Dialogue processing method, device, equipment, storage medium and program

Abstract

The application provides a dialogue processing method, device, equipment, storage medium and program, which are applied to the field of artificial intelligence. The method comprises the following steps: obtaining dialogue data corresponding to a current session, wherein the dialogue data comprises dialogue information of at least one round of dialogue between a target user and a customer service; obtaining user behavior data corresponding to the current session, wherein the user behavior data comprises behavior information of at least one behavior of the target user within a preset time length before the current time; determining whether the current session needs to be converted into high-priority processing according to the dialogue data and the user behavior data; if yes, establishing a session connection between the target user and a high-priority processing flow, so that the high-priority processing flow provides services to the target user. Through the conversion of the current session into high-priority processing, the problem of the user can be quickly and accurately solved, and the satisfaction of the user is improved.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to a dialogue processing method, apparatus, device, storage medium, and program. Background Technology

[0002] With the development of artificial intelligence technology, intelligent customer service systems can provide services to users in many scenarios.

[0003] The process by which an intelligent customer service system provides services to users is as follows: The intelligent customer service system obtains the questions entered by users, performs semantic analysis on the questions to identify the user's inquiry intent, and then outputs the corresponding answers to the questions to the users based on the user's inquiry intent.

[0004] In practical applications, when users' questions are complex, the intelligent customer service system is unable to provide satisfactory answers quickly and accurately, resulting in low user satisfaction. Summary of the Invention

[0005] This application provides a dialogue processing method, apparatus, device, storage medium, and program that can predict in a timely manner whether the current conversation needs to be converted to high-priority processing during the dialogue between the target user and customer service, and convert the current conversation to high-priority processing when necessary, so as to improve user satisfaction.

[0006] In a first aspect, embodiments of this application provide a dialogue processing method, including:

[0007] Obtain the dialogue data corresponding to the current session; the dialogue data includes: dialogue information of at least one round of dialogue between the target user and customer service;

[0008] Obtain user behavior data corresponding to the current session; the user behavior data includes: behavioral information of at least one behavior of the target user within a preset time period before the current moment;

[0009] Based on the dialogue data and the user behavior data, determine whether the current session needs to be transferred to high-priority processing;

[0010] If the current session needs to be switched to high-priority processing, a session connection is established between the target user and the high-priority processing flow.

[0011] In one possible implementation, before determining whether the current session needs to be transferred to high-priority processing based on the dialogue data and the user behavior data, the method further includes:

[0012] Obtain the associated object data corresponding to the current session, wherein the associated object data includes: object information of at least one object related to the current session;

[0013] Based on the dialogue data and the user behavior data, determining whether the current session needs to be transferred to high-priority processing includes:

[0014] Based on the dialogue data, the user behavior data, and the associated object data, determine whether the current session needs to be converted to high-priority processing.

[0015] In one possible implementation, determining whether the current session needs to be converted to high-priority processing based on the dialogue data, the user behavior data, and the associated object data includes:

[0016] Based on the dialogue data, the user behavior data, and the associated object data, a first session feature corresponding to the current session is generated;

[0017] Based on the first session characteristics, determine whether the current session needs to be transferred to high-priority processing.

[0018] In one possible implementation, the dialogue information for each round of conversation includes: a first statement sent by the target user and a second statement output by the customer service representative in response to the first statement; based on the dialogue data, the user behavior data, and the associated object data, a first conversation feature corresponding to the current conversation is generated, including:

[0019] Based on the dialogue data, a first dialogue sequence feature and a second dialogue sequence feature are generated; the first dialogue sequence feature includes: the feature expression result corresponding to at least one first statement in the dialogue data, and the second dialogue sequence feature includes: the feature expression result corresponding to at least one second statement in the dialogue data;

[0020] Based on the user behavior data, a user behavior sequence feature is generated; the user behavior sequence feature includes: the feature expression result corresponding to at least one behavior information in the user behavior data;

[0021] Based on the associated object data, associated object features are generated; the associated object features include: feature expression results corresponding to at least one object information in the associated object data;

[0022] Based on the first dialogue sequence features, the second dialogue sequence features, the user behavior sequence features, and the associated object features, a first session feature corresponding to the current session is generated.

[0023] In one possible implementation, determining whether the current session needs to be transferred to high-priority processing based on the first session characteristics includes:

[0024] The first session features are processed by a pre-trained prediction model to obtain the probability that the target user expects the current session to be converted to a high-priority processing.

[0025] Based on the probability, determine whether the current session needs to be switched to high-priority processing.

[0026] In one possible implementation, the prediction model is trained in the following manner:

[0027] Obtain training samples, which include: second session features corresponding to the sample session and first annotation data corresponding to the sample session. The first annotation data is used to indicate whether the sample user instructs to switch to high-priority processing during the sample session.

[0028] The second session features are processed by the prediction model to be trained to obtain a first prediction result, which is used to indicate the probability that the sample user expects the sample session to be converted to high priority processing.

[0029] Based on the first prediction result and the first labeled data, the model parameters of the prediction model are updated, and the update objective includes minimizing the difference between the first prediction result and the first labeled data.

[0030] In one possible implementation, determining whether the current session needs to be transferred to high-priority processing based on the first session characteristics includes:

[0031] The first session features are processed by a pre-trained prediction model to obtain the probability that the target user expects the current session to be converted to a high-priority processing, and the target user's satisfaction level with the current session;

[0032] Based on the probability and the level of satisfaction, determine whether the current session needs to be transferred to high-priority processing.

[0033] In one possible implementation, the prediction model is trained in the following manner:

[0034] Obtain training samples, which include: second session features corresponding to the sample session, first labeled data corresponding to the sample session, and second labeled data corresponding to the sample session; wherein, the first labeled data is used to indicate whether the sample user instructs to switch to high-priority processing during the sample session, and the second labeled data is used to indicate the sample user's satisfaction level with the sample session;

[0035] The second session features are processed by the prediction model to be trained to obtain a first prediction result and a second prediction result; wherein, the first prediction result is used to indicate the probability that the sample user expects the sample session to be converted to high priority processing, and the second prediction result is used to indicate the sample user's satisfaction with the sample session;

[0036] Based on the first prediction result, the first labeled data, the second prediction result, and the second labeled data, the model parameters of the prediction model are updated. The update objectives include minimizing the difference between the first prediction result and the first labeled data, and minimizing the difference between the second prediction result and the second labeled data.

[0037] In one possible implementation, obtaining the user behavior data corresponding to the current session includes:

[0038] Based on the current session, the identifier of the target user and the target time period are determined, wherein the start time of the target time period is before the current session and the end time of the target time period is the current time.

[0039] Based on the target user's identifier and the target time period, the user's behavior information within the target time period is queried from a preset database to obtain the user behavior data; the preset database is used to store historical behavior information of different users.

[0040] Secondly, embodiments of this application provide a dialogue processing apparatus, including:

[0041] The first acquisition module is used to acquire the dialogue data corresponding to the current session; the dialogue data includes: dialogue information of at least one round of dialogue between the target user and the customer service representative;

[0042] The second acquisition module is used to acquire user behavior data corresponding to the current session; the user behavior data includes: behavior information of at least one behavior of the target user within a preset time period before the current moment;

[0043] The determination module is used to determine whether the current session needs to be transferred to high-priority processing based on the dialogue data and the user behavior data;

[0044] The control module is used to establish a session connection between the target user and the high-priority processing flow if the current session needs to be switched to high-priority processing.

[0045] Thirdly, embodiments of this application provide an electronic device, including: a memory and at least one processor, wherein the memory stores a computer program configured to be executed by the processor to implement the method as described in any of the first aspects.

[0046] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method described in any of the first aspects.

[0047] Fifthly, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the method described in any of the first aspects.

[0048] This application provides a dialogue processing method, apparatus, device, storage medium, and program. The method includes: acquiring dialogue data corresponding to the current session, the dialogue data including dialogue information of at least one round of dialogue between a target user and customer service; acquiring user behavior data corresponding to the current session, the user behavior data including behavior information of at least one behavior of the target user within a preset time period before the current moment; determining whether the current session needs to be transferred to high-priority processing based on the dialogue data and user behavior data; if so, establishing a session connection between the target user and the high-priority processing flow so as to provide services to the target user through the high-priority processing flow. In the above process, when it is predicted based on the dialogue data and user behavior data that the current session needs to be transferred to high-priority processing, the system automatically switches to providing services to the user through the high-priority processing flow. On the one hand, this allows the user's problems to be resolved quickly and accurately, improving user satisfaction; on the other hand, since the current session is transferred to high-priority processing only when it needs to be transferred, the transfer rate of the high-priority processing flow can be reduced, thus reducing the customer service cost of the high-priority processing flow. Furthermore, when predicting whether to prioritize a task, both dialogue data and user behavior data are considered simultaneously, resulting in higher accuracy in the predictions. This helps to minimize the impact of the dialogue process on the user experience and improve user satisfaction. Attached Figure Description

[0049] Figure 1 A schematic diagram illustrating an application scenario provided in an embodiment of this application;

[0050] Figure 2 A flowchart illustrating a dialogue processing method provided in an embodiment of this application;

[0051] Figure 3 A flowchart illustrating another dialogue processing method provided in an embodiment of this application;

[0052] Figure 4 A schematic diagram illustrating the generation process of a first session feature provided in an embodiment of this application;

[0053] Figure 5 A schematic diagram illustrating the input and output of a prediction model provided in an embodiment of this application;

[0054] Figure 6 A schematic diagram of the input and output of another prediction model provided in an embodiment of this application;

[0055] Figure 7 A schematic diagram illustrating a dialogue processing procedure provided in an embodiment of this application;

[0056] Figure 8 This is a schematic diagram of the structure of a dialogue processing device provided in an embodiment of this application;

[0057] Figure 9 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0058] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0059] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in a sequence other than that illustrated or described herein.

[0060] It should be understood that the terms “comprising” and “having” as used herein, and any variations thereof, are intended to cover non-exclusive inclusion, for example, a process, method, system, product, or apparatus that includes a series of steps or units is not necessarily limited to those steps or units that are expressly listed, but may include other steps or units that are not expressly listed or that are inherent to such process, method, product, or apparatus.

[0061] In the description of the embodiments of this application, the term "correspondence" may indicate that there is a direct or indirect correspondence between two things, or that there is an association between two things, or that there is a relationship of instruction and being instructed, configuration and being configured, etc.

[0062] To facilitate understanding of the technical solutions of this application, the concepts and terms involved in the embodiments of this application will be explained first.

[0063] Intelligent customer service systems refer to customer service systems that use intelligent robots to replace human customer service representatives in providing services to users. In some scenarios, intelligent customer service systems can also be called intelligent robots. Intelligent customer service systems can be applied to various scenarios, including but not limited to: e-commerce customer service scenarios, insurance customer service scenarios, financial customer service scenarios, and catering customer service scenarios. Taking e-commerce customer service scenarios as an example, intelligent customer service systems can replace the store / merchant's human customer service representatives to provide users with services such as product consultation and after-sales service.

[0064] Human-machine collaboration refers to the collaborative work between chatbots and human customer service representatives, using collaborative mechanisms to jointly resolve user issues.

[0065] Switching to high-priority processing means that during the process of customer service providing services to a user, the customer service representative is switched to a high-priority processing flow, where the high-priority process continues to provide services to the user. Here, "customer service representative" can refer to an automated customer service system. "High-priority processing" can refer to human customer service or other high-priority processing flows.

[0066] In this embodiment, the customer service representative can be deployed on a local server or a cloud server. The dialogue process between the user and the customer service representative is as follows: The user inputs a question, such as "What is product A made of?", which is sent to the customer service representative. The customer service representative performs semantic analysis on the question to identify the user's inquiry intent. Then, based on the inquiry intent and a preset knowledge base, the customer service representative generates the corresponding answer to the question, for example, the generated answer is "Product A is made of pure cotton," and provides this answer back to the user.

[0067] The above process facilitates dialogue between customer service and users. In practice, when users' questions are complex, customer service representatives are unable to provide the answers they need quickly and accurately, resulting in lower user satisfaction.

[0068] In some possible solutions, the user's emotions could be identified based on the conversation data during the interaction with customer service. When negative emotions are detected, the conversation could be transferred to a human agent for assistance. However, the detection of negative emotions indicates that the user experience has been negatively impacted; therefore, even with this approach, user satisfaction remains low.

[0069] This application provides a dialogue processing method, apparatus, device, storage medium, and program to improve user satisfaction with the dialogue process. The following is in conjunction with... Figure 1 The technical concept of this application is explained.

[0070] Figure 1 This is a schematic diagram illustrating an application scenario provided by an embodiment of this application. For example... Figure 1 As shown, during a conversation between a user and customer service, the customer service representative can predict whether the current conversation needs to be prioritized for high-priority processing based on the conversation data and user behavior data corresponding to the current session. The conversation data includes information from at least one round of dialogue between the user and customer service. The user behavior data includes behavioral information from at least one action taken by the user within a preset time period prior to the current moment. If it is predicted that high-priority processing is required, a session connection is established between the user and the high-priority processing flow, meaning that services are provided to the user through the high-priority processing flow.

[0071] In this technical solution, the need for switching to high-priority processing is predicted in real time during the conversation. When this is predicted, the system automatically switches to high-priority processing to provide service to the user. This allows user issues to be resolved quickly and accurately, improving user satisfaction. Furthermore, because the current conversation is only switched to high-priority processing when it is necessary, the transfer rate of high-priority processing is reduced, lowering customer service costs. Moreover, the prediction of whether to switch to high-priority processing considers both conversation data and user behavior data, resulting in high accuracy and minimizing the impact of the conversation on user experience, thus further enhancing user satisfaction.

[0072] The collection, storage, use, processing, transmission, provision, and disclosure of dialogue data and user behavior data involved in the technical solution of this application all comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0073] The technical solutions provided by the embodiments of this application will be described in detail below through specific examples. It should be noted that the technical solutions provided by the embodiments of this application may include some or all of the following contents. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0074] Figure 2 This is a flowchart illustrating a dialogue processing method provided in an embodiment of this application. Figure 2 As shown, the method in this example includes:

[0075] S201: Obtain the dialogue data corresponding to the current session; the dialogue data includes: dialogue information of at least one round of dialogue between the target user and the customer service representative.

[0076] In this embodiment, the current session refers to the ongoing conversation between the target user and the customer service representative. A single session may include multiple rounds of dialogue. Optionally, the customer service representative mentioned above can be an intelligent customer service system.

[0077] For example, in an e-commerce customer service scenario, the first round of dialogue is: the user types "Hello," and the customer service replies "Hello." The second round of dialogue is: the user types "What is the material of product A?" and the customer service rep replies "Product A is made of pure cotton."

[0078] Optionally, in each round of the above dialogue, the user can input dialogue statements through the first terminal device. In other words, the dialogue data may include dialogue information from at least one round of dialogue between the target user's corresponding first terminal device and the customer service representative.

[0079] In some possible implementations, the dialogue processing method provided in this embodiment can be executed periodically during the dialogue between the target user and customer service. For example, the dialogue processing method provided in this embodiment can be executed once every preset time interval.

[0080] In other possible implementations, the dialogue processing method provided in this embodiment can be executed after each round of dialogue during the conversation between the target user and customer service. For example, after the first round of dialogue is completed, the dialogue processing method provided in this embodiment is executed once; after the second round of dialogue is completed, the dialogue processing method provided in this embodiment is executed once; and so on.

[0081] Each time the dialogue processing method provided in this embodiment is executed, in S201, dialogue information of one or more rounds of dialogue completed between the target user and customer service is obtained. Referring to the above example, if the method of this embodiment is executed after the completion of the second round of dialogue, the dialogue data obtained in S201 includes dialogue information from the first round and dialogue information from the second round.

[0082] The dialogue information in the first round includes: "Hello", "Hello";

[0083] The dialogue information in the second round included: "What is product A made of?" and "Product A is made of pure cotton."

[0084] It should be understood that the dialogue information in each round of dialogue can be in text form, voice form, or other forms, and this embodiment does not limit this.

[0085] S202: Obtain user behavior data corresponding to the current session; user behavior data includes: behavior information of at least one behavior of the target user within a preset time period before the current moment.

[0086] Optionally, the aforementioned at least one behavior can be an action performed by the target user through the first terminal device. That is, the user behavior data can include: behavior information of at least one behavior performed by the target user through the first terminal device within a preset time period prior to the current moment.

[0087] In this embodiment, considering whether the dialogue between the target user and customer service needs to be switched to high-priority processing, in addition to the dialogue information of each round of dialogue, it is also related to the target user's behavior in the most recent time period. Therefore, in S202, behavioral information of each behavior of the target user within a preset time period before the current moment can also be obtained, and this behavioral information is combined with the dialogue information of each round of dialogue to comprehensively predict whether the current session needs to be switched to high-priority processing.

[0088] Taking e-commerce customer service as an example, the aforementioned at least one behavior may include, but is not limited to: the target user's click behavior on the display page of the first terminal device, the behavior of adding the product to favorites, the behavior of adding the product to the shopping cart, the behavior of placing an order for the product, etc. Optionally, the aforementioned at least one behavior may be a sequence of behaviors arranged in chronological order.

[0089] In some possible implementations, the first terminal device detects the target user's behavior. When a behavior is detected, the behavior information is stored in a preset database. The behavior information for each behavior may include: the type of the behavior, the object operated on by the behavior, and the execution timestamp of the behavior. The preset database is used to store historical behavior information for different users.

[0090] In one example, the contents of the preset database can be as shown in Table 1.

[0091] Table 1

[0092]

[0093]

[0094] Thus, in S202, user behavior data corresponding to the current session can be retrieved from a preset database. For example, based on the current conversation, the target user's identifier and target time period can be determined. The start time of the target time period is before the current session, and the end time of the target time period is the current time. For instance, assuming the target user initiated the current session at 8:00, and the current time is 8:02, then the target time period could be from 7:30 to 8:02. Based on the target user's identifier and the target time period, the user's behavior information within the target time period is queried from the preset database to obtain user behavior data.

[0095] It should be noted that the execution order of S201 and S202 is not limited; their order can be interchanged, or they can be executed in parallel.

[0096] S203: Based on dialogue data and user behavior data, determine whether the current session needs to be transferred to high-priority processing.

[0097] In this embodiment, switching to high-priority processing can mean switching to human customer service or switching to a high-priority processing flow.

[0098] For example, a first session feature corresponding to the current session can be determined based on dialogue data and user behavior data. For instance, the first session feature can be obtained by extracting features from the dialogue data and user behavior data. Then, based on the first session feature, it can be determined whether the current session needs to be transferred to high-priority processing. In some possible implementations, some recognition rules can be pre-set, and the determination of whether the current session needs to be transferred to high-priority processing can be made based on the first session feature and the pre-set recognition rules. For example, if the first session feature meets the pre-set recognition rules, then it is determined that the current session needs to be transferred to high-priority processing. If the first session feature does not meet the pre-set recognition rules, then it is determined that the current session does not need to be transferred to high-priority processing.

[0099] In other possible implementations, a prediction model can be pre-trained. This model has the ability to predict whether the current session needs to be prioritized. The features of the first session can be processed using the prediction model, and the output of the model can be used to determine whether the current session needs to be prioritized.

[0100] In this embodiment, since both dialogue data and user behavior data are considered when predicting whether the current session needs to be switched to high-priority processing, the prediction results are more accurate.

[0101] S204: If the current session needs to be transferred to high-priority processing, then establish a session connection between the target user and the high-priority processing flow.

[0102] Here, a high-priority processing flow refers to a processing flow with high priority. For example, a high-priority processing flow could be a human customer service representative, or it could be any other high-priority processing flow.

[0103] Optionally, when the high-priority processing flow is human customer service, establishing a session connection between the target user and the high-priority processing flow may include: establishing a session connection between the first terminal device corresponding to the target user and the second terminal device corresponding to the human customer service representative.

[0104] It should be understood that after establishing a session connection between the target user and the high-priority processing flow, the high-priority processing flow replaces customer service in conducting subsequent rounds of dialogue with the target user, thereby quickly and accurately resolving the target user's problem.

[0105] In some possible implementations, taking a high-priority processing flow with human customer service as an example, when multiple human customer service representatives exist, the following approach can be used: Based on the status of the multiple human customer service representatives, at least one candidate human customer service representative is identified from among them, and the candidate human customer service representative is in an idle state. According to a preset scheduling rule, a target human customer service representative is determined from the at least one candidate human customer service representative, and a session connection is established between the first terminal device and the second terminal device corresponding to the target human customer service representative. Through the above process, it is ensured that the service provided to the target user is provided by an idle human customer service representative, avoiding long waiting times for the target user and further improving the target user's satisfaction.

[0106] In some possible implementations, if the current session does not need to be transferred to high-priority processing, the customer service representative continues to provide service to the target user.

[0107] The dialogue processing method provided in this embodiment includes: acquiring dialogue data corresponding to the current session, the dialogue data including dialogue information of at least one round of dialogue between the target user and customer service; acquiring user behavior data corresponding to the current session, the user behavior data including behavior information of at least one behavior of the target user within a preset time period before the current moment; determining whether the current session needs to be transferred to high-priority processing based on the above dialogue data and user behavior data; if so, establishing a session connection between the target user and the high-priority processing flow so as to provide services to the target user through the high-priority processing flow. In the above process, when it is predicted based on the dialogue data and user behavior data that the current session needs to be transferred to high-priority processing, the service is automatically switched to the high-priority processing flow to provide services to the user. On the one hand, this enables the user's problem to be resolved quickly and accurately, improving user satisfaction; on the other hand, since the current session is transferred to high-priority processing only when it needs to be transferred to high-priority processing, the transfer rate of the high-priority processing flow can be reduced, thus reducing the customer service cost of the high-priority processing flow. Furthermore, when predicting whether to prioritize a task, both dialogue data and user behavior data are considered simultaneously, resulting in higher accuracy in the predictions. This helps to minimize the impact of the dialogue process on the user experience and improve user satisfaction.

[0108] In the above Figure 2 Based on the illustrated embodiment, the following is combined with Figures 3 to 7 A more detailed description of the technical solution in this application is provided.

[0109] Figure 3This is a flowchart illustrating another dialogue processing method provided in an embodiment of this application. Figure 3 As shown, the method in this example includes:

[0110] S301: Obtain the dialogue data corresponding to the current session; the dialogue data includes: dialogue information of at least one round of dialogue between the target user and the customer service representative.

[0111] S302: Obtain user behavior data corresponding to the current session; user behavior data includes: behavior information of at least one behavior of the target user within a preset time period before the current moment.

[0112] It should be understood that the implementation methods of S301 and S302 mentioned above can be found in the relevant descriptions of S201 and S202, and will not be repeated here.

[0113] S303: Obtain the associated object data corresponding to the current session. The associated object data includes: object information of at least one object related to the current session.

[0114] The objects related to the current session include, but are not limited to: the objects discussed in the current session, the objects mentioned in the current session, and the objects represented by the customer service representative. Taking e-commerce customer service as an example, at least one object related to the current session may include one or more of the following: the products discussed / mentioned in the current session, the orders discussed / mentioned in the current session, and the merchants / stores represented by the customer service representative.

[0115] For each object associated with the current session, the object information may include: the object's identifier, the object's category, the user satisfaction of the historical sessions associated with the object, and whether the historical sessions associated with the object have been converted to high-priority processing, etc.

[0116] It should be understood that the execution order of S301 to S303 above can be arbitrary, and this embodiment does not limit it.

[0117] In this embodiment, after obtaining the dialogue data, user behavior data, and associated object data corresponding to the current session, it can be determined whether the current session needs to be transferred to high-priority processing based on the dialogue data, user behavior data, and associated object data. See the detailed descriptions in S304 to S305 for more details.

[0118] S304: Generate the first session feature corresponding to the current session based on dialogue data, user behavior data, and associated object data.

[0119] For example, the first session feature corresponding to the current session can be obtained by extracting or representing features from dialogue data, user behavior data, and associated object data.

[0120] The dialogue information for each round includes: the first statement sent by the target user, and the second statement output by the customer service representative in response to the first statement. Figure 4 This is a schematic diagram illustrating the generation process of a first session feature provided in an embodiment of this application. Figure 4 As shown, the first session features can be generated in the following way:

[0121] (1) Generate a first dialogue sequence feature and a second dialogue sequence feature based on the dialogue data. The first dialogue sequence feature includes the feature expression result corresponding to at least one first statement in the dialogue data; the second dialogue sequence feature includes the feature expression result corresponding to at least one second statement in the dialogue data.

[0122] For example, suppose the dialogue data includes dialogue information from the first round of dialogue and dialogue information from the second round of dialogue. In the first round of dialogue, the first statement is "Hello" and the second statement is "Hello"; in the second round of dialogue, the first statement is "What is the material of product A?" and the second statement is "Product A is made of pure cotton".

[0123] The first statement "Hello" in the first round of dialogue and the first statement "What is the material of product A?" in the second round of dialogue are processed for feature representation, resulting in feature vectors (i.e., feature representation results) corresponding to the two first statements. The feature vectors are then arranged according to the chronological order of the two first statements to form the first dialogue sequence features, as follows:

[0124] {Feature vector corresponding to “Hello”, feature vector corresponding to “What material is product A made of”}.

[0125] The second statement "Hello" in the first round of dialogue and the second statement "Product A is made of pure cotton" in the second round of dialogue are processed for feature representation, resulting in feature vectors (i.e., feature representation results) corresponding to the two second statements. The feature vectors are then arranged according to the order of the two second statements to form the second dialogue sequence features, as follows:

[0126] {Feature vector corresponding to “Hello”, feature vector corresponding to “Product A is made of pure cotton”}.

[0127] (2) Generate user behavior sequence features based on user behavior data; user behavior sequence features include: feature expression results corresponding to at least one behavior information in the user behavior data.

[0128] For example, suppose user behavior data includes the following three actions: clicking on product A, adding product A to the shopping cart, and clicking on product B. Feature representation processing can be performed on the behavior information of each action to obtain a feature vector (i.e., the feature representation result) corresponding to each action. Arranging the corresponding feature vectors according to the order of the three actions forms the user behavior sequence features, as follows:

[0129] {Feature vector corresponding to “Click on product A”, feature vector corresponding to “Add product A to cart”, and feature vector corresponding to “Click on product B”}.

[0130] (3) Generate associated object features based on associated object data; associated object features include: feature expression results corresponding to at least one object information in the associated object data.

[0131] For example, suppose the associated object data includes object information for two objects: Product A and Product B. Feature representation processing can be performed on the object information of each object to obtain a feature vector (i.e., feature representation result) for each object. Based on the feature vectors corresponding to the two objects above, the associated object features are generated as follows:

[0132] {Feature vector corresponding to “Product A”, feature vector corresponding to “Product B”}.

[0133] (4) Generate the first session feature corresponding to the current session based on the first dialogue sequence feature, the second dialogue sequence feature, the user behavior sequence feature, and the associated object feature.

[0134] For example, the first dialogue sequence features, the second dialogue sequence features, the user behavior sequence features, and the associated object features can be concatenated dimensionally to obtain the first dialogue feature corresponding to the current dialogue.

[0135] In the above implementation, the accuracy of the first session features is improved by converting the dialogue data and user behavior data into the form of sequence features, which can accurately express the behavioral characteristics of the target user before and during the current session.

[0136] In some possible implementations, the features in steps (1) to (3) above may cover multiple data types, such as discrete types, continuous types, sequence types, etc. Therefore, before performing dimensional concatenation on each feature, the features can be processed first to improve their quality and facilitate the processing of the prediction model in subsequent steps.

[0137] For example, for discrete types, the processing method may include: converting discrete information into one-dimensional discrete features through dictionary-based processing, and then performing N-dimensional transformation processing on the one-dimensional discrete features to obtain an N-dimensional feature vector. N is a preset integer.

[0138] For continuous data types, considering that the numerical value ranges of different features may be different, the processing methods may include: normalization.

[0139] Regarding sequence types, considering that different sequences are typically of variable length, and prediction models cannot handle variable-length feature sequences during prediction, processing methods can include converting the variable-length sequence into a fixed-length sequence. For example, by performing a mask transformation on the variable-length sequence to convert it into a fixed-length sequence, using a preset character (e.g., 0) as a placeholder. This way, during prediction, the influence of the mask position can be ignored based on the preset character, allowing the model to still read the feature values ​​of the original variable-length sequence.

[0140] S305: Based on the characteristics of the first session, determine whether the current session needs to be transferred to high-priority processing.

[0141] In this embodiment, a pre-trained prediction model can be used to predict whether the current session needs to be switched to high-priority processing based on the first session features.

[0142] In some possible implementations, the determination of whether the current session needs to be prioritized can be made as follows: The features of the first session are processed using a pre-trained prediction model to obtain the probability that the target user expects the current session to be prioritized. For example, Figure 5 This is a schematic diagram illustrating the input and output of a prediction model provided in an embodiment of this application. Figure 5 As shown, the first session features are input into the prediction model, and the prediction model outputs the probability that the target user expects the current session to be converted to high-priority processing.

[0143] Furthermore, based on the probabilities output by the prediction model, it can be determined whether the current session needs to be switched to high-priority processing. For example, if the probability is greater than or equal to a preset threshold, it is determined that the current session needs to be switched to high-priority processing. If the probability is less than the preset threshold, it is determined that the current session does not need to be switched to high-priority processing.

[0144] The aforementioned prediction model is obtained by learning the session features of historical sessions. During the training process, the prediction model can learn under what circumstances users expect to switch to high-priority processing and under what circumstances they do not expect to switch to high-priority processing, thus having the ability to predict whether the current session needs to be switched to high-priority processing.

[0145] In some possible implementations, the above prediction model can be trained in the following way:

[0146] (1) Obtain training samples, which include: second session features corresponding to the sample session and first annotation data corresponding to the sample session. The first annotation data is used to indicate whether the sample user indicates to switch to high priority processing during the sample session.

[0147] For example, multiple sets of training samples can be generated using relevant data from sample sessions, where the sample sessions can be historical sessions. The generation process for each set of training samples can be as follows: First, obtain the dialogue data corresponding to the sample session, which includes dialogue information from at least one round of dialogue between the sample user and customer service. Second, obtain the user behavior data corresponding to the sample session, which includes behavioral information from at least one action taken by the sample user within a preset time period prior to the current moment. Third, generate a second session feature corresponding to the sample session based on the dialogue data and the user behavior data. Finally, determine whether the sample user indicated a shift to high-priority processing during the sample session, and obtain the first labeled data corresponding to the sample session based on the determination result.

[0148] (2) The second session features are processed by the prediction model to be trained to obtain the first prediction result, which is used to indicate the probability that the sample user expects to transfer the sample session to high priority processing.

[0149] For example, the second session features are input into the prediction model, the prediction model makes a prediction based on the second session features, and outputs a first prediction result.

[0150] (3) Update the model parameters of the prediction model based on the first prediction result and the first labeled data; the goal of the update includes minimizing the difference between the first prediction result and the first labeled data.

[0151] For example, a first loss function can be determined based on the first prediction result and the first labeled data. The first loss function is used to indicate the difference between the first prediction result and the first labeled data. The model parameters of the prediction model are updated with the goal of minimizing the loss function.

[0152] Following step (3) above, the process may further include: determining whether the updated prediction model meets the preset convergence condition. If the preset convergence condition is not met, steps (1) to (3) above are repeated until the updated prediction model meets the preset convergence condition, at which point the updated prediction model is used as the trained prediction model. In this way, the trained prediction model has the following capability: predicting whether the current session needs to be switched to high-priority processing.

[0153] This is understandable. Since the prediction model learns from the sample users' behavior of prioritizing sample sessions during training, it can predict whether the current session needs to be prioritized, thus improving the accuracy of the prediction results.

[0154] In some possible implementations, determining whether the current session needs to be prioritized can be achieved by processing the features of the first session using a pre-trained prediction model to obtain the probability that the target user expects the current session to be prioritized and the target user's level of satisfaction with the current session. For example, Figure 6 This is a schematic diagram illustrating the input and output of another prediction model provided in an embodiment of this application. For example... Figure 6 As shown, the first session features are input into the prediction model, and the prediction model outputs the following two items: the probability that the target user expects to transfer the current session to a high priority, and the target user's satisfaction with the current session.

[0155] Furthermore, based on the probabilities and satisfaction levels output by the prediction model, it can be determined whether the current session needs to be switched to high-priority processing. For example, if the satisfaction level is less than or equal to a first preset threshold and the probability is greater than or equal to a second preset threshold, then the current session needs to be switched to high-priority processing. If the satisfaction level is greater than the first preset threshold and / or the probability is less than the second preset threshold, then the current session does not need to be switched to high-priority processing.

[0156] In some possible implementations, the above prediction model can be trained in the following way:

[0157] (1) Obtain training samples, which include: second session features corresponding to the sample session, first annotation data corresponding to the sample session, and second annotation data corresponding to the sample session; wherein, the first annotation data is used to indicate whether the sample user indicates to switch to high priority processing during the sample session; the second annotation data is used to indicate the sample user's satisfaction with the sample session.

[0158] It should be understood that, based on the aforementioned training method step (1), the sample sessions can also be labeled according to the satisfaction scores of the sample users for the sample sessions, thus obtaining the second labeled data.

[0159] (2) The second session features are processed by the prediction model to be trained to obtain the first prediction result and the second prediction result. The first prediction result is used to indicate the probability that the sample user expects the sample session to be transferred to high priority processing, and the second prediction result is used to indicate the sample user's satisfaction with the sample session.

[0160] For example, the second session features are input into the prediction model, and the prediction model outputs a first prediction result and a second prediction result.

[0161] (3) Update the model parameters of the prediction model based on the first prediction result, the first labeled data, the second prediction result and the second labeled data; the update objectives include minimizing the difference between the first prediction result and the first labeled data, and minimizing the difference between the second prediction result and the second labeled data.

[0162] For example, a first loss function can be generated based on a first prediction result and first labeled data, the first loss function indicating the difference between the first prediction result and the first labeled data; a second loss function can be generated based on a second prediction result and second labeled data, the second loss function indicating the difference between the second prediction result and the second labeled data; a target loss function is determined based on the first loss function and the second loss function. The model parameters of the prediction model are updated with the goal of minimizing the target loss function.

[0163] After the above training process, the trained prediction model has the following capabilities: predicting the target user's satisfaction level with the current session and predicting whether the current session needs to be transferred to high-priority processing.

[0164] This is understandable. Because the prediction model learns not only the sample users' tendency to prioritize sample sessions during training, but also their satisfaction with those sessions, it can predict the probability that a target user expects the current session to be prioritized and the target user's level of satisfaction with the current session. By combining these two prediction results, it can comprehensively determine whether the current session needs to be prioritized, thus improving the accuracy of the prediction results.

[0165] It should be noted that the network structure of the prediction model is not limited in the embodiments of this application. For example, the prediction model can employ a convolutional neural network model, a recurrent neural network model, or other network structures. Optionally, the prediction model can employ a click-through rate (CTR) prediction model. The CTR model has good expressive power for sequence features.

[0166] S306: If the current session needs to be transferred to high-priority processing, then establish a session connection between the target user and the high-priority processing flow.

[0167] It should be understood that the implementation of S306 can be found in [reference needed]. Figure 2 The relevant description of S204 will not be repeated here.

[0168] In this embodiment, when predicting whether the current session needs to be transferred to high-priority processing, in addition to considering human dialogue data and user behavior data, associated object data is also considered. By considering data from multiple dimensions, the prediction results are more accurate, and the current session can be transferred to high-priority processing at the appropriate time, thereby further improving user satisfaction.

[0169] Based on any of the above embodiments, the following is combined with Figure 7 The technical solution of this application is illustrated with examples. Figure 7 In the case of e-commerce, let's assume that the high-priority processing flow is human customer service.

[0170] Figure 7 This is a schematic diagram of a dialogue processing procedure provided in an embodiment of this application. In this embodiment, a preset database can be constructed in advance based on the massive amount of dialogue logs, user behavior logs, product inquiry and purchase logs, order status logs, etc. accumulated in the intelligent customer service system.

[0171] When a target user initiates a conversation, the intelligent customer service system first receives the target user. During the dialogue between the target user and the intelligent customer service system, after the completion of the i-th round of dialogue, the following steps (1) to (6) are triggered. Here, i takes the values ​​of integers such as 1, 2, 3, etc.

[0172] (1) Obtain dialogue data. Dialogue data includes dialogue information from the previous i rounds of the current session.

[0173] (2) Obtain user behavior data. See [link / reference] Figure 7 It can determine the target user's identifier and target time period based on the current session, and query the target user's behavior information (such as clicking on product A, adding product A to the shopping cart, clicking on product B, etc.) from the preset database to obtain user behavior data.

[0174] (3) Retrieve data from associated objects. See also Figure 7 Based on the current session, it can identify one or more objects related to the current session, such as the identifier of the product discussed / mentioned in the current session, the identifier of the order discussed / mentioned in the current session, the identifier of the merchant / store, etc. It then queries the object information of each object from a preset database to obtain the associated object data.

[0175] (4) Generate the first session features based on dialogue data, user behavior data, and associated object data. The process for generating the first session features can be found in [reference needed]. Figure 4 The relevant descriptions of the embodiments shown are not repeated here.

[0176] (5) Process the features of the first session using a prediction model to obtain the prediction results. The prediction results may include: the probability that the target user expects the current session to be transferred to a human customer service representative. Alternatively, the prediction results may include: the probability that the target user expects the current session to be transferred to a human customer service representative, and the target user's satisfaction level with the current session.

[0177] (6) Determine whether the current conversation needs to be transferred to a human customer service representative based on the prediction results. If so, a human customer service representative will provide service to the target user. If not, the intelligent customer service system will continue to provide service to the user.

[0178] It should be understood that the implementation methods of steps (5) and (6) above can be found in [reference needed]. Figure 3 The detailed description of the illustrated embodiments is omitted here.

[0179] In this embodiment, the first conversation feature corresponding to the current conversation is generated from multiple dimensions such as dialogue dimension, user behavior dimension, merchant dimension, product dimension, and order dimension. The feature dimensions considered are more comprehensive and detailed, which makes the prediction result of transferring to human customer service more accurate. This allows the current conversation to be transferred to human customer service at the appropriate time, which improves user satisfaction and the efficiency of human-machine collaboration.

[0180] The dialogue processing method provided by the embodiments of this application has been described above. The dialogue processing apparatus provided by the embodiments of this application will be described below.

[0181] This application embodiment can divide the dialogue processing device into functional modules according to the above method embodiment. For example, each function can be divided into a separate functional module, or two or more functions can be integrated into one processing module. The integrated module can be implemented in hardware or as a software functional module.

[0182] It should be noted that the module division in this embodiment is illustrative and represents only one logical functional division. In actual implementation, other division methods may be used. The following explanation uses the division of functional modules according to their respective functions as an example.

[0183] Figure 8 This is a schematic diagram of the structure of a dialogue processing device provided in an embodiment of this application. Figure 8 As shown, the dialogue processing device 800 provided in this embodiment includes:

[0184] The first acquisition module 801 is used to acquire the dialogue data corresponding to the current session; the dialogue data includes: dialogue information of at least one round of dialogue between the target user and the customer service representative;

[0185] The second acquisition module 802 is used to acquire user behavior data corresponding to the current session; the user behavior data includes: behavior information of at least one behavior of the target user within a preset time period before the current moment;

[0186] The determining module 803 is used to determine whether the current session needs to be transferred to high-priority processing based on the dialogue data and the user behavior data;

[0187] The control module 804 is used to establish a session connection between the target user and the high-priority processing flow if the current session needs to be switched to high-priority processing.

[0188] In one possible implementation, the apparatus of this embodiment further includes:

[0189] The third acquisition module is used to acquire the associated object data corresponding to the current session, the associated object data including: object information of at least one object related to the current session;

[0190] The determination module 803 is specifically used to: determine whether the current session needs to be converted to high-priority processing based on the dialogue data, the user behavior data, and the associated object data.

[0191] In one possible implementation, the determining module 803 is specifically used for:

[0192] Based on the dialogue data, the user behavior data, and the associated object data, a first session feature corresponding to the current session is generated;

[0193] Based on the first session characteristics, determine whether the current session needs to be transferred to high-priority processing.

[0194] In one possible implementation, the dialogue information for each round of conversation includes: a first statement sent by the target user, and a second statement output by the customer service representative in response to the first statement; the determining module 803 is specifically used for:

[0195] Based on the dialogue data, a first dialogue sequence feature and a second dialogue sequence feature are generated; the first dialogue sequence feature includes: the feature expression result corresponding to at least one first statement in the dialogue data, and the second dialogue sequence feature includes: the feature expression result corresponding to at least one second statement in the dialogue data;

[0196] Based on the user behavior data, a user behavior sequence feature is generated; the user behavior sequence feature includes: the feature expression result corresponding to at least one behavior information in the user behavior data;

[0197] Based on the associated object data, associated object features are generated; the associated object features include: feature expression results corresponding to at least one object information in the associated object data;

[0198] Based on the first dialogue sequence features, the second dialogue sequence features, the user behavior sequence features, and the associated object features, a first session feature corresponding to the current session is generated.

[0199] In one possible implementation, the determining module 803 is specifically used for:

[0200] The first session features are processed by a pre-trained prediction model to obtain the probability that the target user expects the current session to be converted to a high-priority processing.

[0201] Based on the probability, determine whether the current session needs to be switched to high-priority processing.

[0202] In one possible implementation, the prediction model is trained in the following manner:

[0203] Obtain training samples, which include: second session features corresponding to the sample session and first annotation data corresponding to the sample session. The first annotation data is used to indicate whether the sample user instructs to switch to high-priority processing during the sample session.

[0204] The second session features are processed by the prediction model to be trained to obtain a first prediction result, which is used to indicate the probability that the sample user expects the sample session to be converted to high priority processing.

[0205] Based on the first prediction result and the first labeled data, the model parameters of the prediction model are updated, and the update objective includes minimizing the difference between the first prediction result and the first labeled data.

[0206] In one possible implementation, the determining module 803 is specifically used for:

[0207] The first session features are processed by a pre-trained prediction model to obtain the probability that the target user expects the current session to be converted to a high-priority processing, and the target user's satisfaction level with the current session;

[0208] Based on the probability and the level of satisfaction, determine whether the current session needs to be transferred to high-priority processing.

[0209] In one possible implementation, the prediction model is trained in the following manner:

[0210] Obtain training samples, which include: second session features corresponding to the sample session, first labeled data corresponding to the sample session, and second labeled data corresponding to the sample session; wherein, the first labeled data is used to indicate whether the sample user instructs to switch to high-priority processing during the sample session, and the second labeled data is used to indicate the sample user's satisfaction level with the sample session;

[0211] The second session features are processed by the prediction model to be trained to obtain a first prediction result and a second prediction result; wherein, the first prediction result is used to indicate the probability that the sample user expects the sample session to be converted to high priority processing, and the second prediction result is used to indicate the sample user's satisfaction with the sample session;

[0212] Based on the first prediction result, the first labeled data, the second prediction result, and the second labeled data, the model parameters of the prediction model are updated. The update objectives include minimizing the difference between the first prediction result and the first labeled data, and minimizing the difference between the second prediction result and the second labeled data.

[0213] In one possible implementation, the second acquisition module 802 is specifically used for:

[0214] Based on the current session, the identifier of the target user and the target time period are determined, wherein the start time of the target time period is before the current session and the end time of the target time period is the current time.

[0215] Based on the target user's identifier and the target time period, the user's behavior information within the target time period is queried from a preset database to obtain the user behavior data; the preset database is used to store historical behavior information of different users.

[0216] The dialogue processing apparatus provided in this embodiment can execute the dialogue processing method provided in any of the above method embodiments. Its implementation principle and technical effect are similar, and will not be described again here.

[0217] Figure 9 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Figure 9 As shown, the electronic device 900 provided in this embodiment includes: a memory 901 and at least one processor 902; the memory 901 stores a computer program, which is configured to be executed by the processor 902 to implement the dialogue processing method provided in any of the above method embodiments. The implementation principle and technical effect are similar, and will not be described again here.

[0218] Optionally, the memory 901 can be either standalone or integrated with the processor 902. When the memory 901 is a device independent of the processor 902, the electronic device 900 also includes a bus 903 for connecting the memory 901 and the processor 902.

[0219] This application also provides a computer-readable storage medium storing a computer program thereon. The computer program is executed by a processor to implement the dialogue processing method provided in any of the foregoing method embodiments. The implementation principle and technical effects are similar, and will not be described in detail here.

[0220] This application provides a computer program product, including a computer program. When the computer program is executed by a processor, it implements the dialogue processing method provided in any of the foregoing method embodiments. The implementation principle and technical effects are similar, and will not be described in detail here.

[0221] This application also provides a chip, including a memory and a processor. The memory stores a computer program, and the processor runs the computer program to implement the dialogue processing method provided in any of the above method embodiments. The implementation principle and technical effect are similar, and will not be repeated here.

[0222] It should be understood that the aforementioned processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. A general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this invention can be directly manifested as being executed by a hardware processor, or executed by a combination of hardware and software modules within the processor.

[0223] The memory may include high-speed RAM, and may also include non-volatile storage (NVM), such as at least one disk storage device, and may also be a USB flash drive, external hard drive, read-only memory, disk or optical disc, etc.

[0224] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of illustration, the buses shown in the accompanying drawings are not limited to a single bus or a single type of bus.

[0225] The aforementioned storage medium can be implemented from any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The storage medium can be any available medium accessible to general-purpose or special-purpose computers.

[0226] An exemplary storage medium is coupled to a processor, enabling the processor to read information from and write information to the storage medium. Alternatively, the storage medium can be an integral part of the processor. Both the processor and the storage medium can reside in an application-specific integrated circuit (ASIC). Alternatively, the processor and storage medium can exist as discrete components in an electronic device.

[0227] Those skilled in the art will understand that all or part of the steps of the above-described method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0228] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

Claims

1. A dialog processing method characterized by, include: Retrieve the dialogue data corresponding to the current session; The dialogue data includes: dialogue information from at least one round of conversation between the target user and customer service; Obtain user behavior data corresponding to the current session; the user behavior data includes: behavioral information of at least one behavior of the target user within a preset time period before the current moment; Obtain the associated object data corresponding to the current session. The associated object data includes: object information of at least one object related to the current session; at least one object related to the current session includes: the product discussed / mentioned in the current session, the order discussed / mentioned in the current session, or the merchant / store represented by customer service; the object information includes: the object's identifier, the object's category, the user satisfaction of the historical sessions associated with the object, and whether the historical sessions associated with the object have been converted to high priority processing; Based on the dialogue data, the user behavior data, and the associated object data, determine whether the current session needs to be converted to high-priority processing; If the current session needs to be transferred to high-priority processing, then a session connection is established between the target user and the high-priority processing flow; The step of determining whether the current session needs to be converted to high-priority processing based on the dialogue data, the user behavior data, and the associated object data includes: Based on the dialogue data, the user behavior data, and the associated object data, a first session feature corresponding to the current session is generated; The first session features are processed by a pre-trained prediction model to obtain the probability that the target user expects the current session to be converted to a high-priority processing, and the target user's satisfaction level with the current session; Based on the probability and the level of satisfaction, determine whether the current session needs to be transferred to high-priority processing.

2. The method of claim 1, wherein, The dialogue information for each round of conversation includes: a first statement sent by the target user, and a second statement output by the customer service representative in response to the first statement; based on the dialogue data, the user behavior data, and the associated object data, a first conversation feature corresponding to the current conversation is generated, including: Based on the dialogue data, a first dialogue sequence feature and a second dialogue sequence feature are generated; the first dialogue sequence feature includes: the feature expression result corresponding to at least one first statement in the dialogue data, and the second dialogue sequence feature includes: the feature expression result corresponding to at least one second statement in the dialogue data; Based on the user behavior data, a user behavior sequence feature is generated; the user behavior sequence feature includes: the feature expression result corresponding to at least one behavior information in the user behavior data; Based on the associated object data, associated object features are generated; the associated object features include: feature expression results corresponding to at least one object information in the associated object data; Based on the first dialogue sequence features, the second dialogue sequence features, the user behavior sequence features, and the associated object features, a first session feature corresponding to the current session is generated.

3. The method according to claim 1, characterized in that, The prediction model was trained in the following manner: Obtain training samples, which include: second session features corresponding to the sample session, first labeled data corresponding to the sample session, and second labeled data corresponding to the sample session; wherein, the first labeled data is used to indicate whether the sample user instructs to switch to high-priority processing during the sample session, and the second labeled data is used to indicate the sample user's satisfaction level with the sample session; The second session features are processed by the prediction model to be trained to obtain a first prediction result and a second prediction result; wherein, the first prediction result is used to indicate the probability that the sample user expects the sample session to be converted to high priority processing, and the second prediction result is used to indicate the sample user's satisfaction with the sample session; Based on the first prediction result, the first labeled data, the second prediction result, and the second labeled data, the model parameters of the prediction model are updated. The update objectives include minimizing the difference between the first prediction result and the first labeled data, and minimizing the difference between the second prediction result and the second labeled data.

4. The method according to any one of claims 1 to 3, characterized in that, Obtaining user behavior data corresponding to the current session includes: Based on the current session, the identifier of the target user and the target time period are determined, wherein the start time of the target time period is before the current session and the end time of the target time period is the current time. Based on the target user's identifier and the target time period, the user's behavior information within the target time period is queried from a preset database to obtain the user behavior data; the preset database is used to store historical behavior information of different users.

5. A dialogue processing device, characterized in that, include: The first acquisition module is used to acquire the dialogue data corresponding to the current session; The dialogue data includes: dialogue information from at least one round of conversation between the target user and customer service; The second acquisition module is used to acquire user behavior data corresponding to the current session; the user behavior data includes: behavior information of at least one behavior of the target user within a preset time period before the current moment; The third acquisition module is used to acquire the associated object data corresponding to the current session. The associated object data includes: object information of at least one object related to the current session; at least one object related to the current session includes: the product discussed / mentioned in the current session, the order discussed / mentioned in the current session, or the merchant / store represented by customer service; the object information includes: the object's identifier, the object's category, the user satisfaction of the historical sessions associated with the object, and whether the historical sessions associated with the object have been converted to high-priority processing; The determination module is used to determine whether the current session needs to be converted to high-priority processing based on the dialogue data, the user behavior data, and the associated object data; The control module is used to establish a session connection between the target user and the high-priority processing flow if the current session needs to be transferred to high-priority processing. The determining module is specifically used to generate a first session feature corresponding to the current session based on the dialogue data, the user behavior data, and the associated object data; process the first session feature through a pre-trained prediction model to obtain the probability that the target user expects to transfer the current session to high priority processing and the target user's satisfaction level with the current session; and determine whether the current session needs to be transferred to high priority processing based on the probability and the satisfaction level.

6. An electronic device, characterized in that, include: Memory and at least one processor; The memory stores a computer program configured to be executed by the at least one processor to implement the method as described in any one of claims 1 to 4.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the method as described in any one of claims 1 to 4.

8. A computer program product, characterized in that, Includes a computer program that, when executed by a processor, implements the method as described in any one of claims 1 to 4.

Images