Intelligent customer service method, device, equipment and medium

Abstract

The present invention provides an intelligent customer service method, device, equipment and medium, the intelligent customer service method comprising: determining a target product database according to an inquiry object, obtaining product data and product service data from the target product database; incrementally updating a local knowledge base according to the product data and product service data, and obtaining a first entity set before the incremental update and a second entity set after the incremental update, wherein the entities include entity relationships; predicting according to the inquiry object using a graph neural network to obtain a prediction result, wherein the prediction result indicates the optimal correlation between the inquiry object and the product data and product service data; generating a customer service reply result from the prediction result and sending it to the client, wherein the graph neural network is obtained through training. The beneficial effects of the present invention are: improving the accuracy of customer service replies based on real-time data.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence and online customer service technology, and in particular to an intelligent customer service method, device, equipment and medium. Background Art

[0002] Currently, customer service is either manual or using artificial intelligence. First of all, most manual methods can only deal with a single product, such as automobile after-sales service. Although artificial intelligence methods can provide customers with faster and more convenient customer service based on user portraits and product information.

[0003] However, with the rapid update of products such as auto parts and auto repair services, the current customer service methods can only respond based on existing databases or service data provided by manufacturers, lacking real-time dynamic customer service and personalized services. Summary of the invention

[0004] The main purpose of the embodiments of the present invention is to propose an intelligent customer service method, device, equipment and medium, which improves the accuracy of customer service replies based on real-time data and provides personalized replies.

[0005] One aspect of the present invention provides an intelligent customer service method, comprising:

[0006] According to the customer service inquiry request of the user client, the inquiry object is obtained, wherein the inquiry object includes the product identification and inquiry information;

[0007] Determine a target product database according to the query object, and obtain product data and product service data from the target product database, wherein the product service data includes at least one of product service, user comments, and service comments;

[0008] Incrementally updating the local knowledge base according to the product data and the product service data, and obtaining a first entity set before the incremental update and a second entity set after the incremental update, wherein the entity set includes entities and entity relationships, wherein the entity relationship is used to represent at least one association relationship between the product data and product services, user reviews, and service reviews;

[0009] According to the query object, the second entity set is predicted using a graph neural network to obtain a prediction result, wherein the prediction result represents an optimal correlation relationship between the query object and the product data and the product service data;

[0010] Generate a customer service response result from the prediction result and send it to the client;

[0011] The predictions of the graph neural network include:

[0012] Determine a dynamic knowledge base according to the local knowledge base, the first entity set and the second entity set, wherein the dynamic knowledge base includes an entity set, an entity relationship set, and a time series of incremental updates of the first entity set;

[0013] Determine entity events according to the time series of incremental updates of the first entity set, entities and entity relationships, obtain an entity event set of the dynamic knowledge base, and obtain an embedding vector according to the entity set, the entity relationship set, the time series of incremental updates of the first entity set and the entity event set;

[0014] A graph neural network-based time series encoder is used to take the embedding vector as input to encode and learn the interactive features of entities, entity relationships, and time series.

[0015] A time series decoder based on graph neural network is used to aggregate the neighborhood features of the interaction features to obtain the time dimension features and similarity of the embedded vector;

[0016] The optimal entity relationship of the query object is predicted in the dynamic knowledge base according to the time dimension characteristics and similarity.

[0017] According to the intelligent customer service method, the target product database is determined according to the inquiry object, and product data and product service data are obtained from the target product database, including:

[0018] Determine the service platform of the product according to the product identification, obtain the interface configuration information of the service platform, pull the product data and product service data from the service platform according to the interface configuration information, and record the pulling time each time data is pulled.

[0019] According to the intelligent customer service method, the local knowledge base is incrementally updated according to the product data and the product service data, and a first entity set before the incremental update and a second entity set after the incremental update are obtained, including:

[0020] Obtaining an entity set and an entity relationship set of a first entity set before incremental update, and constructing a first triple set and a first space vector according to the entity set and the entity relationship set;

[0021] Selecting a second triplet similar to the first entity set according to the entity set and the entity relationship set of the second entity set;

[0022] Determine the spatial vectors of entities added to the first entity set and the spatial positions of entity relationships by training according to the first triple set, the second triple and the first entity set;

[0023] A second spatial vector of the second triple is determined according to the spatial vector of the added entity and the spatial position of the entity relationship.

[0024] According to the intelligent customer service method, determining a dynamic knowledge base according to the local knowledge base, the first entity set and the second entity set includes:

[0025] Obtain a third entity set consisting of the first entity set and the second entity set, an entity relationship set included in the third entity set, and obtain a time set for each update of the local knowledge base, and determine the dynamic knowledge base based on the third entity set, the relationship set of the third entity set and the time set.

[0026] According to the intelligent customer service method, a time series encoder based on a graph neural network is used to take an embedding vector as input to encode and learn the interactive features of entities, entity relationships and time series, including:

[0027] Determine, according to the time set, events obtained through corresponding entity relationships between any two associated entities in the third entity set, and then obtain an event set corresponding to the third entity set, wherein the events include generation time, network location, comment object, comment method, and subsequent result of the comment of the service data;

[0028] Determine an embedding vector according to the third entity set, the relationship set of the third entity set, the event set, and the time set, wherein the embedding vector is a four-tuple;

[0029] The encoder of the graph convolutional neural network is used to encode and learn the quadruple to obtain the interactive features of entities, entity relationships, and time series.

[0030] According to the intelligent customer service method, a time series decoder based on a graph neural network is used to aggregate the neighborhood features of the interaction features to obtain the time dimension features and similarity of the embedded vector, including:

[0031] Obtaining a central node of the interactive feature, obtaining a neighborhood feature through aggregating fields of the central node, and obtaining a multidimensional feature of the quadruple according to the neighborhood feature, wherein the multidimensional feature at least includes a time dimension feature;

[0032] The multidimensional features of the quadruple are learned by using a convolutional neural network with a boundary loss training method to obtain the similarity of the multidimensional features of the quadruple.

[0033] According to the intelligent customer service method, generating a customer service response result from the prediction result and sending it to the client includes:

[0034] The keywords of the prediction results are extracted, and the keywords are reassembled using NLP and a preset corpus to obtain the reply result.

[0035] Another aspect of an embodiment of the present invention provides an intelligent customer service device, including:

[0036] The first module is used to obtain an inquiry object according to a customer service inquiry request from a user client, wherein the inquiry object includes a product identifier and inquiry information;

[0037] The second module is used to determine a target product database according to the query object, and obtain product data and product service data from the target product database, wherein the product service data includes at least one of product service, user comments, and service comments;

[0038] A third module is used to incrementally update the local knowledge base according to the product data and the product service data, and obtain a first entity set before the incremental update and a second entity set after the incremental update, wherein the entity set includes entities and entity relationships, wherein the entity relationship is used to characterize at least one association relationship between the product data and the product service, the user review, and the service review;

[0039] A fourth module is used to predict the second entity set using a graph neural network according to the query object to obtain a prediction result, wherein the prediction result represents the optimal correlation between the query object and the product data and the product service data;

[0040] The fifth module is used to generate a customer service reply result based on the prediction result and send it to the client;

[0041] The prediction of the graph neural network includes the sixth module, the seventh module, the eighth module, the ninth module and the tenth module;

[0042] The sixth module is used to determine a dynamic knowledge base according to the local knowledge base, the first entity set and the second entity set, wherein the dynamic knowledge base includes an entity set, an entity relationship set, and a time series of incremental updates of the first entity set;

[0043] The seventh module is used to determine entity events according to the time series of the incremental update of the first entity set, entities and entity relationships, obtain the entity event set of the dynamic knowledge base, and obtain the embedding vector according to the entity set, the entity relationship set, the time series of the incremental update of the first entity set and the entity event set;

[0044] The eighth module is used to use a time series encoder based on a graph neural network to take an embedding vector as input, encode and learn the interactive features of entities, entity relationships and time series;

[0045] The ninth module uses a time series decoder based on a graph neural network to aggregate neighborhood features of the interaction features to obtain the time dimension features and similarity of the embedded vector;

[0046] The tenth module is used to predict the optimal entity relationship of the query object in the dynamic knowledge base according to the time dimension characteristics and similarity.

[0047] Another aspect of an embodiment of the present invention provides an electronic device, including a processor and a memory;

[0048] The memory is used to store programs;

[0049] The processor executes the program to implement the method described above.

[0050] The embodiment of the present invention also discloses a computer program product or a computer program, which includes a computer instruction stored in a computer-readable storage medium. A processor of a computer device can read the computer instruction from the computer-readable storage medium, and the processor executes the computer instruction, so that the computer device executes the method described above.

[0051] The beneficial effects of the present invention are: according to user inquiry requests, the service data and product data of multiple manufacturers are incrementally updated to build a user personalized dynamic knowledge base; a graph convolutional neural network is used to predict entities, entity relationships, time and time dimensions of the user personalized dynamic knowledge base, and provide customer service and real-time services that meet user personalization. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] The above and / or additional aspects and advantages of the present invention will become apparent and easily understood from the description of the embodiments in conjunction with the following drawings, in which:

[0053] Figure 1 It is a schematic diagram of the intelligent customer service process of an embodiment of the present invention.

[0054] Figure 2 It is a schematic diagram of the incremental update process of the local knowledge base according to an embodiment of the present invention.

[0055] Figure 3 It is a schematic diagram of the multi-dimensional relationship prediction process of the dynamic knowledge base in an embodiment of the present invention.

[0056] Figure 4 It is a schematic diagram of the time dimension feature and similarity prediction process of an embodiment of the present invention.

[0057] Figure 5 Schematic diagram of an intelligent customer service device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0058] The embodiments of the present invention are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements with the same or similar functions. In the subsequent description, the use of suffixes such as "module", "component" or "unit" used to represent elements is only for the purpose of facilitating the description of the present invention, and has no specific meaning in itself. Therefore, "module", "component" or "unit" can be used in a mixed manner. "First", "second" and the like are only used for the purpose of distinguishing technical features, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of indicated technical features or implicitly indicating the order of the indicated technical features. In this subsequent description, the continuous numbering of the method steps is for the convenience of review and understanding. In combination with the overall technical solution of the present invention and the logical relationship between the various steps, adjusting the implementation order between the steps will not affect the technical effect achieved by the technical solution of the present invention. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and cannot be understood as limiting the present invention.

[0059] refer to Figure 1 , Figure 1 : is a schematic diagram of the intelligent customer service process of an embodiment of the present invention. It includes but is not limited to steps S100 to S900, wherein steps S100 to S500 are the reply statement generation process of the intelligent customer service service:

[0060] S100, obtaining an inquiry object according to a customer service inquiry request from a user client, wherein the inquiry object includes a product identifier and inquiry information.

[0061] In some embodiments, the query object is a user's query request, for example, "I want to know the most cost-effective car tires recently, and my car is A car all-season tires." Therefore, A car, all-season tire product identification, and the query information is the most cost-effective, and it also has a time dimension requirement of "recently".

[0062] S200, determining a target product database according to an inquiry object, and acquiring product data and product service data from the target product database, wherein the product service data includes at least one of product service, user comments, and service comments.

[0063] In some embodiments, the service platform of the product is determined based on the product identification, the interface configuration information of the service platform is obtained, the product data and product service data are pulled from the service platform based on the interface configuration information, and the pulling time is recorded each time data is pulled.

[0064] It is understandable that according to the above example, it is necessary to pull the product services, user reviews, and service reviews of car model A, and all-season tires need to be matched according to the user's car model and personalized needs. The product services, user reviews, and service reviews of one or more all-season tires are pulled by calling the corresponding API interface. It is understandable that the user's personalized needs can be determined based on the personalized needs in the user's historical information and inquiry information.

[0065] S300, incrementally update the local knowledge base according to the product data and product service data, and obtain a first entity set before the incremental update and a second entity set after the incremental update, wherein the entity set includes entities and entity relationships, wherein the entity relationship is used to characterize at least one association relationship between product data and product services, user reviews, and service reviews.

[0066] In some embodiments, reference Figure 2 The incremental update process of the local knowledge base shown in FIG. includes but is not limited to steps S310 to S340:

[0067] S310, obtaining an entity set and an entity relationship set of a first entity set before incremental update, and constructing a first triple set and a first space vector according to the entity set and the entity relationship set;

[0068] S320, selecting a second triple that is similar to the first entity set according to the entity set and the entity relationship set of the second entity set;

[0069] S330, determining the spatial vectors of entities added to the first entity set and the spatial positions of entity relationships by using a training method according to the first triple set, the second triple and the first entity set;

[0070] S340: Determine a second spatial vector of a second triplet according to the spatial vector of the added entity and the spatial position of the entity relationship.

[0071] In some embodiments, the first triple set and the first space vector represent two entities and the entity relationship between the two entities when the local data is updated, such as (tire A, car A, a user's comment on car A using tire A), where the entity relationship between the two entities can also be a user's after-sales service evaluation of car A or tire A, etc.

[0072] It is understandable that the second triplet can only determine the data newly added to the knowledge base, and the second triplet cannot determine the knowledge base after the incremental update. It is necessary to determine the spatial vector of the entity in the second triplet and the spatial position of the entity relationship through the second spatial vector to obtain the incrementally updated knowledge base and facilitate subsequent predictions.

[0073] S400, predicting the second entity set using a graph neural network according to the query object to obtain a prediction result, wherein the prediction result represents the optimal correlation between the query object and the product data and the product service data.

[0074] In some embodiments, the incrementally updated database is predicted through a graph neural network to obtain the optimal response result for "I want to know the most cost-effective car tires recently, and my car is A car all-season tires", such as "A car", "B brand all-season tires" and "recent time".

[0075] S500: Generate a customer service response result based on the prediction result and send it to the client.

[0076] According to the above embodiment, a reply generated on the user client is "It is recommended that you buy all-season tires of brand C released in a certain year and month". In some embodiments, it can be added to the customer service reply through a product data related link.

[0077] S600~S1000 are the prediction processes of the graph neural network, which include but are not limited to:

[0078] S600, determining a dynamic knowledge base according to a local knowledge base, a first entity set and a second entity set, wherein the dynamic knowledge base includes an entity set, an entity relationship set and a time series of incremental updates of the first entity set.

[0079] In some embodiments, a third entity set consisting of a first entity set and a second entity set, an entity relationship set included in the third entity set, and a time set for each update of the local knowledge base are obtained, and a dynamic knowledge base is determined based on the third entity set, the relationship set of the third entity set, and the time set.

[0080] S700, determine entity events according to the time series of incremental updates of the first entity set, entities and entity relationships, obtain an entity event set of the dynamic knowledge base, and obtain an embedding vector according to the entity set, entity relationship set, the time series of incremental updates of the first entity set and the entity event set.

[0081] S800 uses a graph neural network-based time series encoder with an embedded vector as input to encode and learn the interactive features of entities, entity relationships, and time series.

[0082] In some embodiments, reference Figure 3 The multi-dimensional relationship prediction process diagram of the dynamic knowledge base shown includes but is not limited to steps S810 to S830:

[0083] S810, determining events obtained through corresponding entity relationships between any two associated entities in the third entity set according to the time set, and then obtaining an event set corresponding to the third entity set, wherein the events include generation time, network location, comment object, comment method, and subsequent result of the comment of the service data;

[0084] S820, determining an embedding vector according to the third entity set, the relationship set of the third entity set, the event set, and the time set, wherein the embedding vector is a four-tuple;

[0085] S830, using the encoder of the graph convolutional neural network to encode and learn the quadruple to obtain the interactive features of entities, entity relationships and time series.

[0086] S900 uses a time series decoder based on a graph neural network to aggregate neighborhood features of interactive features to obtain the time dimension features and similarity of the embedded vector.

[0087] In some embodiments, by selecting a central node, the quaternary group including the central node is divided, that is, the quaternary group includes a head entity, an entity relationship, a tail entity and time, the head entity is the central node, and the head entity and the tail entity represent their chronological relationship, such as the relationship between a vehicle model and a tire model; the embedded vector representation of the quaternary group is processed using a linear transformation matrix to reduce the matrix dimension.

[0088] In some embodiments, the contribution of the quadruple to the central node, i.e., the attention value, is learned through an entity embedding method, and the attention value is normalized, wherein the contribution can be calculated using a ReLU function or a LeakyRelu function.

[0089] In some embodiments, the embedding vector representation is iteratively updated by aggregating the attention values ​​of the central node, and the hinge loss is used for training. The loss training is performed by adding negative samples to the original third entity set, where the negative samples are quadruplets with no entity relationship or wrong entity relationship. Figure 4 The schematic diagram of the time dimension feature and similarity prediction process shown in FIG. 1 includes but is not limited to steps S910 to S920:

[0090] S910, obtaining a central node of the interactive feature, obtaining a neighborhood feature through the central node aggregation field, and obtaining a multi-dimensional feature of a quadruple according to the neighborhood feature, wherein the multi-dimensional feature at least includes a time dimension feature;

[0091] S920, using a convolutional neural network to learn the multi-dimensional features of the quadruple using a boundary loss training method to obtain the similarity of the multi-dimensional features of the quadruple.

[0092] S1000, predicting the optimal entity relationship of the query object in the dynamic knowledge base according to the time dimension characteristics and similarity.

[0093] In some embodiments, the central entity captures features of multiple dimensions in the quadruple by aggregating neighborhood features, especially time dimension features. The embedded quadruple is used as the input matrix, and the convolution layer uses convolution kernels of various sizes to extract the quadruple features. The calculated feature vector is converted into a numerical value through the matrix to evaluate the similarity of the quadruple to determine the entity relationship of the query object.

[0094] In some embodiments, keywords of the prediction results are extracted, and the keywords are reassembled using NLP and a preset corpus to obtain a response result.

[0095] For example, if the result is car model A, candidate tire model B and tire model C, the service life of tire model B is better than that of car model A, and the after-sales attitude of tire model B is not good. Therefore, the customer service response obtained by reassembling and reorganizing the corpus through NLP is "It is recommended to use tire model B for car model A, but tire model B is not good. Tire model C has average adaptability but good after-sales service."

[0096] Figure 5 Schematic diagram of an intelligent customer service device according to an embodiment of the present invention. The device includes a first module 510, a second module 520, a third module 530, a fourth module 540, a fifth module 550, a sixth module 560, a seventh module 570, an eighth module 580, a ninth module 590 and a tenth module 5100.

[0097] Among them, the first module 510 is used to obtain the inquiry object according to the customer service inquiry request of the user client, wherein the inquiry object includes the product identification and inquiry information; the second module 520 is used to determine the target product database according to the inquiry object, and obtain product data and product service data from the target product database, wherein the product service data includes at least one of product services, user comments, and service comments; the third module 530 is used to incrementally update the local knowledge base according to the product data and product service data, and obtain the first entity set before the incremental update and the second entity set after the incremental update, wherein the entity set includes entities and entity relationships, wherein the entity relationship is used to characterize the association between product data and at least one of product services, user comments, and service comments; the fourth module 540 is used to predict the second entity set using a graph neural network according to the inquiry object to obtain a prediction result, wherein the prediction result represents the optimal association between the inquiry object and the product data and product service data; the fifth module 550 is used to generate a customer service reply result from the prediction result and send it to the client; wherein the prediction of the graph neural network includes the sixth module Block 560, seventh module 570, eighth module 580, ninth module 590 and tenth module 5100; the sixth module 560 is used to determine the dynamic knowledge base according to the local knowledge base, the first entity set and the second entity set, and the dynamic knowledge base includes the entity set, the entity relationship set, and the time series of the incremental update of the first entity set; the seventh module 570 is used to determine the entity event according to the time series of the incremental update of the first entity set, the entity and the entity relationship, and obtain the entity event set of the dynamic knowledge base, and obtain the embedding vector according to the entity set, the entity relationship set, the time series of the incremental update of the first entity set and the entity event set; the eighth module 580 is used to use the time series encoder based on the graph neural network to use the embedding vector as input, encode and learn to obtain the interactive features of the entity, entity relationship and time series; the ninth module 590 uses the time series decoder based on the graph neural network to aggregate the neighborhood features of the interactive features to obtain the time dimension features and similarity of the embedding vector; the tenth module 5100 is used to predict the optimal entity relationship of the query object in the dynamic knowledge base according to the time dimension features and similarity.Exemplarily, with the cooperation of the first module 510 to the tenth module 5100 in the device, the embodiment device can implement any of the aforementioned intelligent customer service methods, namely, obtaining an inquiry object according to a customer service inquiry request of a user client, wherein the inquiry object includes a product identification and inquiry information; determining a target product database according to the inquiry object, and obtaining product data and product service data from the target product database, wherein the product service data includes at least one of product services, user comments, and service comments; incrementally updating a local knowledge base according to the product data and product service data, and obtaining a first entity set before the incremental update and a second entity set after the incremental update, wherein the entity set includes entities and entity relationships, wherein the entity relationship is used to characterize at least one association relationship between product data and product services, user comments, and service comments; predicting the second entity set using a graph neural network according to the inquiry object to obtain a prediction result, wherein the prediction result represents an association between the inquiry object and the product data and product services. The optimal association relationship of data; the prediction result is generated into a customer service reply result and sent to the client; the prediction of the graph neural network includes: determining the dynamic knowledge base according to the local knowledge base, the first entity set and the second entity set, the dynamic knowledge base includes the entity set, the entity relationship set, and the time series of the incremental update of the first entity set; determining the entity event according to the time series of the incremental update of the first entity set, the entity and the entity relationship, and obtaining the entity event set of the dynamic knowledge base, and obtaining the embedding vector according to the entity set, the entity relationship set, the time series of the incremental update of the first entity set and the entity event set; using the time series encoder based on the graph neural network to take the embedding vector as input, encode and learn to obtain the interactive features of entities, entity relationships and time series; using the time series decoder based on the graph neural network to aggregate the neighborhood features of the interactive features to obtain the time dimension features and similarity of the embedding vector; predicting the optimal entity relationship of the query object in the dynamic knowledge base according to the time dimension features and similarity. The beneficial effects of the present invention are: according to user inquiry requests, the service data and product data of multiple manufacturers are incrementally updated to build a user personalized dynamic knowledge base; a graph convolutional neural network is used to predict entities, entity relationships, time and time dimensions of the user personalized dynamic knowledge base, and provide customer service and real-time services that meet user personalization.

[0098] An embodiment of the present invention further provides an electronic device, the electronic device comprising a processor and a memory;

[0099] The memory stores a program;

[0100] The processor executes the program to execute the aforementioned intelligent customer service method; the electronic device has the function of carrying and running the software system of the intelligent customer service provided by the embodiment of the present invention, for example, a personal computer, a minicomputer, a main frame, a workstation, a network or distributed computing environment, a separate or integrated computer platform, or communicates with a charged particle tool or other imaging device, etc.

[0101] An embodiment of the present invention further provides a computer-readable storage medium, wherein the storage medium stores a program, and the program is executed by a processor to implement the intelligent customer service method as described above.

[0102] In some selectable embodiments, the function / operation mentioned in the block diagram may not occur in the order mentioned in the operation diagram. For example, depending on the function / operation involved, the two boxes shown in succession can actually be executed substantially simultaneously or the boxes can sometimes be executed in reverse order. In addition, the embodiment presented and described in the flow chart of the present invention is provided by way of example, for the purpose of providing a more comprehensive understanding of technology. The disclosed method is not limited to the operation and logic flow presented herein. Selectable embodiments are expected, wherein the order of various operations is changed and the sub-operation of a part for which is described as a larger operation is performed independently.

[0103] The embodiment of the present invention also discloses a computer program product or a computer program, which includes a computer instruction stored in a computer-readable storage medium. A processor of a computer device can read the computer instruction from the computer-readable storage medium, and the processor executes the computer instruction, so that the computer device executes the aforementioned intelligent customer service method.

[0104] In addition, although the present invention is described in the context of functional modules, it should be understood that, unless otherwise specified, one or more of the functions and / or features described may be integrated into a single physical device and / or software module, or one or more functions and / or features may be implemented in separate physical devices or software modules. It is also understood that a detailed discussion of the actual implementation of each module is unnecessary for understanding the present invention. More specifically, in view of the properties, functions, and internal relationships of the various functional modules in the device disclosed herein, the actual implementation of the module will be understood within the conventional skills of the engineer. Therefore, those skilled in the art can implement the present invention set forth in the claims without excessive experimentation using ordinary techniques. It is also understood that the specific concepts disclosed are merely illustrative and are not intended to limit the scope of the present invention, which is determined by the full scope of the appended claims and their equivalents.

[0105] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program codes.

[0106] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0107] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0108] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0109] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0110] Although the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the claims and their equivalents.

[0111] The above is a specific description of the preferred implementation of the present invention, but the present invention is not limited to the described embodiments. Those skilled in the art may make various equivalent modifications or substitutions without violating the spirit of the present invention. These equivalent modifications or substitutions are all included in the scope defined by the claims of this application.

Claims

1. An intelligent customer service method, characterized in that: include: According to the customer service inquiry request of the user client, the inquiry object is obtained, wherein the inquiry object includes the product identification and inquiry information; Determine a target product database according to the query object, and obtain product data and product service data from the target product database, wherein the product service data includes at least one of product service, user comments, and service comments; Incrementally updating the local knowledge base according to the product data and the product service data, and obtaining a first entity set before the incremental update and a second entity set after the incremental update, wherein the entity set includes entities and entity relationships, wherein the entity relationship is used to represent at least one association relationship between the product data and product services, user reviews, and service reviews; According to the query object, the second entity set is predicted using a graph neural network to obtain a prediction result, wherein the prediction result represents an optimal correlation relationship between the query object and the product data and the product service data; Generate a customer service response result from the prediction result and send it to the client; The predictions of the graph neural network include: Determine a dynamic knowledge base according to the local knowledge base, the first entity set and the second entity set, wherein the dynamic knowledge base includes an entity set, an entity relationship set, and a time series of incremental updates of the first entity set; Determine entity events according to the time series of incremental updates of the first entity set, entities and entity relationships, obtain an entity event set of the dynamic knowledge base, and obtain an embedding vector according to the entity set, the entity relationship set, the time series of incremental updates of the first entity set and the entity event set; A graph neural network-based time series encoder is used to take the embedding vector as input to encode and learn the interactive features of entities, entity relationships, and time series. A time series decoder based on graph neural network is used to aggregate the neighborhood features of the interaction features to obtain the time dimension features and similarity of the embedded vector; Predicting the optimal entity relationship of the query object in the dynamic knowledge base according to the time dimension characteristics and similarity; Determining a dynamic knowledge base according to the local knowledge base, the first entity set, and the second entity set includes: Acquire a third entity set consisting of the first entity set and the second entity set, an entity relationship set included in the third entity set, and acquire a time set for each update of the local knowledge base, and determine the dynamic knowledge base according to the third entity set, the relationship set of the third entity set and the time set; The time series encoder based on the graph neural network uses the embedding vector as input to encode and learn the interactive features of entities, entity relationships and time series, including: Determine, according to the time set, events obtained through corresponding entity relationships between any two associated entities in the third entity set, and then obtain an event set corresponding to the third entity set, wherein the events include generation time, network location, comment object, comment method, and subsequent result of the comment of the service data; Determine an embedding vector according to the third entity set, the relationship set of the third entity set, the event set, and the time set, wherein the embedding vector is a four-tuple; The encoder of the graph convolutional neural network is used to encode and learn the quadruple to obtain the interactive features of entities, entity relationships and time series; The time series decoder based on the graph neural network is used to aggregate the neighborhood features of the interaction features to obtain the time dimension features and similarity of the embedded vector, including: Obtaining a central node of the interactive feature, obtaining a neighborhood feature through aggregating fields of the central node, and obtaining a multidimensional feature of the quadruple according to the neighborhood feature, wherein the multidimensional feature at least includes a time dimension feature; The multidimensional features of the quadruple are learned by using a convolutional neural network with a boundary loss training method to obtain the similarity of the multidimensional features of the quadruple.

2. The intelligent customer service method according to claim 1, characterized in that: The step of determining a target product database according to the query object and acquiring product data and product service data from the target product database includes: Determine the service platform of the product according to the product identification, obtain the interface configuration information of the service platform, pull the product data and product service data from the service platform according to the interface configuration information, and record the pulling time each time data is pulled.

3. The intelligent customer service method according to claim 2, characterized in that: The incrementally updating the local knowledge base according to the product data and the product service data, and obtaining the first entity set before the incremental update and the second entity set after the incremental update, includes: Obtaining an entity set and an entity relationship set of a first entity set before incremental update, and constructing a first triple set and a first space vector according to the entity set and the entity relationship set; Selecting a second triplet similar to the first entity set according to the entity set and the entity relationship set of the second entity set; Determine the spatial vectors of entities added to the first entity set and the spatial positions of entity relationships by training according to the first triple set, the second triple and the first entity set; A second spatial vector of the second triple is determined according to the spatial vector of the added entity and the spatial position of the entity relationship.

4. The intelligent customer service method according to claim 1, characterized in that: The generating a customer service response result from the prediction result and sending the result to the client includes: The keywords of the prediction results are extracted, and the keywords are reassembled using NLP and a preset corpus to obtain the reply result.

5. An intelligent customer service device, characterized in that: include: The first module is used to obtain an inquiry object according to a customer service inquiry request from a user client, wherein the inquiry object includes a product identifier and inquiry information; The second module is used to determine a target product database according to the query object, and obtain product data and product service data from the target product database, wherein the product service data includes at least one of product service, user comments, and service comments; A third module is used to incrementally update the local knowledge base according to the product data and the product service data, and obtain a first entity set before the incremental update and a second entity set after the incremental update, wherein the entities include entity relationships, wherein the entity relationships are used to characterize at least one association relationship between the product data and the product service, the user review, and the service review; A fourth module is used to use a graph neural network to perform prediction based on the query object to obtain a prediction result, wherein the prediction result represents the optimal correlation between the query object and the product data and the product service data; The fifth module is used to generate a customer service reply result based on the prediction result and send it to the client; The prediction of the graph neural network includes the sixth module, the seventh module, the eighth module, the ninth module and the tenth module; The sixth module is used to determine a dynamic knowledge base according to the local knowledge base, the first entity set and the second entity set, wherein the dynamic knowledge base includes an entity set, an entity relationship set, and a time series of incremental updates of the first entity set; The seventh module is used to determine entity events according to the time series of the incremental update of the first entity set, entities and entity relationships, obtain the entity event set of the dynamic knowledge base, and obtain the embedding vector according to the entity set, the entity relationship set, the time series of the incremental update of the first entity set and the entity event set; The eighth module is used to use a time series encoder based on a graph neural network to take an embedding vector as input, encode and learn the interactive features of entities, entity relationships and time series; The ninth module is used to use a time series decoder based on a graph neural network to aggregate neighborhood features of the interaction features to obtain time dimension features and similarity of the embedded vector; The tenth module is used to predict the entity relationship of the query object according to the time dimension characteristics and similarity; Determining a dynamic knowledge base according to the local knowledge base, the first entity set, and the second entity set includes: Acquire a third entity set consisting of the first entity set and the second entity set, an entity relationship set included in the third entity set, and acquire a time set for each update of the local knowledge base, and determine the dynamic knowledge base according to the third entity set, the relationship set of the third entity set and the time set; The time series encoder based on the graph neural network uses the embedding vector as input to encode and learn the interactive features of entities, entity relationships and time series, including: Determine, according to the time set, events obtained through corresponding entity relationships between any two associated entities in the third entity set, and then obtain an event set corresponding to the third entity set, wherein the events include generation time, network location, comment object, comment method, and subsequent result of the comment of the service data; Determine an embedding vector according to the third entity set, the relationship set of the third entity set, the event set, and the time set, wherein the embedding vector is a four-tuple; The encoder of the graph convolutional neural network is used to encode and learn the quadruple to obtain the interactive features of entities, entity relationships and time series; The time series decoder based on the graph neural network is used to aggregate the neighborhood features of the interaction features to obtain the time dimension features and similarity of the embedded vector, including: Obtaining a central node of the interactive feature, obtaining a neighborhood feature through aggregating fields of the central node, and obtaining a multidimensional feature of the quadruple according to the neighborhood feature, wherein the multidimensional feature at least includes a time dimension feature; The multidimensional features of the quadruple are learned by using a convolutional neural network with a boundary loss training method to obtain the similarity of the multidimensional features of the quadruple.

6. An electronic device, characterized in that: including a processor and a memory; The memory is used to store programs; The processor executes the program to implement the intelligent customer service method according to any one of claims 1 to 4.

7. A computer-readable storage medium, characterized in that: The storage medium stores a program, and the program is executed by a processor to implement the intelligent customer service method according to any one of claims 1 to 4.

Images