Method, apparatus, and computer-readable storage medium for determining liability in disputes over online car-hailing services

Abstract

The present invention provides a method, device, and computer-readable storage medium for determining liability for online ride-hailing disputes. The method includes the following steps: establishing a neural network model for determining liability for disputes; inputting order, driver, and passenger information into the neural network model to determine liability for disputes. This invention can improve the efficiency and accuracy of determining liability for online ride-hailing disputes.

Description

Technical Field

[0001] The present invention relates to the technical field of information technology, and in particular to a method, device, and computer-readable storage medium for determining liability for disputes involving online ride-hailing services. Background Art

[0002] In the online ride-hailing business, passenger complaints, driver complaints, and disputes between passengers and drivers occur frequently.

[0003] One method of determining liability in related technologies is to assign responsibility to the tail driver with the highest complaint rate. Another method involves determining liability for each complaint. Based on the annotations by human judgment experts, combined with features such as the temporal and spatial information of the trip, basic order information, communication between the driver and passenger, and audio or video recordings of the trip, machine learning is used to model and restore abnormal scenarios, ultimately generating a final determination of liability.

[0004] Currently, the shortcomings of related technologies are as follows. The recall rate for the responsibility determination method that identifies tail drivers with high complaint rates as the driver's responsibility is low, making it difficult to recall and hold accountable part-time drivers and occasional drivers. The responsibility determination method based on manual judgment experts has a low utilization rate of labeled information and does not take into account the intermediate information produced by manual judgment experts during the labeling process. Therefore, a large number of labeled samples are required to obtain a better machine model. In addition, the multi-model judgment method in related technologies is relatively complex, and the model update process is complex and has a long cycle.

[0005] Therefore, it is very necessary to provide a method that can efficiently and accurately determine the liability for online ride-hailing disputes. Summary of the Invention

[0006] The present invention aims to solve at least one of the above technical problems.

[0007] To this end, the first purpose of the present invention is to provide a method for determining dispute liability in online ride-hailing.

[0008] The second purpose of the present invention is to provide a device for determining dispute responsibility for online ride-hailing services.

[0009] A third object of the present invention is to provide a computer-readable storage medium.

[0010] To achieve the first purpose of the present invention, an embodiment of the present invention provides a method for determining dispute liability for online ride-hailing, comprising the following steps: establishing a neural network model for determining dispute liability, inputting order dimension information, driver dimension information, and passenger dimension information into the neural network model to determine dispute liability.

[0011] The method for determining liability in online ride-hailing disputes, provided in this embodiment, aims to efficiently and accurately determine liability in disputes between passengers and drivers. This embodiment inputs information from three dimensions into a neural network model. This neural network model shares the aforementioned order, driver, and passenger characteristics. It then learns and predicts liability attribution, severity, order facts, and complaint reasons in the dispute, thereby improving efficiency and accuracy in determining liability.

[0012] In addition, the technical solutions provided by the above embodiments of the present invention may also have the following additional technical features:

[0013] In the above technical solution, a neural network model for dispute responsibility identification is established, and the order dimension information, driver dimension information and passenger dimension information are input into the neural network model to perform the steps of dispute responsibility identification, including: extracting the text features and table features of the order dimension information, driver dimension information and passenger dimension information respectively as input layer information; encoding the input layer information to obtain coding layer information; converting the coding layer information into a vector to obtain representation layer information; using the representation layer information to perform a prediction task to obtain the result of dispute responsibility identification.

[0014] In this embodiment, the input layer includes information or features across three dimensions: orders, drivers, and passengers. The encoding layer encodes sequential text features and the driver / passenger's historical behavior, and the representation layer outputs a vector. After obtaining the representation layer information, prediction tasks are performed based on this information, leading to the step of determining dispute responsibility.

[0015] In any of the above technical solutions, when executing any prediction task, all the presentation layer information is used to predict the result of dispute responsibility determination.

[0016] When executing each prediction task, this embodiment uses the representation layer information of each dimension to predict the result of the task, thereby improving the efficiency and accuracy of the result prediction.

[0017] In any of the above technical solutions, the prediction tasks include: complaint reason prediction task, responsibility fact prediction task, responsibility attribution prediction task and severity prediction task; the presentation layer information includes: order dimension presentation layer information, driver dimension presentation layer information and passenger dimension presentation layer information.

[0018] In this embodiment, when predicting any of the complaint cause, liability determination, responsibility attribution, and severity tasks, all representation-level information from the three dimensions of order, driver, and passenger is used. This allows the neural network model to share all shallow-level features, enabling multi-task learning and prediction with a single model.

[0019] In any of the above technical solutions, when executing any prediction task, the prediction result output of any prediction task is used as the feature input of any other prediction task, so that any prediction task is correlated with any other prediction task.

[0020] The purpose of this embodiment is to enhance the correlation between various prediction tasks.

[0021] In any of the above technical solutions, the prediction result output of the complaint cause prediction task is used as the feature input of the blame fact prediction task, the responsibility attribution prediction task and the severity prediction task; and / or the prediction result output of the blame fact prediction task is used as the feature input of the responsibility attribution prediction task and the severity prediction task; and / or the prediction result output of the responsibility attribution prediction task is used as the feature input of the severity prediction task.

[0022] In this embodiment, any one of the complaint cause prediction task, the blame fact prediction task, the responsibility attribution prediction task, and the severity prediction task is interconnected or connected in series with any other one.

[0023] In any of the above technical solutions, in the step of using the representation layer information to perform a prediction task to obtain the result of dispute responsibility determination: using an activation function to process the representation layer information to map the representation layer information to the range of 0 to 1, and performing the prediction task to obtain the result of dispute responsibility determination.

[0024] In any of the above technical solutions, the prediction tasks include: complaint cause prediction task, blame fact prediction task, responsibility attribution prediction task and severity prediction task; wherein, the steps of establishing a neural network model for dispute responsibility identification include: based on user feedback data, separately training the complaint cause prediction task, and based on user feedback data, synchronously training the blame fact prediction task, responsibility attribution prediction task and severity prediction task.

[0025] In this embodiment, the complaint cause prediction task is first trained separately. This training process, independent of human expert annotation, utilizes extensive user feedback data to conduct large-scale learning, thereby obtaining initial parameters from the input layer to the representation layer. Furthermore, this embodiment simultaneously trains the task of predicting the fact of fault, the task of predicting responsibility, and the task of predicting severity to initialize the neural network model.

[0026] To achieve the second purpose of the present invention, an embodiment of the present invention provides a device for determining dispute liability for an online ride-hailing vehicle, comprising: a memory storing a computer program; a processor executing the computer program; wherein, when the processor executes the computer program, it implements the steps of the method for determining dispute liability for an online ride-hailing vehicle as described in any embodiment of the present invention.

[0027] The dispute liability determination device for online ride-hailing according to an embodiment of the present invention implements the steps of the dispute liability determination method for online ride-hailing according to any embodiment of the present invention. Therefore, it has all the beneficial effects of the dispute liability determination method for online ride-hailing according to any embodiment of the present invention, and will not be repeated here.

[0028] To achieve the third purpose of the present invention, an embodiment of the present invention provides a computer-readable storage medium, which is characterized in that it includes: the computer-readable storage medium stores a computer program, and when the computer program is executed, it implements the steps of the method for determining dispute liability for online ride-hailing as in any embodiment of the present invention.

[0029] The computer-readable storage medium of an embodiment of the present invention implements the steps of the method for determining dispute liability for an online car-hailing vehicle as in any embodiment of the present invention, and therefore has all the beneficial effects of the method for determining dispute liability for an online car-hailing vehicle as in any embodiment of the present invention, which will not be repeated here.

[0030] Additional aspects and advantages of the invention will become apparent from the description which follows, or may be learned by practice of the invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the following description of the embodiments with reference to the accompanying drawings, in which:

[0032] Figure 1 This is a first flow chart of a method for determining liability for online ride-hailing disputes in some embodiments of the present invention;

[0033] Figure 2 A second flow chart of a method for determining liability for online ride-hailing disputes according to some embodiments of the present invention;

[0034] Figure 3 This is a neural network model for some embodiments of the present invention. DETAILED DESCRIPTION

[0035] In order to more clearly understand the above-mentioned objects, 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, in the absence of conflict, the embodiments of the present application and the features therein can be combined with each other.

[0036] 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 scope of protection of the present invention is not limited to the specific embodiments disclosed below.

[0037] Refer to the following Figures 1 to 3 Describe the technical solutions of some embodiments of the present invention.

[0038] Embodiments of the present invention provide methods, devices, and computer-readable storage media for determining liability in online ride-hailing disputes. The purpose of the methods provided in embodiments of the present invention is to efficiently and accurately determine liability in disputes between online ride-hailing passengers and drivers.

[0039] Specifically, disputes between passengers and drivers frequently arise in the ride-hailing industry. The efficiency and accuracy of liability determination methods used in related technologies are not ideal. Therefore, embodiments of the present invention are intended to improve the efficiency and accuracy of liability determination in ride-hailing disputes.

[0040] Example 1

[0041] like Figure 1 As shown, this embodiment provides a method for determining liability for online car-hailing disputes, which includes the following steps:

[0042] Step S102: Establishing a neural network model for dispute responsibility determination;

[0043] Step S104. Input the order dimension information, driver dimension information and passenger dimension information into the neural network model to determine the dispute responsibility.

[0044] Dispute liability in this embodiment refers to complaints filed by drivers or users with the ride-hailing platform, or disputes between drivers and users. These complaints can be fare or service complaints. For example, passenger complaints include charges for missed rides, excessive surcharges, premature billing, delayed billing termination, fare increases, detours, and overcharges. Complaints can also be filed by drivers regarding unexplained cancellations or unpaid fares.

[0045] The dispute responsibility determination in this embodiment refers to the process conducted by the online ride-hailing platform to determine the party responsible for the dispute. The responsible party can be the driver, the passenger, or the online ride-hailing platform,

[0046] In this embodiment, a neural network model is established and information of three dimensions, namely, orders, drivers, and passengers, is simultaneously input into the neural network model to determine the responsibility for disputes.

[0047] Compared with the dispute responsibility determination method of establishing multiple models separately in related technologies, this embodiment inputs information of three dimensions into a neural network model. The neural network model shares the above three order, driver and passenger characteristics, thereby learning and predicting the responsibility attribution, severity, order facts and complaint reasons in the dispute, thereby improving the efficiency and accuracy of responsibility.

[0048] It should be noted that the responsibility in this embodiment can be attributed to any one of the driver, passenger or online ride-hailing platform, and the severity includes malicious responsibility, negligence responsibility or responsibility for providing false information.

[0049] Example 2

[0050] like Figure 2 As shown, in addition to the technical features of the above-mentioned embodiment 1, this embodiment further includes the following technical features.

[0051] The steps of establishing a neural network model for dispute responsibility identification and inputting order dimension information, driver dimension information, and passenger dimension information into the neural network model for dispute responsibility identification include:

[0052] Step S202: Extract text features and table features of order dimension information, driver dimension information, and passenger dimension information as input layer information;

[0053] Step S204: Encode the input layer information to obtain encoding layer information;

[0054] Step S206: Convert the coding layer information into a vector to obtain the representation layer information;

[0055] Step S208: Use the presentation layer information to perform the prediction task to obtain the result of dispute responsibility determination.

[0056] Specifically, in this embodiment, the input layer includes information or features across three dimensions: orders, drivers, and passengers. The encoding layer encodes sequential text features and driver / passenger historical behavior, and the representation layer outputs a vector. After obtaining the representation layer information, prediction tasks are performed based on this information, leading to the step of determining dispute responsibility.

[0057] Example 3

[0058] In addition to the technical features of the above-mentioned embodiment 2, this embodiment further includes the following technical features.

[0059] When executing any prediction task, all the presentation layer information is used to predict the results of dispute responsibility determination.

[0060] That is, the number of prediction tasks is at least two. In step S208 of embodiment 2, when executing each prediction task, the representation layer information of each dimension is used to predict the result of the task, thereby improving the efficiency and accuracy of the result prediction.

[0061] Example 4

[0062] In addition to the technical features of any of the above embodiments, this embodiment further includes the following technical features.

[0063] The prediction tasks include: complaint reason prediction task, judgment fact prediction task, responsibility attribution prediction task and severity prediction task; the presentation layer information includes: order dimension presentation layer information, driver dimension presentation layer information and passenger dimension presentation layer information.

[0064] Specifically, in this embodiment, the complaint reason prediction task involves predicting the reasons for passenger or driver complaints, such as complaints about fares or service issues. Examples of complaint reasons include, but are not limited to, charges for missed rides, overcharging, premature billing, untimely billing termination, fare increases, and overcharging due to detours. These issues are represented in a structured format, making them directly accessible during accountability assessment or responsibility determination. In this embodiment, the accountability fact prediction task involves identifying or restoring the facts of an incident. A single complaint or dispute may involve multiple facts. For example, a dispute may involve multiple facts, such as a full ride, a mid-trip drop-off, an intoxicated passenger, or a passenger giving directions. Labels for these facts are obtained through manual accountability expert annotation. In this embodiment, the responsibility attribution prediction task involves determining whether responsibility belongs to the passenger or the driver. Responsibility can be attributed to any of the following: the driver, the passenger, or the ride-hailing platform, and is specifically annotated by manual accountability experts. In this embodiment, the severity prediction task refers to predicting the severity of a dispute or incident, where the severity includes malicious responsibility, error responsibility, or responsibility for providing false information, etc.

[0065] In this embodiment, when predicting any of the complaint cause, liability determination, responsibility attribution, and severity tasks, all representation-level information from the three dimensions of order, driver, and passenger is used. This allows the neural network model to share all shallow-level features, enabling multi-task learning and prediction with a single model.

[0066] Example 5

[0067] In addition to the technical features of any of the above embodiments, this embodiment further includes the following technical features.

[0068] When executing any prediction task, the prediction result output of any prediction task is used as the feature input of another prediction task, so that any prediction task is correlated with another prediction task.

[0069] The purpose of this embodiment is to enhance the correlation between various prediction tasks. For example, any one of the complaint cause prediction task, the blame fact prediction task, the responsibility attribution prediction task, and the severity prediction task is interconnected or connected in series with any other. In particular, the output of any prediction task can be used as the input of any other prediction task and directly merged through the upper connection layer (i.e., the concat layer).

[0070] Example 6

[0071] In addition to the technical features of any of the above embodiments, this embodiment further includes the following technical features.

[0072] Use the prediction result output of the complaint reason prediction task as the feature input of the blame fact prediction task, the responsibility attribution prediction task and the severity prediction task; and / or use the prediction result output of the blame fact prediction task as the feature input of the responsibility attribution prediction task and the severity prediction task; and / or use the prediction result output of the responsibility attribution prediction task as the feature input of the severity prediction task.

[0073] Specifically, in this embodiment, Task 1 is the complaint reason prediction task, Task 2 is the blame fact prediction task, Task 3 is the responsibility attribution prediction task, and Task 4 is the severity prediction task. Task 1 is connected in series with Task 2, Task 3, and Task 4, respectively, and serves as the input features for Task 2, Task 3, and Task 4. Task 2 is connected in series with Task 3 and Task 4, respectively, and serves as the input features for Task 3 and Task 4. Task 3 is connected in series with Task 4, and serves as the input features for Task 4.

[0074] Example 7

[0075] In addition to the technical features of any of the above embodiments, this embodiment further includes the following technical features.

[0076] In the step of using the representation layer information to perform a prediction task to obtain a result of dispute responsibility determination: the representation layer information is processed using an activation function to map the representation layer information to a range of 0 to 1, and the prediction task is performed to obtain a result of dispute responsibility determination. In this embodiment, before performing each prediction task, the concat layer is first used to merge the information, and then the activation function (i.e., sigmoid function) is used for processing.

[0077] Example 8

[0078] In addition to the technical features of any of the above embodiments, this embodiment further includes the following technical features.

[0079] The prediction tasks include: complaint cause prediction task, blame fact prediction task, responsibility attribution prediction task and severity prediction task; among them, the steps of establishing a neural network model for dispute responsibility identification include: based on user feedback data, separate training of complaint cause prediction task, and based on user feedback data, simultaneous training of blame fact prediction task, responsibility attribution prediction task and severity prediction task.

[0080] In this embodiment, the complaint cause prediction task is first trained separately. During this training process, it does not rely on manual expert labeling. Instead, it conducts large-scale learning based on a large amount of user feedback data to obtain the initial parameters from the input layer to the representation layer. Furthermore, this embodiment simultaneously trains the task of predicting the fact of judgment, the task of predicting the attribution of responsibility, and the task of predicting the severity to initialize the neural network model. Specifically, this embodiment uses the labeled data of manual expert judgment to learn the parameters of the upper layer and uses a linear weighted approach to calculate the model loss. The weights of different tasks are equal.

[0081] Example 9

[0082] This embodiment provides a device for determining liability for online ride-hailing disputes, comprising a memory and a processor. The memory stores a computer program. The processor executes the computer program. When executing the computer program, the processor implements the steps of the method for determining liability for online ride-hailing disputes as described in any embodiment of the present invention.

[0083] Example 10

[0084] This embodiment provides a computer-readable storage medium, including: the computer-readable storage medium stores a computer program, and when the computer program is executed, it implements the steps of the method for determining dispute liability for online ride-hailing as in any embodiment of the present invention. Specific embodiments

[0086] This embodiment provides a method for determining liability in online ride-hailing disputes. This embodiment is suitable for determining liability in passenger complaints, driver complaints, and disputes between passengers and drivers.

[0087] In this embodiment, a neural network model for dispute responsibility determination is first established and trained. Then, order dimension information, driver dimension information, and passenger dimension information are input into the neural network model to determine dispute responsibility.

[0088] Among them, such as Figure 3As shown, this embodiment provides a structure of a neural network model, which performs a total of four prediction tasks including complaint cause prediction task, blame fact prediction task, responsibility attribution prediction task and severity prediction task to determine dispute responsibility. This embodiment adopts a single-model multi-task learning modeling method, so that the machine model can simultaneously learn and predict responsibility attribution, severity, order facts, complaint causes, etc. Among them, the neural network model shares shallow features, and the complaint cause prediction task, blame fact prediction task, responsibility attribution prediction task and severity prediction task are connected in series. The prediction output of one task is used as the feature input of another task to increase the correlation between tasks. In this embodiment, the input layer information of the neural network structure includes feature information of three dimensions: order, passenger, and driver. The encoding layer encodes the sequence text features and the driver / passenger historical sequence behavior. The final output representation layer information is a vector.

[0089] In summary, the beneficial effects of the embodiments of the present invention are:

[0090] 1. The embodiments of the present invention improve the efficiency of determining liability in online ride-hailing disputes by establishing a multi-task neural network model structure.

[0091] 2. The neural network model established and used in the embodiment of the present invention is based on a single-model, multi-task approach and can solve complex complaint judgment and prediction problems.

[0092] 3. The embodiments of the present invention are trained based on user feedback data, ensuring the accuracy of dispute responsibility determination.

[0093] Throughout this specification, terms such as "one embodiment," "some embodiments," and "specific embodiments" mean that the specific features, structures, materials, or characteristics described in conjunction with that embodiment or example are included in at least one embodiment or example of the present invention. In this specification, schematic representations of these terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0094] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

Claims

1. A method for determining liability for online ride-hailing disputes, characterized by: The following steps are involved: Establish a neural network model for dispute liability determination; Inputting order dimension information, driver dimension information, and passenger dimension information into the neural network model, performing a prediction task, and determining the responsible party, which includes the driver, passenger, or online ride-hailing platform, including: Extracting text features and table features of the order dimension information, the driver dimension information, and the passenger dimension information respectively as input layer information; Encoding the input layer information to obtain encoding layer information; the encoding layer is encoding the sequence text features and the driver / passenger historical sequence behavior; Converting the coding layer information into a vector to obtain representation layer information; the representation layer outputs the vector; Using the representation layer information, executing a prediction task to determine the responsible party includes merging the representation layer information of all prediction tasks through a contact layer, processing the merged representation layer information using an activation function to map the representation layer information to a range from 0 to 1, and executing the prediction task to obtain the responsible party; The prediction tasks include: complaint cause prediction tasks, blame fact prediction tasks, severity prediction tasks, and responsibility attribution prediction tasks; when executing any of the prediction tasks, the prediction result output of any of the prediction tasks is used as the feature input of any of the prediction tasks, and is directly merged through the connection layer to make any of the prediction tasks mutually associated with any of the prediction tasks, including: The prediction result output of the complaint cause prediction task serves as the feature input of the responsibility fact prediction task, the severity prediction task, and the responsibility attribution prediction task; The prediction result output of the responsibility fact prediction task is used as the feature input of the severity prediction task and the responsibility attribution prediction task; Using the prediction result output of the responsibility attribution prediction task as the feature input of the severity prediction task; The steps of establishing a neural network model for dispute responsibility determination include: based on user feedback data, separately training the complaint cause prediction task, and based on user feedback data, synchronously training the responsibility fact prediction task, the responsibility attribution prediction task and the severity prediction task; during the synchronous training, the labeled data of the manual expert judgment is used to learn the upper-layer parameters, and the loss of the neural network model is calculated using a linear weighted method; wherein, the weights of different prediction tasks are equal.

2. The method for determining liability for online car-hailing disputes according to claim 1, characterized in that: When executing any of the prediction tasks, all of the presentation layer information is used to predict the result of the dispute responsibility determination.

3. The method for determining liability for online car-hailing disputes according to claim 2, characterized in that: The presentation layer information includes: order dimension presentation layer information, driver dimension presentation layer information and passenger dimension presentation layer information.

4. A device for determining liability for disputes in online ride-hailing services, characterized in that: include: a memory storing a computer program; a processor, configured to execute the computer program; Wherein, when executing the computer program, the processor implements the steps of the method for determining dispute liability of online car-hailing as described in any one of claims 1 to 3.

5. A computer-readable storage medium, characterized in that include: The computer-readable storage medium stores a computer program, and when the computer program is executed, the steps of the method for determining dispute liability for online car-hailing as described in any one of claims 1 to 3 are implemented.

Images