment unit 806 is specifically used to determine whether the complaint information is valid based on the order information and using the complaint handling model.

[0137] In this technical solution, for other types of scenarios that mainly have the following problems, a pre-trained complaint handling model can be used to determine whether the complaint information is valid. The problems include: the scenario is complex, and it is difficult to identify false information through certain voice and video recognition or formulating rules; it is difficult to explain the reasons for identifying false information in complex scenarios; in addition to considering whether the complaint is false, it is also necessary to determine whether the driver's complaint can be approved. Whether it can be approved refers to whether the driver can be exempted from punishment based on the complaint information. False complaints will definitely not be approved, and real complaints may not be approved. In other words, for complex scenarios, the complaint handling model can be used to determine whether the complaint information is valid, thereby improving the effectiveness of false complaint information identification. The reason why the complaint information is not valid can be the reason item with the highest probability of determining that the driver's complaint will not be approved.

[0138] Embodiment 2: Fig. 9 FIG. 8 is a schematic block diagram of a complaint handling system 800 according to a second embodiment of the present invention. The complaint handling system 800 includes:

[0139] Information acquisition unit 802, used to acquire order information and complaint information;

[0140] A scenario identification unit 804 is used to identify the complaint scenario according to the complaint information and obtain a processing strategy corresponding to the complaint scenario;

[0141] A judgment unit 806, used to determine whether the complaint information is valid according to the processing strategy and order information;

[0142] The prompting unit 808 is used to prompt the reason why the appeal information is not established when it is determined that the appeal information is not established.

[0143] In this technical solution, when it is determined that the complaint information is not valid, the prompt unit 808 will prompt the reason why the complaint information is not valid, so that the driver and the manual customer service can know the criteria for judgment.

[0144] In any of the above technical solutions, the judgment unit 806 is specifically used to obtain the voice or video information in the order information, and use a processing strategy of comparing the complaint information with the voice or video information to determine whether the complaint information is valid.

[0145] In any of the above technical solutions, the judgment unit 806 is specifically used to determine whether the complaint information is valid by using a processing strategy for judging whether the order information complies with pre-stored rules.

[0146] In any of the above technical solutions, the judgment unit 806 is specifically used to determine whether the complaint information is valid based on the order information and using the complaint handling model.

[0147] Embodiment 3, Fig.10 FIG. 8 is a schematic block diagram of a complaint handling system 800 according to a third embodiment of the present invention. The complaint handling system 800 includes:

[0148] Information acquisition unit 802, used to acquire order information and complaint information;

[0149] A scenario identification unit 804 is used to identify the complaint scenario according to the complaint information and obtain a processing strategy corresponding to the complaint scenario;

[0150] A judgment unit 806, configured to determine whether the complaint information is valid based on the order information and using the complaint handling model;

[0151] A prompting unit 808, for prompting the reason why the appeal information is not established when it is determined that the appeal information is not established;

[0152] The first model building unit 810 is used to obtain sampling order information and sampling complaint information corresponding to the sampling order information; parse the sampling complaint information to obtain the sampling complaint label and sampling complaint additional information; and use the sampling order information, sampling complaint label and sampling complaint additional information to establish a complaint handling model.

[0153] In this technical solution, the first model building unit 810 is used to build an accurate complaint handling model, and the complaint handling model can be a deep neural network model of multi-task learning. In the feature part of multi-task learning, it mainly includes order features and complaint features. For order features, it mainly includes time, distance, coordinates, trajectory, congestion, navigation and other features to characterize the true situation of this order. The complaint features include complaint labels and complaint filling information (complaint filling information includes text information and / or image information). The above features are deeply trained at the same time, and the main tasks are whether the complaint is passed, whether the complaint item is established (that is, whether it is false), and the reason for the driver's judgment.

[0154] Embodiment 4: Fig.11 FIG. 8 is a schematic block diagram of a complaint handling system 800 according to a fourth embodiment of the present invention. The complaint handling system 800 includes:

[0155] Information acquisition unit 802, used to acquire order information and complaint information;

[0156] The scene recognition unit 804 is used to parse the complaint information, obtain the complaint label and the complaint supplementary information, and use the scene recognition model to identify the complaint scene according to the complaint label and the complaint supplementary information, and obtain the processing strategy corresponding to the complaint scene;

[0157] A judgment unit 806, used to determine whether the complaint information is valid according to the processing strategy and order information;

[0158] A prompting unit 808, for prompting the reason why the appeal information is not established when it is determined that the appeal information is not established;

[0159] The second model building unit 812 is used to obtain sampling appeal information; parse the sampling appeal information to obtain sampling appeal labels and sampling appeal additional information; perform scene annotation on the sampling appeal labels; and establish a scene recognition model using the sampling appeal additional information and the scene annotated sampling appeal labels.

[0160] In this technical solution, the second model building unit 812 is used to build a scene recognition model, which mainly includes two steps: the first step is to map the scene corresponding to the complaint label selected by the driver, specifically including: (1) mapping the same type of labels to the same scene; (2) correcting the labels selected incorrectly by the driver; (3) for the part where the driver selects the label "other", according to the complaint reason added by the driver, it is mapped to the corresponding scene. The second step is text classification or image classification, and the complaint reason added by the driver can be text information or image information. Using the complaint label as a marker, using the text classification algorithm, the text information added by the driver is preprocessed, feature represented, model trained, and model evaluated, so as to train a multi-classification text classification model, that is, a scene recognition model. Among them, the text classification algorithm can adopt an end-to-end deep learning model, such as TextCNN, TestRNN, RCNN, Bert, etc. Or by image recognition, image classification is performed, using the complaint label as a marker, and a multi-classification image classification model, that is, a scene recognition model, is obtained by the image classification algorithm. Among them, the image classification algorithm can adopt GoogleNet, ResNet, DenseNet, etc.

[0161] In this technical solution, when identifying a complaint scenario, the complaint information includes a complaint label and complaint additional information. The complaint label is the complaint option selected by the driver by clicking in the terminal application. The complaint additional information is the information additionally filled in by the driver, which may include text information and / or image information. Based on the complaint label and complaint additional information, the scene recognition model is used to identify the complaint scenario, thereby providing a processing basis for identifying the authenticity of the complaint information.

[0162] It should be noted that, in the fourth embodiment, a first model building unit 810 may also be included.

[0163] An embodiment of the third aspect of the present invention proposes a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements a complaint handling method as described in any of the above technical solutions when executing the computer program.

[0164] The computer device provided by the present invention implements the steps of the complaint handling method of any of the above technical solutions when the processor executes the computer program, so the computer device includes all the beneficial effects of the complaint handling method of any of the above technical solutions.

[0165] An embodiment of the fourth aspect of the present invention provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the complaint handling method of any of the above technical solutions is implemented.

[0166] The computer-readable storage medium provided by the present invention implements the steps of the complaint handling method of any of the above-mentioned technical solutions when the computer program is executed by the processor. Therefore, the computer-readable storage medium includes all the beneficial effects of the complaint handling method of any of the above-mentioned technical solutions.

[0167] In the description of this specification, the terms "first" and "second" are used for descriptive purposes only and cannot be understood as indicating or implying relative importance unless otherwise clearly specified and limited; the terms "connection", "installation", "fixation" and the like should be understood in a broad sense, for example, "connection" can be a fixed connection, a detachable connection, or an integral connection; it can be directly connected or indirectly connected through an intermediate medium. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0168] In the description of this specification, the description of the terms "one embodiment", "some embodiments", "specific embodiments", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0169] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for handling complaints, characterized in that: include: Obtain order information and complaint information; the complaint information includes a complaint label and complaint additional information; The complaint information includes text information and / or image information; Identify the complaint scenario according to the complaint information, and obtain a processing strategy corresponding to the complaint scenario; Determining whether the complaint information is valid according to the processing strategy and the order information; The step of identifying the complaint scenario according to the complaint information specifically includes: Parsing the complaint information to obtain the complaint label and the complaint additional information; According to the complaint label and the complaint supplementary information, using a scene recognition model to identify the complaint scene; The steps of establishing the scene recognition model include: Obtain sampling complaint information; Parse the sampling complaint information to obtain the sampling complaint label and sampling complaint supplementary information; Performing scene annotation on the sampled complaint labels; Using the sampling complaint filled information and the sampling complaint label after scene annotation, establish the scene recognition model; The step of labeling the sample complaint label with a scene includes: Mapping the sampled complaint labels of the same type to the same scene; Correct the incorrectly selected sampling complaint label; If the sample complaint label is "other", the sample complaint label is mapped to the corresponding scenario according to the additional complaint reason; The step of establishing the scene recognition model by using the sampling complaint filling information and the sampling complaint label after scene annotation includes: If the sample complaint added information is text information, the scene recognition model is obtained by training with a text classification algorithm according to the sample complaint label annotated with the usage scenario; If the sampling complaint added information is image information, the scene recognition model is obtained by training with an image classification algorithm according to the sampling complaint label annotated with the usage scenario.

2. The method for handling complaints according to claim 1, characterized in that: Also includes: When it is determined that the complaint information is not valid, the reason why the complaint information is not valid is prompted.

3. The complaint handling method according to claim 1, characterized in that: The step of determining whether the complaint information is valid according to the processing strategy and the order information specifically includes: Obtaining the audio or video information in the order information; A processing strategy of comparing the complaint information with the voice or video information is used to determine whether the complaint information is valid.

4. The complaint handling method according to claim 1, characterized in that: The step of determining whether the complaint information is valid according to the processing strategy and the order information specifically includes: A processing strategy is used to determine whether the order information complies with pre-stored rules to determine whether the complaint information is valid.

5. The method for handling complaints according to claim 1, characterized in that: The step of determining whether the complaint information is valid according to the processing strategy and the order information specifically includes: Based on the order information, a complaint handling model is used to determine whether the complaint information is valid.

6. The method for handling complaints according to claim 5, characterized in that: Also includes: Obtaining sampling order information and sampling complaint information corresponding to the sampling order information; Parse the sampling complaint information to obtain the sampling complaint label and sampling complaint supplementary information; The complaint handling model is established using the sampling order information, the sampling complaint label and the sampling complaint supplementary information.

7. A complaint handling system, characterized in that: include: An information acquisition unit, used to acquire order information and complaint information; the complaint information includes a complaint label and complaint additional information; The complaint information includes text information and / or image information; A scene recognition unit, used to identify the complaint scene according to the complaint information, and obtain a processing strategy corresponding to the complaint scene; A judgment unit, used to determine whether the complaint information is valid according to the processing strategy and the order information; A scene recognition unit, specifically used to parse the complaint information, obtain a complaint label and complaint supplementary information, and identify the complaint scene using a scene recognition model according to the complaint label and the complaint supplementary information; The second model building unit is used to obtain sampling appeal information; parse the sampling appeal information to obtain sampling appeal labels and sampling appeal additional information; Performing scene annotation on the sampling complaint label; establishing the scene recognition model using the sampling complaint filled-in information and the sampling complaint label after scene annotation; The second model building unit is used to map the same type of sampling appeal labels to the same scene; correct the incorrectly selected sampling appeal label; if the sampling appeal label is "other", map the sampling appeal label to the corresponding scene according to the additional appeal reason; if the sampling appeal added information is text information, use the text classification algorithm to train according to the sampling appeal label annotated with the usage scene to obtain the scene recognition model; if the sampling appeal added information is image information, use the image classification algorithm to train according to the sampling appeal label annotated with the usage scene to obtain the scene recognition model.

8. The complaint handling system according to claim 7, characterized in that: Also includes: The prompting unit is used to prompt the reason why the complaint information is not established when it is determined that the complaint information is not established.

9. The complaint handling system according to claim 7, characterized in that: The judgment unit is specifically used to obtain the voice or video information in the order information, and use a processing strategy of comparing the complaint information with the voice or video information to determine whether the complaint information is valid.

10. The complaint handling system according to claim 7, characterized in that: The judgment unit is specifically used to determine whether the complaint information is valid by using a processing strategy for judging whether the order information complies with a pre-stored rule.

11. The complaint handling system according to claim 7, characterized in that: The judgment unit is specifically used to determine whether the complaint information is valid based on the order information and using the complaint handling model.

12. The complaint handling system according to claim 11, characterized in that: Also includes: The first model building unit is used to obtain sampling order information and sampling complaint information corresponding to the sampling order information; parse the sampling complaint information to obtain sampling complaint labels and sampling complaint filling information; and use the sampling order information, the sampling complaint labels and the sampling complaint filling information to establish the complaint handling model.

13. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the complaint handling method according to any one of claims 1 to 6 is implemented.

14. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the complaint handling method according to any one of claims 1 to 6 is implemented.