Intelligent customer service method and device for sustainable learning, storage medium and program product

Abstract

The invention provides an intelligent customer service method and device for sustainable learning, a storage medium and a program product, and relates to the technical field of artificial intelligence and customer service. The method comprises the following steps: acquiring session data between an intelligent customer service and a customer; determining a session type corresponding to the intelligent customer service session data based on a session round and a session end statement corresponding to the session data; determining a session labeling strategy corresponding to the session data based on the session type, and executing the session labeling strategy to obtain a data labeling result corresponding to the session data; the session labeling strategy comprises a first session strategy corresponding to a single-round session type, a second session strategy corresponding to a multi-round session type or a third session strategy corresponding to a man-to-man session type; and updating a preset intelligent customer service model and / or a preset session knowledge base based on the data annotation result and an update operation indicated by the session annotation strategy. According to the invention, the problem of poor interaction experience between the customer and the intelligent customer service can be solved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence and customer service technology, and in particular to a sustainable learning intelligent customer service method, device, storage medium and program product. Background Art

[0002] Compared with manual customer service, intelligent customer service can provide services around the clock and respond to user requests immediately, which is suitable for handling common problems with high frequency. At the same time, the use of intelligent customer service can significantly reduce operating costs and reduce the demand for manual customer service, which is suitable for enterprises to use in high-traffic environments. Therefore, the application of intelligent customer service products is becoming more and more widespread.

[0003] However, existing intelligent customer service systems focus on specific application scenarios, such as simple questions and answers or consultation on specific products. When faced with personalized or core business services, the limitations of intelligent customer service are more significant. The core business requirements are updated and iterated quickly and the business logic is complex. When faced with new problems, they are unable to learn and update their knowledge base in a timely and accurate manner, resulting in a poor interaction experience between customers and intelligent customer service. Summary of the invention

[0004] In view of this, the embodiments of the present invention provide a method, device, storage medium and program product for intelligent customer service capable of continuous learning to eliminate or improve one or more defects existing in the prior art.

[0005] One aspect of the present invention provides a sustainable learning intelligent customer service method, the method comprising the following steps:

[0006] Acquire the conversation data between the intelligent customer service and the customer; the conversation data includes the customer conversation corresponding to the customer and the intelligent customer service response generated by the intelligent customer service based on the preset conversation knowledge base and the preset intelligent customer service model;

[0007] Determine the session type corresponding to the intelligent customer service session data based on the session round and session end statement corresponding to the session data; the session type of the session data includes a single-round session type, a multi-round session type, or a manual session type;

[0008] Determine the session annotation strategy corresponding to the session data based on the session type, and execute the session annotation strategy to obtain the data annotation result corresponding to the session data; wherein the session annotation strategy is used to indicate at least one annotation strategy acting on the intelligent customer service reply and the update operation corresponding to each annotation strategy; the session annotation strategy includes a first session strategy corresponding to a single-round session type, a second session strategy corresponding to a multi-round session type, or a third session strategy corresponding to a manual session type;

[0009] Based on the data annotation results and the update operation indicated by the conversation annotation strategy, the preset intelligent customer service model and / or the preset conversation knowledge base is updated.

[0010] In some embodiments of the present invention, when the session type of the session data is a single-round session type, the session data includes a customer session and an intelligent customer service reply; executing the session annotation strategy to obtain a data annotation result corresponding to the session data includes:

[0011] Sending the session data to the first target device, so that after receiving the session data, the first target device classifies and annotates an intelligent customer service reply based on the first session strategy and a customer session, obtains a data annotation result and returns it;

[0012] Receive data annotation results, which include a response classification label corresponding to an intelligent customer service response.

[0013] In some embodiments of the present invention, updating a preset intelligent customer service model and / or a preset conversation knowledge base based on the data annotation result and the update operation indicated by the conversation annotation strategy includes:

[0014] When the response classification label indicates that the answer exists in the preset conversation knowledge base and the response is wrong, updating the preset intelligent customer service model;

[0015] or,

[0016] When the response classification label indicates that there is a lack of relevant responses in the preset conversation knowledge base, the preset conversation knowledge base and / or the preset intelligent customer service model is updated.

[0017] In some embodiments of the present invention, when the session type of the session data is a multi-round session type, the session data includes at least one customer session and at least one intelligent customer service reply; executing the session annotation strategy to obtain the data annotation result corresponding to the session data includes:

[0018] Sending the session data to the second target device, so that after receiving the session data, the second target device marks the business scenario corresponding to the session data, and marks at least one intelligent customer service reply in sequence based on the second session strategy and at least one customer session, obtains and returns the data marking result;

[0019] Receive data annotation results, which include the response classification label corresponding to each intelligent customer service response and the business scenario label corresponding to the session data.

[0020] In some embodiments of the present invention, updating a preset intelligent customer service model and / or a preset conversation knowledge base based on the data annotation result and the update operation indicated by the conversation annotation strategy includes:

[0021] When the response classification label indicates that the preset intelligent customer service model has an intention recognition error or an entity recognition error, updating the preset intelligent customer service model;

[0022] or,

[0023] When the business scenario label indicates that the business scenario is incorrectly identified, the preset intelligent customer service model is updated;

[0024] or,

[0025] When the business scenario label indicates that the current business scenario is not identified, the preset conversation knowledge base and / or the preset intelligent customer service model are updated.

[0026] In some embodiments of the present invention, when the session type of the session data is a manual session type, the session data includes at least one customer session and at least one intelligent customer service reply; executing the session annotation strategy to obtain the data annotation result corresponding to the session data includes:

[0027] Sending the session data to a third target device, so that after receiving the session data, the third target device annotates the session data based on a third session policy, obtains a data annotation result and returns it;

[0028] Receive data labeling results, where the data labeling results include a manual result label or a manual reason label.

[0029] In some embodiments of the present invention, it is characterized in that updating a preset intelligent customer service model and / or a preset conversation knowledge base based on the data annotation result and the update operation indicated by the conversation annotation strategy includes:

[0030] When the result label of the manual transfer indicates that the manual transfer has failed, the preset intelligent customer service model is updated;

[0031] or,

[0032] When the reason label for transferring to manual service indicates that the intelligent customer service gave an erroneous reply or did not give a reply, the preset intelligent customer service model and the preset conversation knowledge base are updated.

[0033] Another aspect of the present invention provides an intelligent customer service conversation annotation device, including a processor, a memory, and a computer program / instructions stored in the memory, the processor is used to execute the computer program / instructions, and when the computer program / instructions are executed, the device implements the steps of the aforementioned sustainable learning intelligent customer service method.

[0034] Another aspect of the present invention provides a computer-readable storage medium having a computer program / instruction stored thereon, which, when executed by a processor, implements the steps of the aforementioned sustainable learning intelligent customer service method.

[0035] Another aspect of the present invention provides a computer program product, including a computer program / instruction, which, when executed by a processor, implements the steps of the aforementioned continuously learning intelligent customer service method.

[0036] The continuously learning intelligent customer service method, device, storage medium and program product of the present invention can solve the problem of poor interactive experience between customers and intelligent customer service; after annotating and sorting the conversation data through the conversation annotation strategy, the preset conversation knowledge base and the training optimization model are updated in time, and the knowledge base and model are accurately learned and updated, so that the intelligent customer service can continue to learn, improve the accuracy of identifying user intentions and understanding question semantics, and improve the probability of replying correct answers and the service efficiency of intelligent customer service, so as to improve the interactive experience between customers and intelligent customer service; at the same time, by annotating and sorting the conversations between customers and intelligent customer service, by continuously iterating the training model and updating the knowledge base, the intelligent customer service can continue to learn in various core business scenarios, and then apply it in more core business scenarios, so as to realize the use of intelligent customer service in more core business scenarios, thereby improving the service efficiency of customer service.

[0037] Additional advantages, purposes, and features of the present invention will be described in part in the following description, and will become apparent to those skilled in the art after studying the following, or may be learned from the practice of the present invention. The purposes and other advantages of the present invention may be achieved and obtained by the structures specifically indicated in the specification and the accompanying drawings.

[0038] Those skilled in the art will appreciate that the objectives and advantages that can be achieved with the present invention are not limited to the above specific description, and the above and other objectives that can be achieved by the present invention will be more clearly understood from the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] The drawings described herein are used to provide a further understanding of the present invention, constitute a part of the present application, and do not constitute a limitation of the present invention. In the drawings:

[0040] Figure 1 A flowchart of a continuously learning intelligent customer service method provided by one embodiment of the present invention.

[0041] Figure 2 A block diagram of an intelligent customer service device capable of continuous learning provided by another embodiment of the present invention. DETAILED DESCRIPTION

[0042] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the embodiments and the accompanying drawings. Here, the illustrative embodiments of the present invention and their descriptions are used to explain the present invention, but are not intended to limit the present invention.

[0043] It should also be noted that, in order to avoid obscuring the present invention due to unnecessary details, only structures and / or processing steps closely related to the solutions according to the present invention are shown in the accompanying drawings, while other details that are not closely related to the present invention are omitted.

[0044] It should be emphasized that the term “include / comprises” when used herein refers to the presence of features, elements, steps or components, but does not exclude the presence or addition of one or more other features, elements, steps or components.

[0045] It should also be noted that, unless otherwise specified, the term “connection” herein may refer not only to a direct connection but also to an indirect connection involving an intermediate.

[0046] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. In the accompanying drawings, the same reference numerals represent the same or similar components, or the same or similar steps.

[0047] The sustainable learning intelligent customer service method provided by this application is introduced in detail below.

[0048] Optionally, the execution subject of each embodiment in the present application may be an electronic device, which may be a terminal such as a computer, a mobile phone, a tablet computer, a personal computer, or a server. This embodiment does not limit the implementation method of the electronic device.

[0049] like Figure 1 As shown, an embodiment of the present application provides a sustainable learning intelligent customer service method, which at least includes steps S101 to S104:

[0050] Step S101, obtaining conversation data between the intelligent customer service and the customer; the conversation data includes the customer conversation corresponding to the customer and the intelligent customer service reply generated by the intelligent customer service based on a preset conversation knowledge base and a preset intelligent customer service model.

[0051] Among them, intelligent customer service refers to a system that uses artificial intelligence technology (such as natural language processing, machine learning, etc.) to automatically respond to customer inquiries; session data refers to the interaction records between customers and intelligent customer service, including all questions and answers and communication content.

[0052] In some embodiments of the present invention, the session data includes the session records between the current user and the intelligent customer service obtained after the conversation between the current customer and the intelligent customer service ends; or, the session data includes the session records between all customers and the intelligent customer service within a preset time period, such as the session records between all customers and the intelligent customer service within 5 hours, 10 hours or 24 hours.

[0053] The preset conversation knowledge base refers to a pre-stored database of common questions and their answers, such as a company's product information, solutions to common questions, etc. The preset intelligent customer service model refers to a pre-trained model that can understand and generate natural language responses, and is used to generate responses for customer conversations. Intelligent customer service automatically generates responses based on customer input and information in the preset conversation knowledge base.

[0054] Step S102: determining the session type corresponding to the intelligent customer service session data based on the session round and the session end statement corresponding to the session data.

[0055] Among them, the conversation turn refers to the number of interactions between the customer and the intelligent customer service in a conversation; the conversation end statement refers to the last sentence recorded in the conversation data.

[0056] In some implementations of the present invention, the session type of the session data includes a single-round session type, a multi-round session type, or a manual session type.

[0057] Among them, a single-round conversation type refers to a conversation type in which a customer asks a question and an intelligent customer service replies, that is, a conversation type with a conversation round of 1. After the customer asks a question, the intelligent customer service gives a direct and complete reply, which is determined to be a single-round conversation, forming a complete interaction unit.

[0058] Multi-round conversation type refers to the conversation type with multiple interactions between customers and intelligent customer service, that is, the conversation type with more than 1 conversation round. Each round may be further based on the content of the previous round of conversation; multi-round conversations are also regarded as a whole. If there are clear interactions of "entering multi-round" and "ending multi-round", they are judged as valid multi-round conversations.

[0059] The transfer to human operator conversation type refers to the conversation in which the customer requests to be transferred to human operator, that is, the conversation turn is greater than or equal to 1, and the customer's conversation ending statement is a conversation type that requests transfer to human operator. The transfer to human operator process is also a holistic judgment. Only when the "transfer to human operator" operation is completed and the "agent response ends" state is reached, can it be considered a complete transfer to human operator process.

[0060] Step S103: determining a session annotation strategy corresponding to the session data based on the session type, and executing the session annotation strategy to obtain a data annotation result corresponding to the session data.

[0061] The session annotation strategy is used to indicate at least one annotation strategy acting on the intelligent customer service response and the update operation corresponding to each annotation strategy.

[0062] Since the core of intelligent customer service is accurate data support, data labeling is a key link to ensure the efficient and accurate operation of intelligent customer service. Data labeling provides training data for machine learning models by carefully classifying, labeling and interpreting massive amounts of data, enabling them to understand real-world scenarios more accurately, thereby promoting the continuous optimization and upgrading of intelligent customer service and better serving customers. Given that the application scenarios of data labeling are extremely wide and business needs vary, most intelligent products are currently used in scenarios with relatively simple business rules. When there is a large demand for data labeling, they tend to outsource data to third-party companies for specialized labeling processing. However, when faced with personalized or core business services, the limitations of intelligent customer service are also quite significant.

[0063] Based on this, in order to broaden the application scope of intelligent customer service and deepen its effectiveness in various businesses, in some implementations of the present invention, the conversation data is annotated and sorted through the conversation annotation strategy to update the preset conversation knowledge base, or provided to the model as training data for training, accurately learning and updating its own knowledge base and model, so that intelligent customer service can continuously improve the accuracy of identifying user intentions and understanding question semantics, as well as improve the probability of replying correct answers and the service efficiency of intelligent customer service, thereby improving the interactive experience between customers and intelligent customer service.

[0064] In some embodiments of the present invention, different session types correspond to different session annotation strategies, including a first session strategy corresponding to a single-round session type, a second session strategy corresponding to a multi-round session type, or a third preset strategy corresponding to a manual session type.

[0065] In some embodiments of the present invention, since the annotation strategies corresponding to session data of different session types are different, the characteristics and annotation requirements of the sessions are different. Based on this, the annotation devices corresponding to session data of different session types are different. By using special annotation devices to annotate session data of different session types, the efficiency of session data annotation can be improved, and at the same time, the accuracy of data annotation is ensured.

[0066] At this time, the session marking strategy includes determining a target marking device according to a session type of the session data, and sending the session data to the target marking device, so that the target marking device marks the session data according to the corresponding marking strategy after receiving the session data.

[0067] For example: take the example that the session data includes session data a corresponding to customer A and session data b corresponding to customer B; the session type corresponding to session data a is a single-round session type, and the session type corresponding to session data b is a multi-round session type; the session data a is sent to the first target device, and after receiving the session data a, the first target device marks the session data a according to the first marking strategy; the session data b is sent to the second target device, and after receiving the session data b, the second target device marks the session data b according to the second marking strategy.

[0068] In some embodiments of the present invention, the session annotation strategy includes determining a target annotation device for processing current session data in at least one annotation device, and sending the annotation strategy and session data corresponding to the current session type to the target annotation device, so that the target annotation device annotates the session data according to the annotation strategy after receiving the annotation strategy and session data.

[0069] For example: taking the session data including session data c corresponding to customer C as an example, determine that the third annotation device processes the session data c; when the session type of session data c is a single-round session type, send the session data c and the first annotation strategy to the third annotation device, and the third annotation device annotates the session data c according to the first annotation strategy after receiving it; when the session type of session data c is a multi-round session type, send the session data c and the second annotation strategy to the third annotation device, and the third annotation device annotates the session data c according to the second annotation strategy after receiving it.

[0070] In addition, in actual implementation, different annotators may be assigned to session data of different session types to further improve the efficiency and accuracy of annotation.

[0071] The first conversation strategy refers to a strategy for intelligent customer service conversation classification labeling of conversation data of a single-round conversation type. The labeled reply classification label is used to indicate that the intelligent customer service reply is correct, the answer exists in the preset conversation knowledge base and the reply is wrong, or the preset conversation knowledge base lacks relevant replies.

[0072] Among them, the answer exists in the preset conversation knowledge base and the response is wrong, which includes the situation that the intelligent customer service fails to recognize the customer's intention and does not give a response, or the situation that the intelligent customer service fails to recognize the customer's intention and gives an incorrect response.

[0073] The lack of relevant responses in the preset conversation knowledge base includes the case where the preset conversation knowledge base lacks corresponding knowledge and the intelligent customer service does not respond, or the case where the preset conversation knowledge base lacks corresponding knowledge and the intelligent customer service responds incorrectly.

[0074] In some embodiments of the present invention, the reply classification label is also used to indicate that the intelligent customer service has not replied, the intelligent customer service has recognized greetings, polite expressions, etc., the intelligent customer service has recognized business-irrelevant issues, etc.

[0075] In addition, when the conversation data includes a customer's evaluation statement on the intelligent customer service reply (for example, helpful or unhelpful), the reply classification label is also used to indicate the customer's evaluation of the intelligent customer service reply.

[0076] Specifically, when the session type of the session data is a single-round session type, the session data includes a customer session and an intelligent customer service reply; executing the session annotation strategy to obtain a data annotation result corresponding to the session data, including: sending the session data to a first target device, so that after receiving the session data, the first target device classifies and annotates an intelligent customer service reply based on the first session strategy and a customer session, obtains and returns the data annotation result; receiving the data annotation result, the data annotation result includes a reply classification label corresponding to the intelligent customer service reply.

[0077] The second session strategy refers to a strategy for annotating session data for multi-round session types, including but not limited to business scenario annotations corresponding to session data, intelligent customer service intent recognition annotations, and intelligent customer service entity recognition annotations.

[0078] Specifically, when the session type of the session data is a multi-round session type, the session data includes at least one customer session and at least one intelligent customer service reply; executing the session annotation strategy to obtain the data annotation result corresponding to the session data, including: sending the session data to the second target device, so that the second target device annotates the business scenario corresponding to the session data after receiving the session data, and annotates the at least one intelligent customer service reply in turn based on the second session strategy and at least one customer session to obtain and return the data annotation result; receiving the data annotation result, the data annotation result includes the reply classification label corresponding to each intelligent customer service reply and the business scenario label corresponding to the session data.

[0079] The third preset strategy refers to a strategy for marking session data for the manual session type, including whether the marking corresponding to the manual session is identified and whether the manual session process implements the corresponding marking.

[0080] Specifically, when the session type of the session data is a conversion to manual session type, the session data includes at least one customer session and at least one intelligent customer service response; executing the session annotation strategy to obtain the data annotation result corresponding to the session data, including: sending the session data to a third target device, so that the third target device annotates the session data based on the third session strategy after receiving the session data, obtains the data annotation result and returns it; receiving the data annotation result, the data annotation result includes a conversion to manual result label or a conversion to manual reason label.

[0081] Step S104: based on the data annotation result and the update operation indicated by the conversation annotation strategy, update the preset intelligent customer service model and / or the preset conversation knowledge base.

[0082] In some embodiments of the present invention, the update operations corresponding to different data annotation results are the same or different, wherein the update operation includes a preset intelligent customer service model update operation and a preset conversation knowledge base update operation.

[0083] When the data type of the session data is a single-round session type, the data annotation results and corresponding update operations include the following cases:

[0084] First, when the data annotation result includes that the answer exists in the preset conversation knowledge base and the response is wrong, the update operation includes updating the preset intelligent customer service model.

[0085] For example: if the answer the customer needs exists in the preset conversation knowledge base, but the intelligent customer service fails to recognize the customer's true intention and does not give a response, the preset intelligent customer service model is updated; or, if the answer the customer needs exists in the preset conversation knowledge base, but the intelligent customer service fails to recognize the customer's true intention and replies with a wrong answer, the preset intelligent customer service model is updated.

[0086] The second type is that when the data annotation results include the lack of relevant responses in the preset conversation knowledge base, the update operation includes updating the preset conversation knowledge base and the preset intelligent customer service model. For example: when the preset conversation knowledge base lacks the response corresponding to the customer's question, resulting in the robot being unable to reply with the correct answer, the preset conversation knowledge base and the preset intelligent customer service model are updated; or when the preset conversation knowledge base lacks the response corresponding to the customer's question, the intelligent customer service fails to recognize the customer's true intention and responds incorrectly, the preset conversation knowledge base and the preset intelligent customer service model are updated.

[0087] Specifically, based on the update operation indicated by the data annotation results and the conversation annotation strategy, the preset intelligent customer service model and / or the preset conversation knowledge base are updated, including: when the reply classification label indicates that there is an answer in the preset conversation knowledge base and the reply is wrong, the preset intelligent customer service model is updated; or, when the reply classification label indicates that there is a lack of relevant replies in the preset conversation knowledge base, the preset conversation knowledge base and the preset intelligent customer service model are updated.

[0088] In addition, in actual implementation, the reply classification label can also be used to indicate that the intelligent customer service has not replied, the intelligent customer service has recognized greetings, polite expressions, etc., the intelligent customer service has recognized business-irrelevant issues, or the user's service evaluation of the intelligent customer service, etc. Accordingly, when the intelligent customer service has not replied, that is, the intelligent customer service has not recognized the customer's true intention or has not replied after recognizing the intention, the preset intelligent customer service model is updated; when the intelligent customer service recognizes greetings, polite expressions, etc., conversation-irrelevant or business-irrelevant issues, the preset intelligent customer service model and the preset conversation knowledge base are updated, or the interaction of the intelligent customer service is optimized and the corresponding reply is configured; when the conversation data includes the user's service evaluation of the intelligent customer service, the preset intelligent customer service model and the preset conversation knowledge base are updated.

[0089] In some embodiments of the present invention, the labeled session data is used to update a preset session knowledge base and a preset intelligent customer service model; based on this, in order to ensure that the question semantics of the updated knowledge base are non-repetitive and the reply answers are accurate, as well as the accuracy of the replies of the updated intelligent customer service model, before using the labeled session data for the update operation, it is also necessary to manually review the corresponding labeled session data to analyze whether the standard questions, answers, and similar questions are correct, to ensure that the correct training data and direction are provided for the model.

[0090] If the review result indicates that there are errors in the data annotation, instruct the relevant annotation equipment or annotation personnel to re-annotate; if the review result indicates that the data annotation is correct, use the annotated session data to update the preset session knowledge base; use the annotated session data as training data to train and tune the preset intelligent customer service model, and at the same time, use the original session data as a verification set to verify the trained model to ensure the correctness of the output of the trained model.

[0091] The version update of the preset conversation knowledge base and the model iteration of the preset intelligent customer service model are maintained synchronously. After the update operation is completed, the updated preset conversation knowledge base and preset intelligent customer service model are released online as new versions.

[0092] When the data type of the conversation data is a multi-round conversation type, the data annotation result includes the annotation corresponding to each reply of the intelligent customer service in the conversation data. The data annotation result and the corresponding update operation include the following situations:

[0093] The first is to update the preset intelligent customer service model when the data annotation results include intelligent customer service intention recognition errors or entity recognition errors. For example, when the real intention of the user session, keywords / keywords and other entity recognition errors are made, the preset intelligent customer service model is updated.

[0094] The second method is to update the preset conversation knowledge base and the preset intelligent customer service model when the data annotation results include failure to identify the current business scenario. For example, if the business classification corresponding to the current business scenario is not found in the business classification, the corresponding business classification is added to the preset conversation knowledge base, and the preset conversation knowledge base and the preset intelligent customer service model are updated; or if the corresponding business classification exists in the preset indication library and the intelligent customer service intention business scenario is incorrectly identified, the preset intelligent customer service model is updated.

[0095] Specifically, based on the update operation indicated by the data annotation results and the session annotation strategy, the preset intelligent customer service model and / or the preset session knowledge base are updated, including: when the reply classification label indicates an intelligent customer service intention recognition error or an entity recognition error, the preset intelligent customer service model is updated; or, when the business scenario label indicates a business scenario recognition error, the preset intelligent customer service model is updated; or, when the business scenario label indicates that the current business scenario is not recognized, the preset session knowledge base and / or the preset intelligent customer service model is updated.

[0096] In addition, in actual implementation, in multiple rounds of conversations, in order to further communicate with customers and understand their true intentions, intelligent customer service usually needs to reply with some guiding sentences to guide customers to provide correct information to understand their true intentions. Based on this, the reply classification label is also used to indicate the guiding sentences replied by the intelligent customer service.

[0097] When the data type of the session data is a manual session type, the data annotation result includes whether the annotation result corresponding to the manual transition is identified and whether the manual transition process achieves the corresponding annotation result. At this time, the data annotation result and the corresponding update operation include the following situations:

[0098] The first is to update the preset intelligent customer service model when the data annotation results include a failure to transfer to manual service. For example: the customer applies for manual service, but the intelligent customer service fails to recognize the customer's intention; the customer applies for manual service, but the intelligent customer service recognizes the customer's intention but does not assign a manual seat, etc.

[0099] The second type is to update the preset conversation knowledge base and the preset intelligent customer service model when the data annotation results include a valid question raised by the customer and the question is transferred to manual service because the intelligent customer service gave an incorrect answer or did not reply. For example, a customer initiated a business consultation, and the intelligent customer service gave an incorrect answer or did not reply, so manual service was requested.

[0100] Specifically, based on the update operation indicated by the data annotation results and the conversation annotation strategy, the preset intelligent customer service model and / or the preset conversation knowledge base are updated, including: when the transfer to manual result label indicates that the transfer to manual service failed, the preset intelligent customer service model is updated; or, when the transfer to manual reason label indicates that the intelligent customer service responded incorrectly or did not respond, the preset intelligent customer service model and the preset conversation knowledge base are updated.

[0101] In addition, in the actual implementation, the scenario of customer transfer to manual service also includes the situation that the customer asks an invalid question, and the intelligent customer service gives an incorrect reply or does not reply, so the customer is transferred to manual service; or, in multiple rounds of conversations, the customer's question is unclear, and after the intelligent customer service replies with a guiding statement, the customer applies for manual service; or, the customer directly applies for manual service. At this time, subsequent update operations also include optimizing the interaction of the intelligent customer service and configuring the corresponding replies.

[0102] In summary, the continuously learning intelligent customer service method provided by the embodiment of the present invention obtains the session data between the intelligent customer service and the customer; the session data includes the customer session corresponding to the customer and the intelligent customer service reply generated by the intelligent customer service based on the preset session knowledge base and the preset intelligent customer service model; based on the session round and session end statement corresponding to the session data, the session type corresponding to the intelligent customer service session data is determined; the session type of the session data includes a single-round session type, a multi-round session type or a manual session type; based on the session type, the session annotation strategy corresponding to the session data is determined, and the session annotation strategy is executed to obtain the data annotation result corresponding to the session data; wherein the session annotation strategy is used to indicate at least one annotation strategy acting on the intelligent customer service reply and the update operation corresponding to each annotation strategy; the session annotation strategy includes a first session strategy corresponding to the single-round session type, a second session strategy corresponding to the multi-round session type or a manual session type type; based on the data annotation results and the update operation indicated by the session annotation strategy, the preset intelligent customer service model and / or the preset session knowledge base is updated; the problem of poor interaction experience between customers and intelligent customer service can be solved; after the session data is annotated and sorted through the session annotation strategy, the preset session knowledge base and the training optimization model are updated in time, and the knowledge base and model are accurately learned and updated, so that the intelligent customer service can continue to learn, improve the accuracy of identifying user intentions and understanding question semantics, and improve the probability of replying correct answers and the service efficiency of intelligent customer service, so as to improve the interaction experience between customers and intelligent customer service; at the same time, through the annotation and sorting of the conversations between customers and intelligent customer service, through continuous iteration of training models and updating of knowledge bases, intelligent customer service can continue to learn in various core business scenarios, and then apply them in more core business scenarios, so as to realize the use of intelligent customer service in more core business scenarios, thereby improving the service efficiency of customer service.

[0103] Corresponding to the above method, the present invention also provides a continuously learning intelligent customer service device, the continuously learning intelligent customer service device includes a computer device, such as Figure 2 As shown, the computer device includes a processor 210 and a memory 220, wherein the memory 210 stores computer instructions, and the processor 210 is used to execute the computer instructions stored in the memory 220. When the computer instructions are executed by the processor, the continuously learning intelligent customer service device implements the steps of the continuously learning intelligent customer service method as described above.

[0104] Another aspect of the present invention provides a computer-readable storage medium having a computer program / instruction stored thereon, which, when executed by a processor, implements the steps of the aforementioned sustainable learning intelligent customer service method.

[0105] Another aspect of the present invention provides a computer program product, including a computer program / instruction, which, when executed by a processor, implements the steps of the aforementioned continuously learning intelligent customer service method.

[0106] It should be understood by those skilled in the art that the exemplary components, systems and methods described in conjunction with the embodiments disclosed herein can be implemented in hardware, software or a combination of the two. Whether it is performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention. When implemented in hardware, it can be, for example, an electronic circuit, an application specific integrated circuit (ASIC), appropriate firmware, a plug-in, a function card, etc. When implemented in software, the elements of the present invention are programs or code segments used to perform the required tasks. The program or code segment can be stored in a machine-readable medium, or transmitted on a transmission medium or a communication link via a data signal carried in a carrier.

[0107] It should be clear that the present invention is not limited to the specific configuration and processing described above and shown in the figures. For the sake of simplicity, a detailed description of the known method is omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of the present invention is not limited to the specific steps described and shown, and those skilled in the art can make various changes, modifications and additions, or change the order between the steps after understanding the spirit of the present invention.

[0108] In the present invention, features described and / or illustrated for one embodiment may be used in the same or similar manner in one or more other embodiments, and / or combined with features of other embodiments or replace features of other embodiments.

[0109] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the embodiments of the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A sustainable learning intelligent customer service method, characterized in that: The method comprises the following steps: Acquire conversation data between the intelligent customer service and the customer; the conversation data includes the customer conversation corresponding to the customer and the intelligent customer service reply generated by the intelligent customer service based on a preset conversation knowledge base and a preset intelligent customer service model; Determine the session type corresponding to the intelligent customer service session data based on the session round and the session end statement corresponding to the session data; the session type of the session data includes a single-round session type, a multi-round session type, or a manual session type; Determine a session annotation strategy corresponding to the session data based on the session type, and execute the session annotation strategy to obtain a data annotation result corresponding to the session data; wherein the session annotation strategy is used to indicate at least one annotation strategy acting on the intelligent customer service reply and an update operation corresponding to each annotation strategy; the session annotation strategy includes a first session strategy corresponding to the single-round session type, a second session strategy corresponding to the multi-round session type, or a third session strategy corresponding to the manual session type; Based on the data annotation result and the update operation indicated by the conversation annotation strategy, the preset intelligent customer service model and / or the preset conversation knowledge base is updated.

2. The method according to claim 1, characterized in that When the session type of the session data is the single-round session type, the session data includes a customer session and an intelligent customer service reply; The executing the session annotation strategy to obtain a data annotation result corresponding to the session data includes: Sending the session data to a first target device, so that after receiving the session data, the first target device classifies and annotates the intelligent customer service reply based on the first session strategy and the customer session, obtains the data annotation result and returns it; The data annotation result is received, where the data annotation result includes a reply classification label corresponding to the intelligent customer service reply.

3. The method according to claim 2, characterized in that The updating operation based on the data annotation result and the session annotation strategy indication, updating the preset intelligent customer service model and / or the preset session knowledge base, includes: When the reply classification tag indicates that there is an answer in the preset conversation knowledge base and the reply is wrong, updating the preset intelligent customer service model; or, When the reply classification tag indicates that there is a lack of relevant replies in the preset conversation knowledge base, the preset conversation knowledge base and / or the preset intelligent customer service model is updated.

4. The method according to claim 1, characterized in that: In the case where the session type of the session data is the multi-round session type, the session data includes at least one customer session and at least one intelligent customer service response; The executing the session annotation strategy to obtain a data annotation result corresponding to the session data includes: Sending the session data to a second target device, so that after receiving the session data, the second target device marks the business scenario corresponding to the session data, and marks the at least one intelligent customer service reply in sequence based on the second session strategy and the at least one customer session, obtains the data marking result and returns it; The data annotation result is received, where the data annotation result includes a reply classification label corresponding to each intelligent customer service reply and a business scenario label corresponding to the session data.

5. The method according to claim 4, characterized in that The updating operation based on the data annotation result and the session annotation strategy indication, updating the preset intelligent customer service model and / or the preset session knowledge base, includes: When the reply classification label indicates that the preset intelligent customer service model has an intention recognition error or an entity recognition error, updating the preset intelligent customer service model; or, When the business scenario label indicates a business scenario identification error, updating the preset intelligent customer service model; or, When the business scenario label indicates that the current business scenario is not identified, the preset conversation knowledge base and / or the preset intelligent customer service model is updated.

6. The method according to claim 1, characterized in that In the case where the session type of the session data is the converted manual session type, the session data includes at least one customer session and at least one intelligent customer service response; The executing the session annotation strategy to obtain a data annotation result corresponding to the session data includes: Sending the session data to a third target device, so that after receiving the session data, the third target device annotates the session data based on the third session policy, obtains the data annotation result and returns it; The data labeling result is received, where the data labeling result includes a manual result label or a manual reason label.

7. The method according to claim 6, characterized in that The updating operation based on the data annotation result and the session annotation strategy indication, updating the preset intelligent customer service model and / or the preset session knowledge base, includes: When the transfer result label indicates that the transfer to manual service has failed, updating the preset intelligent customer service model; or, When the transfer reason tag indicates that the intelligent customer service has responded incorrectly or has not responded, the preset intelligent customer service model and the preset conversation knowledge base are updated.

8. An intelligent customer service conversation annotation device, comprising a processor, a memory, and a computer program / instruction stored in the memory, characterized in that: The processor is used to execute the computer program / instructions. When the computer program / instructions are executed, the device implements the steps of the method according to any one of claims 1 to 7.

9. A computer-readable storage medium having a computer program / instruction stored thereon, characterized in that: When the computer program / instructions are executed by a processor, the steps of the method as claimed in any one of claims 1 to 7 are implemented.

10. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instructions are executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images