Model training methods, responsibility determination methods, devices, equipment, media and products

Abstract

The disclosed embodiment discloses a model training method, a responsibility determination method, an apparatus, a device, a medium and a product, the method comprising: obtaining voice call data of a call between a user party and a service party before an order operation occurs and a responsibility determination result corresponding to the order operation; performing speech recognition on the voice call data to obtain a speech recognition text; determining at least one target sentence in the speech recognition text; obtaining a responsibility determination model, and taking at least one target sentence as input and the responsibility determination result as output, training the responsibility determination model to obtain a target responsibility determination model. The target responsibility determination model obtained by the scheme can obtain the responsibility determination result, and according to the responsibility determination result, it can be determined whether one of the user party and the service party actively triggers the order operation by inducing the other party to actively trigger the order operation through voice, which helps to improve the accuracy of the party determined to be responsible and improve the user experience.

Description

Technical Field

[0001] The present disclosure relates to the field of data processing technology, and in particular to a model training method, a responsibility determination method, a device, equipment, a medium and a product. Background Art

[0002] With the development of mobile Internet technology, people can book corresponding services online in their daily lives, such as food delivery services, errand services, travel services, etc. In the process of using the above services, if the corresponding order has been formed, if it is necessary to perform corresponding order operations on the formed order, such as canceling the order, modifying the order service address, modifying the order service time, etc., it will increase the cost of the order during the execution process. Therefore, the service user needs to determine whether the user should bear the responsibility for the corresponding order operation or the service provider should bear the responsibility for the corresponding order operation.

[0003] In the related art, when an order operation is triggered, it can be determined whether the order operation is triggered by the user or by the service provider, and it can be determined that the party that triggers the order operation bears the responsibility for the order operation.

[0004] Although the above scheme can determine the party between the user and the service provider who bears the responsibility for the corresponding order operation, considering that in some scenarios, a call may occur between the user and the service provider, during the call, one of the user and the service provider may induce the other party to actively trigger the corresponding order operation, thereby avoiding assuming the corresponding responsibility themselves, resulting in the inability to accurately determine the party that should bear the corresponding responsibility, thereby damaging the user experience. Summary of the invention

[0005] In order to solve the problems in the related art, the embodiments of the present disclosure provide a model training method, a responsibility determination method, an apparatus, a device, a medium and a product.

[0006] In a first aspect, an embodiment of the present disclosure provides a model training method, the method comprising:

[0007] Acquire voice call data of a call between the user party and the service provider before an order operation on an order between the user party and the service provider occurs, and a responsibility determination result corresponding to the order operation;

[0008] Performing speech recognition on the voice call data to obtain speech recognition text;

[0009] Determine at least one target sentence in the speech recognition text, wherein the distance between the target sentence and the target scene word is less than or equal to the scene word distance, and the word frequency of the target operation inducing word in the target sentence is greater than or equal to the inducing word frequency threshold;

[0010] A responsibility determination model is obtained, and at least one target sentence is used as input, and a responsibility determination result is used as output, and the responsibility determination model is trained to obtain a target responsibility determination model.

[0011] In combination with the first aspect, in a first implementation of the first aspect of the present disclosure, obtaining voice call data includes:

[0012] Receive authorization indication information, and acquire voice call data in response to the authorization indication information.

[0013] In combination with the first aspect, in a second implementation of the first aspect of the present disclosure, before obtaining the responsibility determination result corresponding to the order operation, the method further includes:

[0014] Obtain the accountability results corresponding to the order operation;

[0015] Obtain the responsibility determination results corresponding to the order operation, including:

[0016] In response to the fact that the responsible party indicated by the responsibility determination result is the service party, and the service party has not sent any dispute arbitration request information within the responsibility determination time length after the time when the responsibility determination result is generated, a responsibility determination result is generated to indicate that the responsible party for the order operation is the service party.

[0017] In combination with the first aspect, in a third implementation of the first aspect of the present disclosure, before taking at least one target sentence as input and taking the responsibility determination result as output and training the responsibility determination model to obtain the target responsibility determination model, the method further includes:

[0018] Receiving the updated weight parameters sent by the edge server, and updating the responsibility determination model according to the updated weight parameters;

[0019] Taking at least one target sentence as input and taking the responsibility determination result as output, training the responsibility determination model to obtain the target responsibility determination model includes:

[0020] Taking at least one target sentence as input and taking the responsibility determination result as output, training the updated responsibility determination model;

[0021] When the trained responsibility determination model has not converged, a gradient update vector is obtained according to the trained responsibility determination model, and the gradient update vector is sent to the edge server;

[0022] When the trained responsibility determination model converges, the trained responsibility determination model is determined as the target responsibility determination model.

[0023] In a second aspect, an embodiment of the present disclosure provides a method for determining responsibility, the method comprising:

[0024] Acquire voice call data of a call between the user party and the service party before an order operation for a corresponding order between the user party and the service party occurs;

[0025] Performing speech recognition on the voice call data to obtain speech recognition text;

[0026] Determine at least one target sentence in the speech recognition text, wherein the distance between the target sentence and the target scene word is less than or equal to the scene word distance, and the word frequency of the target operation inducing word in the target sentence is greater than or equal to the inducing word frequency threshold;

[0027] Obtain a target training responsibility determination model, and take at least one target statement as input to obtain a responsibility determination result corresponding to the order operation output by the target training responsibility determination model.

[0028] In combination with the second aspect, in a first implementation of the second aspect of the present disclosure, the target training responsibility determination model is obtained through the model training method of any one of the first aspect, the first implementation of the first aspect to the third implementation.

[0029] In combination with the second aspect, in a second implementation of the second aspect of the present disclosure, the responsibility determination method is applied to an online responsibility determination system, and the target training responsibility determination model is obtained by training based on a text convolutional neural network model (textcnn);

[0030] Or, the responsibility determination method is applied to an offline responsibility determination system, and the target training responsibility determination model includes a first target training responsibility determination model obtained by training based on TextCNN and a second target training responsibility determination model obtained by training based on a bidirectional encoding representation model (BERT) of a converter;

[0031] The method also includes:

[0032] A combined responsibility determination result is obtained based on a first responsibility determination result output by the first target training responsibility determination model and a second responsibility determination result output by the second target training responsibility determination model.

[0033] In a third aspect, an embodiment of the present disclosure provides a method for determining responsibility, wherein the method includes:

[0034] Receive responsibility determination result indication information and display the responsibility determination result indication information, the responsibility determination result indication information is used to indicate the responsibility determination result corresponding to the order operation, and the responsibility determination result is obtained through any one of the methods from the second aspect, the first implementation method of the second aspect to the second implementation method.

[0035] In combination with the third aspect, in a first implementation manner of the third aspect of the present disclosure, before receiving the responsibility determination result indication information and displaying the responsibility determination result indication information, the method further includes:

[0036] In response to the input order operation action, order operation instruction information is sent, and the order operation instruction information is used to instruct the order operation to be performed on the order between the user party and the service party.

[0037] In combination with the third aspect, in a second implementation manner of the third aspect of the present disclosure, before receiving the responsibility determination result indication information and displaying the responsibility determination result indication information, the method further includes:

[0038] Sending voice call data of a call between the user party and the service party before an order operation for an order between the user party and the service party occurs.

[0039] In combination with the second implementation of the third aspect, in a third implementation of the third aspect of the present disclosure, before receiving the responsibility determination result indication information and displaying the responsibility determination result indication information, the method further includes:

[0040] In response to the input authorization indication action, authorization indication information is sent, where the authorization indication information is used to instruct the acquisition of voice call data.

[0041] In a fourth aspect, an embodiment of the present disclosure provides a model training device, the device comprising:

[0042] A training data acquisition module is configured to acquire voice call data of a call between the user party and the service party before an order operation for an order between the user party and the service party occurs and a responsibility determination result corresponding to the order operation;

[0043] A first speech recognition module is configured to perform speech recognition on the voice call data to obtain speech recognition text;

[0044] A first sentence determination module is configured to determine at least one target sentence in the speech recognition text, wherein the distance between the target sentence and the target scene word is less than or equal to the scene word distance, and the word frequency of the target operation inducing word in the target sentence is greater than or equal to the inducing word frequency threshold;

[0045] The model training module is configured to obtain a responsibility determination model, take at least one target statement as input, take a responsibility determination result as output, train the responsibility determination model, and obtain a target responsibility determination model.

[0046] In a fifth aspect, an embodiment of the present disclosure provides a responsibility determination device, the device comprising:

[0047] A judgment data acquisition module is configured to acquire voice call data of a call between the user party and the service party before an order operation for a corresponding order between the user party and the service party occurs;

[0048] A second speech recognition module is configured to perform speech recognition on the voice call data to obtain speech recognition text;

[0049] a second sentence determination module configured to determine at least one target sentence in the speech recognition text, wherein the distance between the target sentence and the target scene word is less than or equal to the scene word distance, and the word frequency of the target operation inducing word in the target sentence is greater than or equal to the inducing word frequency threshold;

[0050] The determination result acquisition module is configured to obtain a target training responsibility determination model and take at least one target statement as input to obtain a responsibility determination result corresponding to the order operation output by the target training responsibility determination model.

[0051] In a sixth aspect, an embodiment of the present disclosure provides a responsibility determination device, the device comprising:

[0052] The determination result display module is configured to receive responsibility determination result indication information and display the responsibility determination result indication information. The responsibility determination result indication information is used to indicate the responsibility determination result corresponding to the order operation. The responsibility determination result is obtained through any one of the methods from the second aspect, the first implementation method of the second aspect to the second implementation method.

[0053] In the seventh aspect, an electronic device is provided in an embodiment of the present disclosure, comprising a memory and at least one processor; the memory is used to store one or more computer instructions, and the one or more computer instructions are executed by at least one processor to implement the method steps of any one of the first aspect, the first implementation method of the first aspect to the third implementation method, the second aspect, the first implementation method of the second aspect to the second implementation method, the third aspect, and the first implementation method of the third aspect to the third implementation method.

[0054] In an eighth aspect, a computer-readable storage medium is provided in an embodiment of the present disclosure, on which computer instructions are stored. When the computer instructions are executed by a processor, the method steps of any one of the first aspect, the first implementation method to the third implementation method of the first aspect, the second aspect, the first implementation method to the second implementation method of the second aspect, the third aspect, and the first implementation method to the third implementation method of the third aspect are implemented.

[0055] In the ninth aspect, a computer program product is provided in an embodiment of the present disclosure, comprising a computer program / instruction, which, when executed by a processor, implements the method steps of any one of the first aspect, the first implementation manner to the third implementation manner of the first aspect, the second aspect, the first implementation manner to the second implementation manner of the second aspect, the third aspect, and the first implementation manner to the third implementation manner of the third aspect.

[0056] The technical solution provided by the embodiments of the present disclosure may have the following beneficial effects:

[0057] In the above-mentioned embodiment, by obtaining voice call data of a call between the user party and the service party before the order operation for the order between the user party and the service party occurs and the responsibility determination result corresponding to the order operation; performing speech recognition on the voice call data to obtain speech recognition text, determining at least one target sentence in the speech recognition text, the distance between the target sentence and the target scene word is less than or equal to the scene word distance, and the word frequency of the target operation inducing word in the target sentence is greater than or equal to the inducing word frequency threshold, that is, the shorter target sentence may include the content of communication between the user party and the service party regarding the target scene corresponding to the order operation, and the target sentence may include the content of one party of the user party and the service party inducing the other party to actively trigger the order operation through voice communication; obtaining a responsibility determination model, and taking at least one target sentence as input and the responsibility determination result as output, the responsibility determination model is trained to obtain a target responsibility determination model. The target responsibility determination model obtained by the above scheme can obtain the corresponding responsibility determination results based on the sentences extracted from the voice call data, so as to determine whether there is any content in the call between the user and the service party that induces the other party to actively trigger an order operation, which helps to improve the accuracy of the party determined to be responsible and improve the user experience.

[0058] In the above implementation, by receiving authorization indication information and obtaining voice call data in response to the authorization indication information, it is possible to avoid obtaining the voice call data of the user talking to others without the user's permission, thereby protecting the user's privacy and improving the user experience.

[0059] In the above-mentioned embodiment, by obtaining the responsibility determination result corresponding to the order operation, in response to the fact that the responsible party indicated by the responsibility determination result is the service party, and the service party has not sent dispute arbitration request information within the responsibility determination time length after the generation time of the responsibility determination result, a responsibility determination result indicating that the responsible party for the order operation is the service party is generated. In the case that the service party does not express objection to the responsibility determination result, the responsibility determination result indicating that the responsible party for the order operation is the service party is automatically generated, thereby improving the accuracy of the responsibility determination result.

[0060] In the above technical solution, taking into account the differences in the usage habits of people in different regions, even if the meaning expressed in a sentence is the same, there may be certain differences in the words used in the sentence. For example, the words that express negative meanings are more frequently used in the southwest of my country, such as "can't win" and "don't want to", while in the northeast of my country, "no" and "can't" are more frequently used. Therefore, by receiving the updated weight parameters sent by the edge server, and updating the responsibility determination model according to the updated weight parameters, and taking at least one target statement as input and the responsibility determination result as output, the updated responsibility determination model is trained. When the trained responsibility determination model has not converged, the gradient update vector is obtained according to the trained responsibility determination model, and the gradient update vector is sent to the edge server; when the trained responsibility determination model converges, the trained responsibility determination model is determined as the target responsibility determination model. On the one hand, it can ensure that the target responsibility determination model finally obtained can be a model that learns both common rules and private rules, and the accuracy of the output responsibility determination result is higher; on the other hand, since the process of continuing to train the responsibility determination models on each responsibility determination model training system is jointly executed by the responsibility determination model training system and the edge server, compared with the responsibility determination model training system alone training the responsibility determination model, the required processing resources are less and the training speed is faster.

[0061] In the above-mentioned implementation mode, by acquiring voice call data of a call between the user and the service party before an order operation for an order corresponding to the order between the user and the service party occurs, voice recognition is performed on the voice call data to acquire a voice recognition text, and at least one target sentence in the voice recognition text is determined, wherein the distance between the target sentence and the target scene word is less than or equal to the scene word distance, and the word frequency of the target operation inducing word in the target sentence is greater than or equal to the inducing word frequency threshold, that is, a shorter target sentence may include content of communication between the user and the service party regarding the target scene corresponding to the order operation, and the target sentence may include content in which one of the user and the service party induces the other party to actively trigger the order operation through voice communication; a target training responsibility determination model is acquired, and at least one target sentence is used as input to acquire a responsibility determination result corresponding to the order operation output by the target training responsibility determination model, wherein according to the responsibility determination result, it can be determined whether there is content in which one party induces the other party to actively trigger the order operation in the call between the user and the service party, which helps to improve the accuracy of the determined party that should bear the responsibility and improve the user experience.

[0062] In the above implementation, considering that the online responsibility determination system needs to have a short reaction time to output the responsibility determination result, the processing speed of the target training responsibility determination model should be faster, and the textcnn model has the characteristics of fast calculation speed and low consumption of computing resources. When the responsibility determination method is applied to the online responsibility determination system, the target training responsibility determination model is limited to that obtained based on textcnn training, which ensures that the online responsibility determination system has a short reaction time to output the responsibility determination result, thereby improving the user experience. In addition, considering that the offline responsibility determination system has a relatively loose requirement on the response time of outputting the responsibility determination result, but has a high requirement on the accuracy of the responsibility determination result, the accuracy of the responsibility determination result output by the target training responsibility determination model should be relatively high. Although the BERT model consumes more computing resources and has a slower computing speed, it performs relatively well in recognition accuracy. When the responsibility determination method is applied to the offline responsibility determination system, the target training responsibility determination model is limited to include a first target training responsibility determination model obtained by TextCNN training and a second target training responsibility determination model obtained by BERT training, and a combined responsibility determination result is obtained according to the first responsibility determination result output by the first target training responsibility determination model and the second responsibility determination result output by the second target training responsibility determination model. This can improve the accuracy of the obtained combined responsibility determination result and improve user experience.

[0063] In the above-mentioned embodiment, by receiving responsibility determination result indication information and displaying the responsibility determination result indication information, wherein the responsibility determination result indication information is used to indicate the responsibility determination result corresponding to the order operation, the responsibility determination result is obtained through the following steps: by performing voice recognition on the voice call data of the call between the user party and the service party before the order operation corresponding to the order between the user party and the service party occurs, so as to obtain the voice recognition text, and determine at least one target sentence in the voice recognition text, the distance between the target sentence and the target scene word is less than or equal to the scene word distance, and the word frequency of the target operation inducing word in the target sentence is greater than or equal to the inducing word frequency threshold, that is, the target sentence with a shorter length may include the content of communication between the user party and the service party regarding the target scene corresponding to the order operation, and the target sentence may include the content of one party of the user party and the service party inducing the other party to actively trigger the order operation through voice communication; obtain the target training responsibility determination model, and take at least one target sentence as input to obtain the responsibility determination result corresponding to the order operation output by the target training responsibility determination model. Among them, since the responsibility determination result can determine whether there is content in the call between the user and the service provider that one party induces the other party to actively trigger the order operation, it can ensure that the user can promptly know the party that should bear the responsibility with a higher accuracy rate, thereby improving the user experience.

[0064] In the above implementation, by sending order operation indication information in response to the input order operation action, the responsibility determination server can be promptly informed of the order operation performed by the responsibility determination client, thereby improving the processing efficiency of the responsibility determination server when executing the corresponding responsibility determination method.

[0065] In the above implementation, by sending the voice call data of the conversation between the user party and the service party before the order operation for the order between the user party and the service party occurs, the responsibility determination server can obtain the corresponding voice call data in a timely manner, thereby improving the processing efficiency of the responsibility determination server when executing the corresponding responsibility determination method.

[0066] In the above embodiment, in response to the input authorization indication action, authorization indication information is sent, and the authorization indication information is used to indicate the acquisition of voice call data, which can avoid obtaining the voice call data of the user's conversation with others without the user's permission, thereby protecting the user's privacy and improving the user experience.

[0067] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Other features, objectives and advantages of the present disclosure will become more apparent through the following detailed description of non-limiting embodiments in conjunction with the accompanying drawings. In the accompanying drawings:

[0069] Figure 1 A flowchart of a model training method according to an embodiment of the present disclosure is shown.

[0070] Figure 2 A schematic structural diagram of a responsibility determination system according to an embodiment of the present disclosure is shown.

[0071] Figure 3 A flowchart of a responsibility determination method according to an embodiment of the present disclosure is shown.

[0072] Figure 4 A flowchart of a responsibility determination method according to an embodiment of the present disclosure is shown.

[0073] Figure 5 A flowchart of a responsibility determination method according to an embodiment of the present disclosure is shown.

[0074] Figure 6 A structural block diagram of a model training device according to an embodiment of the present disclosure is shown.

[0075] Figure 7 A structural block diagram of a responsibility determination device according to an embodiment of the present disclosure is shown.

[0076] Figure 8A structural block diagram of a responsibility determination device according to an embodiment of the present disclosure is shown.

[0077] Fig. 9 A structural block diagram of an electronic device according to an embodiment of the present disclosure is shown.

[0078] Fig.10 It is a schematic diagram of the structure of a computer system suitable for implementing the method according to the embodiment of the present disclosure. DETAILED DESCRIPTION

[0079] Hereinafter, exemplary embodiments of the present disclosure will be described in detail with reference to the accompanying drawings so that those skilled in the art can easily implement them. In addition, for the sake of clarity, parts not related to the description of the exemplary embodiments are omitted in the accompanying drawings.

[0080] In the present disclosure, it should be understood that terms such as "include" or "have" are intended to indicate the presence of labels, numbers, steps, behaviors, components, parts, or a combination thereof disclosed in the present specification, and are not intended to exclude the possibility that one or more other labels, numbers, steps, behaviors, components, parts, or a combination thereof exist or are added.

[0081] It should also be noted that, in the absence of conflict, the embodiments and labels in the embodiments of the present disclosure can be combined with each other. The present disclosure will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0082] In order to determine the party between the user and the service provider that should bear the corresponding responsibility for the order operation, the inventor of the present disclosure has considered the following solution.

[0083] Taking the example of users using food delivery services, when the user orders corresponding meals through the user terminal and forms an order corresponding to the meals, the food delivery person, that is, the service provider of the food delivery service, can apply through the delivery person terminal to execute the delivery of the meals in the order. After the food delivery service operator approves the application, it can send delivery instruction information to the delivery person terminal to instruct the delivery person to start delivering the meals in the order. At the same time, the food delivery service operator can also send service status update information to the user terminal to remind the user that the meals in the order have been started to be delivered by the corresponding delivery person. After receiving the delivery instruction information, the delivery person can respond to the delivery instruction information and stop displaying the corresponding order operation (such as canceling delivery, etc.) trigger button, so that the delivery person cannot trigger the corresponding order operation; and after receiving the service status update information, the user can respond to the service status update information and display the corresponding order operation (such as canceling delivery, etc.) trigger button. After the user triggers the cancel delivery button, the user can send a cancel delivery request to the takeaway delivery service operator. The takeaway delivery service operator can respond to the cancel delivery request and send a cancel delivery instruction information to the delivery person to instruct the delivery person to stop executing the delivery service for the order. Since the cancel delivery order operation is triggered by the user, it can be determined that the party that should bear the corresponding responsibility is the user.

[0084] Disadvantages of this solution: Since a call may occur between the user and the service provider, one of the two parties may induce the other to actively trigger the corresponding order operation during the call to avoid assuming the corresponding responsibility, resulting in the inability to accurately determine the party that should bear the corresponding responsibility, thereby damaging the user experience.

[0085] Taking into account the shortcomings of the above schemes, the inventors of the present disclosure have proposed a new scheme: by obtaining voice call data of a call between the user and the service party before the order operation for the order between the user and the service party occurs and the responsibility determination result corresponding to the order operation; performing speech recognition on the voice call data to obtain speech recognition text, determining at least one target sentence in the speech recognition text, the distance between the target sentence and the target scene word is less than or equal to the scene word distance, and the word frequency of the target operation inducing word in the target sentence is greater than or equal to the inducing word frequency threshold, that is, the target sentence of shorter length may include the content of communication between the user and the service party regarding the target scene corresponding to the order operation, and the target sentence may include the content of one of the user and the service party inducing the other party to actively trigger the order operation through voice communication; obtaining a responsibility determination model, and taking at least one target sentence as input and the responsibility determination result as output, training the responsibility determination model to obtain a target responsibility determination model. The target responsibility determination model obtained by the above scheme can obtain the corresponding responsibility determination results based on the sentences extracted from the voice call data, so as to determine whether there is any content in the call between the user and the service party that induces the other party to actively trigger an order operation, which helps to improve the accuracy of the party determined to be responsible and improve the user experience.

[0086] In order to solve the above problems, the present disclosure proposes a model training method, a responsibility determination method, an apparatus, a device, a medium and a product.

[0087] Figure 1 FIG. 1 is a flow chart of a model training method according to an embodiment of the present disclosure. Figure 1 As shown, the model training method includes steps S101-S104.

[0088] In step S101, voice call data of a call between the user and the service provider before an order operation for an order between the user and the service provider occurs and a responsibility determination result corresponding to the order operation are obtained.

[0089] In step S102, voice recognition is performed on the voice call data to obtain voice recognition text.

[0090] In step S103, at least one target sentence in the speech recognition text is determined.

[0091] Among them, the distance between the target sentence and the target scene word is less than or equal to the scene word distance, and the word frequency of the target operation inducing word in the target sentence is greater than or equal to the inducing word frequency threshold.

[0092] In step S104, a responsibility determination model is obtained, and at least one target sentence is used as input, and the responsibility determination result is used as output, and the responsibility determination model is trained to obtain a target responsibility determination model.

[0093] In one embodiment of the present disclosure, the user party can be understood as a user who requests a corresponding service; the service party can be understood as the specific executor of the corresponding service. Exemplarily, taking the corresponding service as an online service as an example, the user party can send a service request indication message to the online service operator through the corresponding application running on the user end of the access network, and the online service operator can send a service execution indication message to the service end of the access network used by the corresponding service party in response to the service request indication message, so as to instruct the service party to provide the corresponding online service to the user. Among them, online services may include meal delivery services, medicine delivery services, errand services, travel services, etc.

[0094] In one embodiment of the present disclosure, an order operation can be understood as modifying at least one item of order information of an order, wherein the order information may include at least one item of the order address, order time, order executor, order status, order quantity, and order type. For example, taking a meal delivery order for a meal delivery service as an example, the order operation can be modifying at least one item of the meal delivery address, meal arrival time, meal delivery person, meal quantity, meal type, and whether the meal delivery order is cancelled.

[0095] In one embodiment of the present disclosure, the call between the user and the service provider can be initiated by either the user or the service provider. It should be noted that in order to protect the privacy of the user, the call between the user and the service provider can be connected by the service operator, so as to ensure that neither the user nor the service provider can know the real communication address of the other party, which can be a telephone number, a social software number, etc.

[0096] In one embodiment of the present disclosure, the responsibility determination result can be used to indicate the probability that one of the user party and the service party induces the other party to actively trigger the corresponding order operation by voice communication in the corresponding voice call data. Exemplarily, the responsibility determination result can be a score ranging from 1 to 0, and the higher the score, the higher the probability that the service party induces the user party to trigger the corresponding order operation by voice communication.

[0097] In one embodiment of the present disclosure, voice recognition is performed on the voice call data, which may be based on a pre-acquired voice recognition algorithm or voice recognition model, or may be sending the voice call data to other devices or systems and receiving the voice recognition results returned by the other devices or systems. Exemplarily, the voice recognition of the voice call data may be achieved by calling a corresponding automatic speech recognition (ASR) service.

[0098] In one embodiment of the present disclosure, scene words can be understood as words that are highly related to the scene of the order between the user and the service provider. For example, when the order between the user and the service provider is a food delivery service, the scene words may include: order, list, task, takeout, meal, dish, food, delivery, take, put, etc.; when the order between the user and the service provider is a travel service, the scene words may include: car, driving, drive to, arrive, walk to, go to, go, position, direction, stop, jam, etc. The target scene words can be understood as the scene words in the speech recognition text.

[0099] Among them, scene words can be pre-set, or the speech recognition text corresponding to at least one segment of voice call data can be segmented, and the modal particles, verbs and other words irrelevant to the scene in the segmentation results can be removed, and the segmentation results after removal can be clustered, and word frequency statistics can be performed based on the clustering results. According to the word frequency statistics results, words with a word frequency greater than or equal to a word frequency threshold are determined as scene words.

[0100] In one embodiment of the present disclosure, the distance between the target sentence and the target scene word can be understood as the number of words or characters between the word closest to the target scene word in the target sentence and the target scene word. The scene word distance can be pre-set or obtained from other devices or systems.

[0101] In one embodiment of the present disclosure, the operation induction words can be understood as words with a high degree of relevance to the content that induces the execution of the corresponding operation. Exemplarily, when the order between the user and the service provider is for a meal delivery service, the operation induction words may include: cancel, late, late, no way, can't deliver, can't arrive, the car is broken, unfamiliar, can't find, random order, delay, delay, etc.; when the order between the user and the service provider is for a travel service, the operation induction words may include: cancel, late, late, no way, can't arrive, the car is broken, unfamiliar, can't find, random order, jam, collision, rub, accident, etc. The target operation induction words can be understood as the operation induction words in the speech recognition text.

[0102] Among them, the operation inducement words can be pre-set, or can be at least one voice call data determined to contain inducement content based on the responsibility determination result, and the voice recognition text corresponding to the at least one voice call data is segmented, and the modal particles, nouns and other words irrelevant to the inducement content in the segmentation results are removed, and the segmentation results after removal are clustered, and word frequency statistics are performed based on the clustering results. According to the word frequency statistics results, the words with a word frequency greater than or equal to the word frequency threshold are determined as operation inducement words.

[0103] In one embodiment of the present disclosure, the frequency of the target operation inducing words can be understood as the number of operation inducing words in the target sentence, or can also be understood as the ratio of the number of operation inducing words in the target sentence to the number of all words in the target sentence. The inducing word frequency threshold can be pre-set or obtained from other devices or systems.

[0104] In one embodiment of the present disclosure, the responsibility determination model can be understood as being pre-acquired or acquired from other devices or systems. The responsibility determination model can be a neural network (NN) model, a convolutional neural network (CNN) model, or a long short-term memory network (LSTM) model, etc. Exemplarily, the responsibility determination model can include a text convolutional neural network model (textConvolutional Neural Networks, textcnn) and a bidirectional encoder representation model based on a transformer (Bidirectional Encoder Representations from Transformers, bert).

[0105] In the above-mentioned embodiment, by obtaining voice call data of a call between the user party and the service party before the order operation for the order between the user party and the service party occurs and the responsibility determination result corresponding to the order operation; performing speech recognition on the voice call data to obtain speech recognition text, determining at least one target sentence in the speech recognition text, the distance between the target sentence and the target scene word is less than or equal to the scene word distance, and the word frequency of the target operation inducing word in the target sentence is greater than or equal to the inducing word frequency threshold, that is, the shorter target sentence may include the content of communication between the user party and the service party regarding the target scene corresponding to the order operation, and the target sentence may include the content of one party of the user party and the service party inducing the other party to actively trigger the order operation through voice communication; obtaining a responsibility determination model, and taking at least one target sentence as input and the responsibility determination result as output, the responsibility determination model is trained to obtain a target responsibility determination model. The target responsibility determination model obtained by the above scheme can obtain the corresponding responsibility determination results based on the sentences extracted from the voice call data, so as to determine whether there is any content in the call between the user and the service party that induces the other party to actively trigger an order operation, which helps to improve the accuracy of the party determined to be responsible and improve the user experience.

[0106] In one embodiment of the present disclosure, in step S101, obtaining voice call data can be achieved by the following steps:

[0107] Receive authorization indication information, and acquire voice call data in response to the authorization indication information.

[0108] In one embodiment of the present disclosure, the authorization indication information can be understood as the call data used to indicate that the corresponding user party allows the call between itself and at least one service party can be used to train the corresponding target responsibility determination model, and the user party can send the authorization indication information through the corresponding user terminal. It can also be understood that the call data used to indicate that the corresponding service party allows the call between itself and at least one user party can be used to train the corresponding target responsibility determination model, and the service party can send the authorization indication information through the corresponding service terminal.

[0109] Exemplarily, before the call between the user and the service provider begins, or after the call between the user and the service provider ends, the authorization request information can be displayed through the user terminal used by the user to prompt the user to input the authorization instruction information through the user terminal.

[0110] In the above implementation, by receiving authorization indication information and obtaining voice call data in response to the authorization indication information, it is possible to avoid obtaining the voice call data of the user talking to others without the user's permission, thereby protecting the user's privacy and improving the user experience.

[0111] In one embodiment of the present disclosure, in step S101, before obtaining the responsibility determination result corresponding to the order operation, the method further includes the following steps:

[0112] Obtain the accountability results corresponding to the order operation;

[0113] In step S101, obtaining the responsibility determination result corresponding to the order operation can be achieved through the following steps:

[0114] In response to the fact that the responsible party indicated by the responsibility determination result is the service party, and the service party has not sent any dispute arbitration request information within the responsibility determination time length after the time when the responsibility determination result is generated, a responsibility determination result is generated to indicate that the responsible party for the order operation is the service party.

[0115] In one embodiment of the present disclosure, the liability determination result can be understood as the result of preliminarily determining the party responsible for the order operation, and the liability determination result can be overturned after corresponding dispute arbitration.

[0116] In one embodiment of the present disclosure, the dispute arbitration request information can be understood as an instruction to request arbitration of the corresponding liability determination results to obtain the liability determination results after arbitration, wherein the arbitration can be performed by the corresponding arbitration personnel, and is also understood to be performed based on the corresponding arbitration algorithm or arbitration model.

[0117] In one embodiment of the present disclosure, when it is necessary to determine the responsibility for the corresponding order operation, the responsible party can be automatically determined as the service party, that is, the determination result indicating that the responsible party is the service party is automatically generated.

[0118] In one embodiment of the present disclosure, the moment when the accountability result is generated can be understood as the moment when the corresponding accountability result is generated, the moment when the accountability result is sent to the server corresponding to the service provider, or the moment when the server corresponding to the service provider receives the accountability result.

[0119] In the above-mentioned embodiment, by obtaining the responsibility determination result corresponding to the order operation, in response to the fact that the responsible party indicated by the responsibility determination result is the service party, and the service party has not sent dispute arbitration request information within the responsibility determination time length after the generation time of the responsibility determination result, a responsibility determination result indicating that the responsible party for the order operation is the service party is generated. In the case that the service party does not express objection to the responsibility determination result, the responsibility determination result indicating that the responsible party for the order operation is the service party is automatically generated, thereby improving the accuracy of the responsibility determination result.

[0120] In one embodiment of the present disclosure, in step S104, before taking at least one target sentence as input and taking the responsibility determination result as output to train the responsibility determination model to obtain the target responsibility determination model, the method further includes the following steps:

[0121] Receiving the updated weight parameters sent by the edge server, and updating the responsibility determination model according to the updated weight parameters;

[0122] In step S104, at least one target sentence is used as input, and the responsibility determination result is used as output to train the responsibility determination model to obtain a target responsibility determination model, which can be achieved by the following steps:

[0123] Taking at least one target sentence as input and taking the responsibility determination result as output, training the updated responsibility determination model;

[0124] When the trained responsibility determination model has not converged, a gradient update vector is obtained according to the trained responsibility determination model, and the gradient update vector is sent to the edge server;

[0125] When the trained responsibility determination model converges, the trained responsibility determination model is determined as the target responsibility determination model.

[0126] In one implementation of the present disclosure, the edge server is used to aggregate the gradient update vectors, and update the weight parameters of the responsibility determination model on the edge server according to the aggregated gradient update vectors to obtain updated weight parameters. The edge server can be a cloud server or a server provided by a responsibility determination service provider. It should be noted that an edge server can correspond to one or more regions. For example, the responsibility determination service provider can divide the area under its jurisdiction into multiple blocks, and the responsibility determination model training system for implementing the model training method in the corresponding area can interact with the edge server corresponding to the area. The responsibility determination model on the edge server can be a neural network model, a convolutional neural network model, a long short-term memory network model, etc. Exemplary responsibility determination models on edge servers may include TextCNN and BERT.

[0127] In one implementation of the present disclosure, the updated weight parameters sent by the edge server and received by the responsibility determination model training system are obtained by aggregating the gradient update vectors sent by multiple different responsibility determination model training systems on the edge server, and updating the weight parameters of the responsibility determination model on the edge server according to the aggregated gradient update vectors. Therefore, the updated responsibility determination model can reflect the common rules between the target statements and the responsibility determination results obtained by multiple different responsibility determination model training systems, which have been learned by the responsibility determination model on the edge server in the previous round of training. Then, at least one target sentence is taken as input, and the responsibility determination result is taken as output to train the updated responsibility determination model. This allows the updated responsibility determination model on the responsibility determination model training system to learn the common rules and also to learn the target sentence and responsibility determination result obtained by the responsibility determination model training system itself in a personalized manner, so that the trained responsibility determination model can learn the private rules between the target sentence and the responsibility determination result obtained by the responsibility determination model training system itself. When the trained responsibility determination model does not converge, it means that the trained responsibility determination model still needs to be trained. By obtaining the gradient update vector according to the trained responsibility determination model and sending the gradient update vector, the edge server can continue to obtain the corresponding update weight parameters based on the gradient update vectors uploaded by multiple responsibility determination model training systems, thereby continuing to train the responsibility determination models on each responsibility determination model training system. When the trained responsibility determination model converges, it can be considered that the converged responsibility determination model can output a relatively accurate responsibility determination result based on the target sentence obtained by the responsibility determination model training system. Therefore, the converged responsibility determination model can be stored as the target responsibility determination model, that is, a model with a relatively accurate output responsibility determination result.

[0128] In the above technical solution, taking into account the differences in the usage habits of people in different regions, even if the meaning expressed in a sentence is the same, there may be certain differences in the words used in the sentence. For example, the words that express negative meanings are more frequently used in the southwest of my country, such as "can't win" and "don't want to", while in the northeast of my country, "no" and "can't" are more frequently used. Therefore, by receiving the updated weight parameters sent by the edge server, and updating the responsibility determination model according to the updated weight parameters, and taking at least one target statement as input and the responsibility determination result as output, the updated responsibility determination model is trained. When the trained responsibility determination model has not converged, the gradient update vector is obtained according to the trained responsibility determination model, and the gradient update vector is sent to the edge server; when the trained responsibility determination model converges, the trained responsibility determination model is determined as the target responsibility determination model. On the one hand, it can ensure that the target responsibility determination model finally obtained can be a model that learns both common rules and private rules, and the accuracy of the output responsibility determination result is higher; on the other hand, since the process of continuing to train the responsibility determination models on each responsibility determination model training system is jointly executed by the responsibility determination model training system and the edge server, compared with the responsibility determination model training system alone training the responsibility determination model, the required processing resources are less and the training speed is faster.

[0129] Figure 2 A schematic structural diagram of a responsibility determination system according to an embodiment of the present disclosure is shown. The responsibility determination system can be applied to other Internet service platforms with online order scenarios such as express delivery services and food delivery services, and can also be applied to transportation services such as taxis, chauffeur services, driver rentals, or any combination thereof. The embodiments of the present disclosure do not specifically limit the application scenarios of the responsibility determination system.

[0130] like Figure 2 As shown, the responsibility determination system includes a responsibility determination server 10 , a responsibility determination client 20 and a network 30 .

[0131] In some embodiments, the responsibility determination server 10 can be a single server or a server group. The server group can be centralized or distributed (for example, the responsibility determination server 10 can be a distributed system). In some embodiments, the responsibility determination server 10 can be local or remote relative to the terminal. For example, the responsibility determination server 10 can access at least one information and / or data stored in the responsibility determination client 20 via a network. As another example, the responsibility determination server 10 can be directly connected to the responsibility determination client 20 to access the stored information and / or data. In some embodiments, the responsibility determination server 10 can be implemented on a cloud platform; as an example only, the cloud platform can include a private cloud, a public cloud, a hybrid cloud, a community cloud, a distributed cloud, an inter-cloud, a multi-cloud, etc., or any combination thereof.

[0132] The network 30 can be used for the exchange of information and / or data. In some embodiments, one or more components in the responsibility determination system (e.g., the responsibility determination server 10, the responsibility determination client 20) can send information and / or data to other components. In some embodiments, the network 30 can be any type of wired or wireless network, or a combination thereof. As an example only, the network 30 can include a wired network, a wireless network, a fiber optic network, a telecommunication network, an intranet, the Internet, a local area network (LAN), a wide area network (WAN), a wireless local area network (WLAN), a metropolitan area network (MAN), a wide area network (WAN), a public switched telephone network (PSTN), a Bluetooth network, a ZigBee network, or a near field communication (NFC) network, etc., or any combination thereof. In some embodiments, the network 30 can include one or more network access points. For example, network 30 may include wired or wireless network access points, such as base stations and / or network switching nodes, through which one or more components of the responsibility determination system may connect to network 30 to exchange data and / or information.

[0133] In some embodiments, the responsibility determination client 20 may include a mobile device, a tablet computer, a laptop computer, or a built-in device in a motor vehicle, or any combination thereof. In some embodiments, the mobile device may include a smart home device, a wearable device, a smart mobile device, a virtual reality device, or an augmented reality device, or any combination thereof. In some embodiments, the smart home device may include a smart lighting device, a control device for a smart electrical device, a smart TV, a smart camera, or an intercom, or any combination thereof. In some embodiments, the wearable device may include a smart bracelet, a smart shoelace, a smart glass, a smart helmet, a smart watch, smart clothing, a smart backpack, a smart accessory, or any combination thereof. In some embodiments, the smart mobile device may include a smart phone, a personal digital assistant (PDA), a gaming device, a navigation device, or a point of sale (POS) device, or any combination thereof. In some embodiments, the virtual reality device and / or the augmented reality device may include a virtual reality helmet, a virtual reality glass, a virtual reality patch, an augmented reality helmet, an augmented reality glass, or an augmented reality patch, or any combination thereof. For example, the virtual reality device and / or the augmented reality device may include various virtual reality products, etc. In some embodiments, the built-in device in the motor vehicle may include an onboard computer, an onboard TV, etc. In some embodiments, the responsibility determination client 20 may be a device with positioning technology.

[0134] Figure 3 A flowchart showing a method for determining responsibility according to an embodiment of the present disclosure is shown. The method for determining responsibility can be applied to Figure 2 As shown in the responsibility determination server. Figure 3 As shown, the responsibility determination method includes steps S201-S204.

[0135] In step S201, voice call data of a call between the user party and the service party before an order operation for a corresponding order between the user party and the service party occurs is obtained.

[0136] In step S202, voice recognition is performed on the voice call data to obtain voice recognition text.

[0137] In step S203, at least one target sentence in the speech recognition text is determined, the distance between the target sentence and the target scene word is less than or equal to the scene word distance, and the word frequency of the target operation inducing word in the target sentence is greater than or equal to the inducing word frequency threshold.

[0138] In step S204, a target training responsibility determination model is obtained, and at least one target statement is used as input to obtain a responsibility determination result corresponding to the order operation output by the target training responsibility determination model.

[0139] In one embodiment of the present disclosure, the user party can be understood as a user who requests a corresponding service; the service party can be understood as the specific executor of the corresponding service. Exemplarily, taking the corresponding service as an online service as an example, the user party can send a service request indication message to the online service operator through the corresponding application running on the user end of the access network, and the online service operator can send a service execution indication message to the service end of the access network used by the corresponding service party in response to the service request indication message, so as to instruct the service party to provide the corresponding online service to the user. Among them, online services may include meal delivery services, medicine delivery services, errand services, travel services, etc.

[0140] In one embodiment of the present disclosure, an order operation can be understood as modifying at least one item of order information of an order, wherein the order information may include at least one item of the order address, order time, order executor, order status, order quantity, and order type. For example, taking a meal delivery order for a meal delivery service as an example, the order operation can be modifying at least one item of the meal delivery address, meal arrival time, meal delivery person, meal quantity, meal type, and whether the meal delivery order is cancelled.

[0141] In one embodiment of the present disclosure, the call between the user and the service provider can be initiated by either the user or the service provider. It should be noted that in order to protect the privacy of the user, the call between the user and the service provider can be connected by the service operator, so as to ensure that neither the user nor the service provider can know the real communication address of the other party, which can be a telephone number, a social software number, etc.

[0142] In one embodiment of the present disclosure, the responsibility determination result can be used to indicate the probability that one of the user party and the service party induces the other party to actively trigger the corresponding order operation by voice communication in the corresponding voice call data. Exemplarily, the responsibility determination result can be a score ranging from 1 to 0, and the higher the score, the higher the probability that the service party induces the user party to trigger the corresponding order operation by voice communication.

[0143] In one embodiment of the present disclosure, voice recognition is performed on the voice call data, which may be based on a pre-acquired voice recognition algorithm or voice recognition model, or may be sending the voice call data to other devices or systems and receiving the voice recognition results returned by the other devices or systems. Exemplarily, the voice recognition of the voice call data may be achieved by calling a corresponding automatic speech recognition (ASR) service.

[0144] In one embodiment of the present disclosure, scene words can be understood as words that are highly related to the scene of the order between the user and the service provider. For example, when the order between the user and the service provider is a food delivery service, the scene words may include: order, list, task, takeout, meal, dish, food, delivery, take, put, etc.; when the order between the user and the service provider is a travel service, the scene words may include: car, driving, drive to, arrive, walk to, go to, go, position, direction, stop, jam, etc. The target scene words can be understood as the scene words in the speech recognition text.

[0145] Among them, scene words can be pre-set, or the speech recognition text corresponding to at least one segment of voice call data can be segmented, and the modal particles, verbs and other words irrelevant to the scene in the segmentation results can be removed, and the segmentation results after removal can be clustered, and word frequency statistics can be performed based on the clustering results. According to the word frequency statistics results, words with a word frequency greater than or equal to a word frequency threshold are determined as scene words.

[0146] In one embodiment of the present disclosure, the distance between the target sentence and the target scene word can be understood as the number of words or characters between the word closest to the target scene word in the target sentence and the target scene word. The scene word distance can be pre-set or obtained from other devices or systems.

[0147] In one embodiment of the present disclosure, the operation induction words can be understood as words with a high degree of relevance to the content that induces the execution of the corresponding operation. Exemplarily, when the order between the user and the service provider is for a meal delivery service, the operation induction words may include: cancel, late, late, no way, can't deliver, can't arrive, the car is broken, unfamiliar, can't find, random order, delay, delay, etc.; when the order between the user and the service provider is for a travel service, the operation induction words may include: cancel, late, late, no way, can't arrive, the car is broken, unfamiliar, can't find, random order, jam, collision, rub, accident, etc. The target operation induction words can be understood as the operation induction words in the speech recognition text.

[0148] Among them, the operation inducement words can be pre-set, or can be at least one voice call data determined to contain inducement content based on the responsibility determination result, and the voice recognition text corresponding to the at least one voice call data is segmented, and the modal particles, nouns and other words irrelevant to the inducement content in the segmentation results are removed, and the segmentation results after removal are clustered, and word frequency statistics are performed based on the clustering results. According to the word frequency statistics results, the words with a word frequency greater than or equal to the word frequency threshold are determined as operation inducement words.

[0149] In one embodiment of the present disclosure, the frequency of the target operation inducing words can be understood as the number of operation inducing words in the target sentence, or can also be understood as the ratio of the number of operation inducing words in the target sentence to the number of all words in the target sentence. The inducing word frequency threshold can be pre-set or obtained from other devices or systems.

[0150] In one embodiment of the present disclosure, the target responsibility determination model can be understood as being pre-acquired or obtained from other devices or systems. The target responsibility determination model can be a neural network (NN) model, a convolutional neural network (CNN) model, or a long short-term memory network (LSTM) model. Exemplarily, the target responsibility determination model can include a text convolutional neural network model (textcnn) and a bidirectional encoder representation model based on a transformer (bert).

[0151] In the above-mentioned implementation mode, by acquiring voice call data of a call between the user and the service party before an order operation for an order corresponding to the order between the user and the service party occurs, voice recognition is performed on the voice call data to acquire a voice recognition text, and at least one target sentence in the voice recognition text is determined, wherein the distance between the target sentence and the target scene word is less than or equal to the scene word distance, and the word frequency of the target operation inducing word in the target sentence is greater than or equal to the inducing word frequency threshold, that is, a shorter target sentence may include content of communication between the user and the service party regarding the target scene corresponding to the order operation, and the target sentence may include content in which one of the user and the service party induces the other party to actively trigger the order operation through voice communication; a target training responsibility determination model is acquired, and at least one target sentence is used as input to acquire a responsibility determination result corresponding to the order operation output by the target training responsibility determination model, wherein according to the responsibility determination result, it can be determined whether there is content in which one party induces the other party to actively trigger the order operation in the call between the user and the service party, which helps to improve the accuracy of the determined party that should bear the responsibility and improve the user experience.

[0152] In one implementation of the present disclosure, the target training responsibility determination model is obtained by any of the above-mentioned model training methods.

[0153] In one implementation of the present disclosure, the responsibility determination method is applied to an online responsibility determination system, and the target training responsibility determination model is obtained by training based on a text convolutional neural network model (textcnn);

[0154] Or, the responsibility determination method is applied to an offline responsibility determination system, and the target training responsibility determination model includes a first target training responsibility determination model obtained by training based on TextCNN and a second target training responsibility determination model obtained by training based on a bidirectional encoding representation model (BERT) of a converter;

[0155] The method further comprises the steps of:

[0156] A combined responsibility determination result is obtained based on a first responsibility determination result output by the first target training responsibility determination model and a second responsibility determination result output by the second target training responsibility determination model.

[0157] In one implementation of the present disclosure, a combined responsibility determination result is obtained based on a first responsibility determination result and a second responsibility determination result. The combined responsibility determination result can be obtained by performing calculations based on the first responsibility determination result and the second responsibility determination result based on a pre-acquired combined responsibility determination algorithm; or the first responsibility determination result and the second responsibility determination result can be input into a pre-acquired combined responsibility determination model to obtain the combined responsibility determination result output by the combined responsibility determination model.

[0158] For example, Figure 4 A flow chart of a method for determining responsibility according to an embodiment of the present disclosure is shown. Figure 4 As shown, the order processing method includes the following steps:

[0159] Obtain anonymous small account call records by identifying the link.

[0160] The anonymous accounts include the anonymous accounts corresponding to the user and the anonymous accounts corresponding to the service provider. The anonymous account call records can be understood as including the voice call data of the call between the user and the service provider before the order operation for the order between the user and the service provider occurs.

[0161] Based on the speech recognition ASR service, speech recognition is performed on the voice call data in the anonymous small account call records to obtain speech recognition text.

[0162] Key sentences are extracted from the speech recognition text to determine at least one target sentence in the speech recognition text.

[0163] The target sentence is input into the model obtained by training based on textcnn, namely the first target training responsibility determination model, to obtain the first responsibility determination result output by the first target training responsibility determination model.

[0164] The target sentence is input into the model obtained based on BERT training, that is, the second target training responsibility determination model, to obtain the second responsibility determination result output by the second target training responsibility determination model.

[0165] The responsibility judgment system can obtain a combined responsibility judgment result based on the first responsibility judgment result output by the first target training responsibility judgment model and the second responsibility judgment result output by the second target training responsibility judgment model, and determine based on the combined responsibility judgment result whether there is any content in the call between the user and the service party in which one party induces the other party to actively trigger an order operation.

[0166] In the above implementation, considering that the online responsibility determination system needs to have a short reaction time to output the responsibility determination result, the processing speed of the target training responsibility determination model should be faster, and the textcnn model has the characteristics of fast calculation speed and low consumption of computing resources. When the responsibility determination method is applied to the online responsibility determination system, the target training responsibility determination model is limited to that obtained based on textcnn training, which ensures that the online responsibility determination system has a short reaction time to output the responsibility determination result, thereby improving the user experience. In addition, considering that the offline responsibility determination system has a relatively loose requirement on the response time of outputting the responsibility determination result, but has a high requirement on the accuracy of the responsibility determination result, the accuracy of the responsibility determination result output by the target training responsibility determination model should be relatively high. Although the BERT model consumes more computing resources and has a slower computing speed, it performs relatively well in recognition accuracy. When the responsibility determination method is applied to the offline responsibility determination system, the target training responsibility determination model is limited to include a first target training responsibility determination model obtained by TextCNN training and a second target training responsibility determination model obtained by BERT training, and a combined responsibility determination result is obtained according to the first responsibility determination result output by the first target training responsibility determination model and the second responsibility determination result output by the second target training responsibility determination model. This can improve the accuracy of the obtained combined responsibility determination result and improve user experience.

[0167] Figure 5 A flowchart showing a method for determining responsibility according to an embodiment of the present disclosure is shown. The method for determining responsibility can be applied to Figure 2 As shown in the responsibility determination client. Figure 5 As shown, the responsibility determination method includes step S301:

[0168] In step S301, responsibility determination result indication information is received and displayed.

[0169] The responsibility determination result indication information is used to indicate the responsibility determination result corresponding to the order operation. Figure 4 The corresponding implementation example is obtained.

[0170] In one embodiment of the present disclosure, the responsibility determination result can be used to indicate the probability that one of the user party and the service party induces the other party to actively trigger the corresponding order operation by voice communication in the corresponding voice call data. Exemplarily, the responsibility determination result can be a score ranging from 1 to 0, and the higher the score, the higher the probability that the service party induces the user party to trigger the corresponding order operation by voice communication.

[0171] In one embodiment of the present disclosure, the user party can be understood as a user who requests a corresponding service; the service party can be understood as the specific executor of the corresponding service. Exemplarily, taking the corresponding service as an online service as an example, the user party can send a service request indication message to the online service operator through the corresponding application running on the user end of the access network, and the online service operator can send a service execution indication message to the service end of the access network used by the corresponding service party in response to the service request indication message, so as to instruct the service party to provide the corresponding online service to the user. Among them, online services may include meal delivery services, medicine delivery services, errand services, travel services, etc.

[0172] In the above-mentioned embodiment, by receiving responsibility determination result indication information and displaying the responsibility determination result indication information, wherein the responsibility determination result indication information is used to indicate the responsibility determination result corresponding to the order operation, the responsibility determination result is obtained through the following steps: by performing voice recognition on the voice call data of the call between the user party and the service party before the order operation corresponding to the order between the user party and the service party occurs, so as to obtain the voice recognition text, and determine at least one target sentence in the voice recognition text, the distance between the target sentence and the target scene word is less than or equal to the scene word distance, and the word frequency of the target operation inducing word in the target sentence is greater than or equal to the inducing word frequency threshold, that is, the target sentence with a shorter length may include the content of communication between the user party and the service party regarding the target scene corresponding to the order operation, and the target sentence may include the content of one party of the user party and the service party inducing the other party to actively trigger the order operation through voice communication; obtain the target training responsibility determination model, and take at least one target sentence as input to obtain the responsibility determination result corresponding to the order operation output by the target training responsibility determination model. Among them, since the responsibility determination result can determine whether there is content in the call between the user and the service provider that one party induces the other party to actively trigger the order operation, it can ensure that the user can promptly know the party that should bear the responsibility with a higher accuracy rate, thereby improving the user experience.

[0173] In an implementation of the present disclosure, in step S301, before receiving the responsibility determination result indication information and displaying the responsibility determination result indication information, the method further includes the following steps:

[0174] In response to the input order operation action, order operation instruction information is sent, and the order operation instruction information is used to instruct the order operation to be performed on the order between the user party and the service party.

[0175] In one embodiment of the present disclosure, an order operation can be understood as modifying at least one item of order information of an order, wherein the order information may include at least one item of the order address, order time, order executor, order status, order quantity, and order type. For example, taking a meal delivery order for a meal delivery service as an example, the order operation can be modifying at least one item of the meal delivery address, meal arrival time, meal delivery person, meal quantity, meal type, and whether the meal delivery order is cancelled.

[0176] In the above implementation, by sending order operation indication information in response to the input order operation action, the responsibility determination server can be promptly informed of the order operation performed by the responsibility determination client, thereby improving the processing efficiency of the responsibility determination server when executing the corresponding responsibility determination method.

[0177] In an implementation of the present disclosure, in step S301, before receiving the responsibility determination result indication information and displaying the responsibility determination result indication information, the method further includes the following steps:

[0178] Sending voice call data of a call between the user party and the service party before an order operation for an order between the user party and the service party occurs.

[0179] In one embodiment of the present disclosure, the call between the user and the service provider can be initiated by either the user or the service provider. It should be noted that in order to protect the privacy of the user, the call between the user and the service provider can be connected by the service operator, so as to ensure that neither the user nor the service provider can know the real communication address of the other party, which can be a telephone number, a social software number, etc.

[0180] In the above implementation, by sending the voice call data of the conversation between the user party and the service party before the order operation for the order between the user party and the service party occurs, the responsibility determination server can obtain the corresponding voice call data in a timely manner, thereby improving the processing efficiency of the responsibility determination server when executing the corresponding responsibility determination method.

[0181] In an implementation of the present disclosure, in step S301, before receiving the responsibility determination result indication information and displaying the responsibility determination result indication information, the method further includes the following steps:

[0182] In response to the input authorization indication action, authorization indication information is sent, where the authorization indication information is used to instruct the acquisition of voice call data.

[0183] In one embodiment of the present disclosure, the authorization indication information can be understood as the call data used to indicate that the corresponding user party allows the call between itself and at least one service party can be used to train the corresponding target responsibility determination model, and the user party can send the authorization indication information through the corresponding user terminal. It can also be understood that the call data used to indicate that the corresponding service party allows the call between itself and at least one user party can be used to train the corresponding target responsibility determination model, and the service party can send the authorization indication information through the corresponding service terminal.

[0184] Exemplarily, before the call between the user and the service provider begins, or after the call between the user and the service provider ends, the authorization request information can be displayed through the user terminal used by the user to prompt the user to input the authorization instruction information through the user terminal.

[0185] In the above embodiment, in response to the input authorization indication action, authorization indication information is sent, and the authorization indication information is used to indicate the acquisition of voice call data, which can avoid obtaining the voice call data of the user's conversation with others without the user's permission, thereby protecting the user's privacy and improving the user experience.

[0186] The following reference Figure 6 A model training device according to an embodiment of the present disclosure is described. Figure 6 A structural block diagram of a model training device according to an embodiment of the present disclosure is shown.

[0187] like Figure 6 As shown, the model training device 100 includes:

[0188] The training data acquisition module 101 is configured to acquire voice call data of a call between the user party and the service party before an order operation for an order between the user party and the service party occurs and a responsibility determination result corresponding to the order operation;

[0189] The first speech recognition module 102 is configured to perform speech recognition on the voice call data to obtain speech recognition text;

[0190] The first sentence determination module 103 is configured to determine at least one target sentence in the speech recognition text, wherein the distance between the target sentence and the target scene word is less than or equal to the scene word distance, and the word frequency of the target operation inducing word in the target sentence is greater than or equal to the inducing word frequency threshold;

[0191] The model training module 104 is configured to obtain a responsibility determination model, and take at least one target statement as input, take a responsibility determination result as output, and train the responsibility determination model to obtain a target responsibility determination model.

[0192] According to the technical solution provided by the embodiment of the present disclosure, by obtaining voice call data of a call between the user party and the service party before an order operation for an order between the user party and the service party occurs and a responsibility determination result corresponding to the order operation; performing speech recognition on the voice call data to obtain a speech recognition text, determining at least one target sentence in the speech recognition text, wherein the distance between the target sentence and the target scene word is less than or equal to the scene word distance, and the word frequency of the target operation inducing word in the target sentence is greater than or equal to the inducing word frequency threshold, that is, the target sentence of shorter length may include the content of communication between the user party and the service party regarding the target scene corresponding to the order operation, and the target sentence may include the content of one party of the user party and the service party inducing the other party to actively trigger the order operation through voice communication; obtaining a responsibility determination model, and taking at least one target sentence as input and the responsibility determination result as output, training the responsibility determination model to obtain a target responsibility determination model. The target responsibility determination model obtained by the above scheme can obtain the corresponding responsibility determination results based on the sentences extracted from the voice call data, so as to determine whether there is any content in the call between the user and the service party that induces the other party to actively trigger an order operation, which helps to improve the accuracy of the party determined to be responsible and improve the user experience.

[0193] Those skilled in the art will understand that referring to Figure 6 The technical solution described can be compared with the above Figure 1 The corresponding embodiments are combined to have Figure 1 The technical effects achieved by the corresponding embodiments. For details, please refer to Figure 1 The specific contents of the description of the corresponding embodiments will not be repeated here.

[0194] The following reference Figure 7 A responsibility determination device according to an embodiment of the present disclosure is described. Figure 7 A structural block diagram of a responsibility determination device according to an embodiment of the present disclosure is shown.

[0195] like Figure 7 As shown, the responsibility determination device 200 includes:

[0196] The judgment data acquisition module 201 is configured to acquire voice call data of a call between the user party and the service party before an order operation for a corresponding order between the user party and the service party occurs;

[0197] The second speech recognition module 202 is configured to perform speech recognition on the voice call data to obtain speech recognition text;

[0198] The second sentence determination module 203 is configured to determine at least one target sentence in the speech recognition text, wherein the distance between the target sentence and the target scene word is less than or equal to the scene word distance, and the word frequency of the target operation inducing word in the target sentence is greater than or equal to the inducing word frequency threshold;

[0199] The determination result acquisition module 204 is configured to obtain a target training responsibility determination model and take at least one target statement as input to obtain a responsibility determination result corresponding to the order operation output by the target training responsibility determination model.

[0200] According to the technical solution provided by the embodiment of the present disclosure, by obtaining voice call data of a call between the user party and the service party before an order operation for an order corresponding to the order between the user party and the service party occurs, voice recognition is performed on the voice call data to obtain a voice recognition text, and at least one target sentence in the voice recognition text is determined, the distance between the target sentence and the target scene word is less than or equal to the scene word distance, and the word frequency of the target operation inducing word in the target sentence is greater than or equal to the inducing word frequency threshold, that is, the target sentence with a shorter length may include the content of communication between the user party and the service party for the target scene corresponding to the order operation, and the target sentence may include the content that one of the user party and the service party induces the other party to actively trigger the order operation through voice communication; obtain a target training responsibility determination model, and use at least one target sentence as input to obtain a responsibility determination result corresponding to the order operation output by the target training responsibility determination model, wherein according to the responsibility determination result, it can be determined whether there is content in which one party induces the other party to actively trigger the order operation in the call between the user party and the service party, which helps to improve the accuracy of the determined party that should bear the responsibility and improve the user experience.

[0201] Those skilled in the art will understand that referring to Figure 7 The technical solution described can be compared with the above Figure 3 The corresponding embodiments are combined to have Figure 3 The technical effects achieved by the corresponding embodiments. For details, please refer to Figure 3 The specific contents of the description of the corresponding embodiments will not be repeated here.

[0202] The following reference Figure 8 A responsibility determination device according to an embodiment of the present disclosure is described. Figure 8 A structural block diagram of a responsibility determination device according to an embodiment of the present disclosure is shown.

[0203] like Figure 8 As shown, the responsibility determination device 300 includes:

[0204] The determination result display module 301 is configured to receive responsibility determination result indication information and display the responsibility determination result indication information. The responsibility determination result indication information is used to indicate the responsibility determination result corresponding to the order operation. The responsibility determination result is obtained through any one of the methods from the second aspect, the first implementation method of the second aspect to the second implementation method.

[0205] In the above-mentioned embodiment, by receiving responsibility determination result indication information and displaying the responsibility determination result indication information, wherein the responsibility determination result indication information is used to indicate the responsibility determination result corresponding to the order operation, the responsibility determination result is obtained through the following steps: by performing voice recognition on the voice call data of the call between the user party and the service party before the order operation corresponding to the order between the user party and the service party occurs, so as to obtain the voice recognition text, and determine at least one target sentence in the voice recognition text, the distance between the target sentence and the target scene word is less than or equal to the scene word distance, and the word frequency of the target operation inducing word in the target sentence is greater than or equal to the inducing word frequency threshold, that is, the target sentence with a shorter length may include the content of communication between the user party and the service party regarding the target scene corresponding to the order operation, and the target sentence may include the content of one party of the user party and the service party inducing the other party to actively trigger the order operation through voice communication; obtain the target training responsibility determination model, and take at least one target sentence as input to obtain the responsibility determination result corresponding to the order operation output by the target training responsibility determination model. Among them, since the responsibility determination result can determine whether there is content in the call between the user and the service provider that one party induces the other party to actively trigger the order operation, it can ensure that the user can promptly know the party that should bear the responsibility with a higher accuracy rate, thereby improving the user experience.

[0206] Those skilled in the art will understand that referring to Figure 8 The technical solution described can be compared with the above Figure 5 The corresponding embodiments are combined to have Figure 5 The technical effects achieved by the corresponding embodiments. For details, please refer to Figure 5 The specific contents of the description of the corresponding embodiments will not be repeated here.

[0207] Fig. 9 A structural block diagram of an electronic device according to an embodiment of the present disclosure is shown.

[0208] The present disclosure also provides an electronic device, such as Fig. 9As shown, the electronic device 400 includes at least one processor 401 and a memory 402 that is communicatively connected to the at least one processor 401. The memory 402 stores instructions that can be executed by the at least one processor 401, and the instructions can be executed by the at least one processor 401 to implement the following steps:

[0209] In a first aspect, an embodiment of the present disclosure provides a model training method, the method comprising:

[0210] Acquire voice call data of a call between the user party and the service provider before an order operation on an order between the user party and the service provider occurs, and a responsibility determination result corresponding to the order operation;

[0211] Performing speech recognition on the voice call data to obtain speech recognition text;

[0212] Determine at least one target sentence in the speech recognition text, wherein the distance between the target sentence and the target scene word is less than or equal to the scene word distance, and the word frequency of the target operation inducing word in the target sentence is greater than or equal to the inducing word frequency threshold;

[0213] A responsibility determination model is obtained, and at least one target sentence is used as input, and a responsibility determination result is used as output, and the responsibility determination model is trained to obtain a target responsibility determination model.

[0214] In combination with the first aspect, in a first implementation of the first aspect of the present disclosure, obtaining voice call data includes:

[0215] Receive authorization indication information, and acquire voice call data in response to the authorization indication information.

[0216] In combination with the first aspect, in a second implementation of the first aspect of the present disclosure, before obtaining the responsibility determination result corresponding to the order operation, the method further includes:

[0217] Obtain the accountability results corresponding to the order operation;

[0218] Obtain the responsibility determination results corresponding to the order operation, including:

[0219] In response to the fact that the responsible party indicated by the responsibility determination result is the service party, and the service party has not sent any dispute arbitration request information within the responsibility determination time length after the time when the responsibility determination result is generated, a responsibility determination result is generated to indicate that the responsible party for the order operation is the service party.

[0220] In combination with the first aspect, in a third implementation of the first aspect of the present disclosure, before taking at least one target sentence as input and taking the responsibility determination result as output and training the responsibility determination model to obtain the target responsibility determination model, the method further includes:

[0221] Receiving the updated weight parameters sent by the edge server, and updating the responsibility determination model according to the updated weight parameters;

[0222] Taking at least one target sentence as input and taking the responsibility determination result as output, training the responsibility determination model to obtain the target responsibility determination model includes:

[0223] Taking at least one target sentence as input and taking the responsibility determination result as output, training the updated responsibility determination model;

[0224] When the trained responsibility determination model has not converged, a gradient update vector is obtained according to the trained responsibility determination model, and the gradient update vector is sent to the edge server;

[0225] When the trained responsibility determination model converges, the trained responsibility determination model is determined as the target responsibility determination model.

[0226] In a second aspect, an embodiment of the present disclosure provides a method for determining responsibility, the method comprising:

[0227] Acquire voice call data of a call between the user party and the service party before an order operation for a corresponding order between the user party and the service party occurs;

[0228] Performing speech recognition on the voice call data to obtain speech recognition text;

[0229] Determine at least one target sentence in the speech recognition text, wherein the distance between the target sentence and the target scene word is less than or equal to the scene word distance, and the word frequency of the target operation inducing word in the target sentence is greater than or equal to the inducing word frequency threshold;

[0230] Obtain a target training responsibility determination model, and take at least one target statement as input to obtain a responsibility determination result corresponding to the order operation output by the target training responsibility determination model.

[0231] In combination with the second aspect, in a first implementation of the second aspect of the present disclosure, the target training responsibility determination model is obtained through the model training method of any one of the first aspect, the first implementation of the first aspect to the third implementation.

[0232] In combination with the second aspect, in a second implementation of the second aspect of the present disclosure, the responsibility determination method is applied to an online responsibility determination system, and the target training responsibility determination model is obtained by training based on a text convolutional neural network model (textcnn);

[0233] Or, the responsibility determination method is applied to an offline responsibility determination system, and the target training responsibility determination model includes a first target training responsibility determination model obtained by training based on TextCNN and a second target training responsibility determination model obtained by training based on a bidirectional encoding representation model (BERT) of a converter;

[0234] The method also includes:

[0235] A combined responsibility determination result is obtained based on a first responsibility determination result output by the first target training responsibility determination model and a second responsibility determination result output by the second target training responsibility determination model.

[0236] In a third aspect, an embodiment of the present disclosure provides a method for determining responsibility, wherein the method includes:

[0237] Receive responsibility determination result indication information and display the responsibility determination result indication information, the responsibility determination result indication information is used to indicate the responsibility determination result corresponding to the order operation, and the responsibility determination result is obtained through any one of the methods from the second aspect, the first implementation method of the second aspect to the second implementation method.

[0238] In combination with the third aspect, in a first implementation manner of the third aspect of the present disclosure, before receiving the responsibility determination result indication information and displaying the responsibility determination result indication information, the method further includes:

[0239] In response to the input order operation action, order operation instruction information is sent, and the order operation instruction information is used to instruct the order operation to be performed on the order between the user party and the service party.

[0240] In combination with the third aspect, in a second implementation manner of the third aspect of the present disclosure, before receiving the responsibility determination result indication information and displaying the responsibility determination result indication information, the method further includes:

[0241] Sending voice call data of a call between the user party and the service party before an order operation for an order between the user party and the service party occurs.

[0242] In combination with the second implementation of the third aspect, in a third implementation of the third aspect of the present disclosure, before receiving the responsibility determination result indication information and displaying the responsibility determination result indication information, the method further includes:

[0243] In response to the input authorization indication action, authorization indication information is sent, where the authorization indication information is used to instruct the acquisition of voice call data.

[0244] like Fig.10As shown, the computer system 500 includes a processing unit 501, which can perform various processes in the embodiments shown in the above figures according to the program stored in the read-only memory (ROM) 502 or the program loaded from the storage part 508 to the random access memory (RAM) 503. In the RAM 503, various programs and data required for the operation of the system 500 are also stored. The CPU 501, the ROM 502, and the RAM 503 are connected to each other through a bus 504. An input / output (I / O) interface 505 is also connected to the bus 504.

[0245] The following components are connected to the I / O interface 505: an input section 506 including a keyboard, a mouse, etc. An output section 507 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc. A storage section 508 including a hard disk, etc. And a communication section 509 including a network interface card such as a LAN card, a modem, etc. The communication section 509 performs communication processing via a network such as the Internet. A drive 510 is also connected to the I / O interface 505 as needed. A removable medium 511, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 510 as needed, so that the computer program read therefrom is installed into the storage section 508 as needed. Among them, the processing unit 501 can be implemented as a processing unit such as a CPU, a GPU, a TPU, an FPGA, an NPU, etc.

[0246] In particular, according to an embodiment of the present disclosure, the method described above with reference to the accompanying drawings can be implemented as a computer software program. For example, an embodiment of the present disclosure includes a computer program product, which includes a computer program tangibly contained on a readable medium thereof, and the computer program includes a program code for executing the method in the accompanying drawings. In such an embodiment, the computer program can be downloaded and installed from a network through a communication portion 509, and / or installed from a removable medium 511. For example, an embodiment of the present disclosure includes a readable storage medium on which computer instructions are stored, and when the computer instructions are executed by a processor, the program code for executing the method in the accompanying drawings is implemented.

[0247] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, functions and operations of the system, method and computer program product according to various embodiments of the present disclosure. In this regard, each square box in the road map or block diagram can represent a part of a module, program segment or code, and the part of the module, program segment or code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the square box can also occur in a different order from the order marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs the specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0248] The units or modules involved in the embodiments described in the present disclosure may be implemented by software or hardware. The units or modules described may also be set in a processor, and the names of these units or modules do not constitute limitations on the units or modules themselves in some cases.

[0249] As another aspect, the present disclosure further provides a computer-readable storage medium, which may be a computer-readable storage medium included in the node in the above-mentioned embodiment. It may also be a computer-readable storage medium that exists independently and is not assembled into a device. The computer-readable storage medium stores one or more programs, and the programs are used by one or more processors to execute the method described in the present disclosure.

[0250] The above description is only a preferred embodiment of the present disclosure and an explanation of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the present disclosure is not limited to the technical solution formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the inventive concept. For example, the above features are replaced with the technical features with similar functions disclosed in the present disclosure (but not limited to) by each other.

Claims

1. A model training method, wherein: The method comprises: Acquire voice call data of a call between the user and the service provider before an order operation for an order between the user and the service provider occurs and a responsibility determination result corresponding to the order operation, wherein the responsibility determination result is used to indicate the probability that one of the user and the service provider induces the other to actively trigger the corresponding order operation in the corresponding voice call data; Performing voice recognition on the voice call data to obtain voice recognition text; Determine at least one target sentence in the speech recognition text, wherein the distance between the target sentence and the target scene word is less than or equal to the scene word distance, and the word frequency of the target operation inducing word in the target sentence is greater than or equal to the inducing word frequency threshold; Acquire a responsibility determination model, and use the at least one target statement as input, use the responsibility determination result as output, and train the responsibility determination model to acquire a target responsibility determination model; Before taking the at least one target statement as input and the responsibility determination result as output to train the responsibility determination model to obtain the target responsibility determination model, the method further includes: Receiving updated weight parameters sent by the edge server, and updating the responsibility determination model according to the updated weight parameters; The step of taking the at least one target statement as input, taking the responsibility determination result as output, and training the responsibility determination model to obtain a target responsibility determination model includes: Taking the at least one target sentence as input and the responsibility determination result as output, training the updated responsibility determination model; When the trained responsibility determination model has not converged, a gradient update vector is obtained according to the trained responsibility determination model, and the gradient update vector is sent to the edge server; When the trained responsibility determination model converges, the trained responsibility determination model is determined as the target responsibility determination model.

2. The model training method according to claim 1, wherein: Acquiring the voice call data includes: Receive authorization indication information, and acquire the voice call data in response to the authorization indication information.

3. The model training method according to claim 1, wherein: Before obtaining the responsibility determination result corresponding to the order operation, the method further includes: Obtaining the responsibility determination result corresponding to the order operation; Obtaining the responsibility determination result corresponding to the order operation, including: In response to the fact that the responsible party indicated by the responsibility determination result is the service party, and the service party has not sent any dispute arbitration request information within the responsibility determination time length after the time when the responsibility determination result is generated, a responsibility determination result is generated to indicate that the responsible party for the order operation is the service party.

4. A method for determining responsibility, wherein: The method comprises: Acquiring voice call data of a call between the user party and the service party before an order operation for a corresponding order between the user party and the service party occurs; Performing voice recognition on the voice call data to obtain voice recognition text; Determine at least one target sentence in the speech recognition text, wherein the distance between the target sentence and the target scene word is less than or equal to the scene word distance, and the word frequency of the target operation inducing word in the target sentence is greater than or equal to the inducing word frequency threshold; Obtain a target training responsibility determination model, and use the at least one target statement as input to obtain a responsibility determination result corresponding to the order operation output by the target training responsibility determination model, wherein the responsibility determination result is used to indicate the probability that one of the user party and the service party induces the other party to actively trigger the corresponding order operation in the corresponding voice call data.

5. The liability determination method according to claim 4, wherein: The target training responsibility determination model is obtained through the model training method described in any one of claims 1-3.

6. The liability determination method according to claim 4, wherein: The responsibility determination method is applied to an online responsibility determination system, and the target training responsibility determination model is obtained by training based on a text convolutional neural network model textcnn; Or, the responsibility determination method is applied to an offline responsibility determination system, and the target training responsibility determination model includes a first target training responsibility determination model obtained by training based on textCNN and a second target training responsibility determination model obtained by training based on a bidirectional encoding representation model BERT of a converter; The method further comprises: A combined responsibility determination result is obtained based on a first responsibility determination result output by the first target training responsibility determination model and a second responsibility determination result output by the second target training responsibility determination model.

7. A method for determining responsibility, wherein: The method comprises: Receive responsibility determination result indication information and display the responsibility determination result indication information, wherein the responsibility determination result is used to indicate the probability that one party between the user party and the service party induces the other party to actively trigger a corresponding order operation in the corresponding voice call data, and the responsibility determination result is obtained by any one of the methods in claims 4-6.

8. The liability determination method according to claim 7, wherein: Before receiving the responsibility determination result indication information and displaying the responsibility determination result indication information, the method further includes: In response to the input order operation action, order operation instruction information is sent, and the order operation instruction information is used to instruct to perform an order operation on the order between the user party and the service party.

9. The liability determination method according to claim 7, wherein: Before receiving the responsibility determination result indication information and displaying the responsibility determination result indication information, the method further includes: Sending voice call data of a call between the user party and the service party before an order operation for an order between the user party and the service party occurs.

10. The method for determining responsibility according to claim 9, wherein: Before receiving the responsibility determination result indication information and displaying the responsibility determination result indication information, the method further includes: In response to the input authorization indication action, authorization indication information is sent, where the authorization indication information is used to instruct the acquisition of the voice call data.

11. An electronic device comprising a memory and at least one processor; wherein: The memory is used to store one or more computer instructions, wherein the one or more computer instructions are executed by the at least one processor to implement the method steps described in any one of claims 1-10.

12. A computer-readable storage medium having computer instructions stored thereon, wherein the computer instructions, when executed by a processor, implement the method steps described in any one of claims 1 to 10.

13. A computer program product comprising a computer program / instructions, wherein: When the computer program / instructions are executed by a processor, the method steps described in any one of claims 1 to 10 are implemented.

Images