Customer service system intention recognition method, device and equipment based on large model

Abstract

The embodiment of the invention relates to a customer service system intention recognition method, device and equipment based on a large model. The method comprises the following steps: acquiring first-round interaction information and vectorizing the first-round interaction information to obtain a first-round interaction information vector; performing vector retrieval from the scene knowledge base according to the first-round interaction information vector to obtain first-round scene information with the highest matching degree and the approximation degree; if the approximation degree of the first-round scene information exceeds a first-round scene threshold value, taking a scene corresponding to the first-round scene information as a target scene, calling a process engine, calling the target scene by the process engine, and starting to execute interaction with the user according to a link corresponding to the target scene; and during each round of link interaction of the target scene, judging whether scene switching or link switching is needed or not. The coverage of user regret, interruption, intention change and the like in a customer service business process is realized; information can be synchronized in time when flow links change, and redevelopment / training required by service adjustment is avoided.

Description

Technical Field

[0001] The embodiments of this specification relate to the field of computer technology, and in particular to a method, device and equipment for customer service system intent recognition based on a large model. Background Art

[0002] Although the widespread application of intelligent customer service systems in the field of customer service has greatly improved service efficiency, they still face many challenges in actual operation, which often affect customer experience. Specifically, they are as follows: ① User needs are diverse, involving business handling, consultation, complaints and other aspects, which are difficult for intelligent customer service to fully cover; ② Traditional customer service systems rely on machine learning or deep learning combined with rules to implement service processes. This model has great limitations in interactive scenarios; ③ The system lacks a flexible process jump and interruption mechanism, resulting in an inability to effectively respond in complex conversations. These defects cause intelligent customer service to be unable to cope with non-preset scenarios, and frequent errors in response, which ultimately leads to an increase in the transfer rate to manual services.

[0003] The root cause is that the existing intelligent customer service system mainly relies on natural language processing (NLP) and rule engines to identify customer intentions, and is often helpless in situations outside the rules, which limits the application scope of intelligent customer service. Summary of the invention

[0004] In order to solve the problems existing in the prior art, the embodiments of this specification provide a method, device and equipment for customer service system intention recognition based on a large model, which integrates multiple technologies such as large models and vector database retrieval to handle abnormal situations in conversations, thereby accurately identifying user intentions. In addition, a mechanism is designed to allow intelligent customer service to switch and jump freely in various business scenarios or links.

[0005] The specific technical solutions of the embodiments of this specification are as follows:

[0006] On the one hand, the embodiments of this specification provide a customer service system intention recognition method based on a large model, including:

[0007] Acquire the user's first-round interaction information, and vectorize the first-round interaction information to obtain a first-round interaction information vector;

[0008] Performing vector retrieval from a scene knowledge base according to the first-round interaction information vector to obtain the first-round scene information with the highest matching degree and the approximation of the first-round scene information, wherein the scene knowledge base pre-stores vectors corresponding to a plurality of scenes;

[0009] If the similarity of the first-round scenario information exceeds the first-round scenario threshold, the scenario corresponding to the first-round scenario information is used as the target scenario, and the process engine is called, the process engine calls up the target scenario, and starts to interact with the user according to the steps corresponding to the target scenario;

[0010] Wherein, in each round of link interaction of the target scene, the current interactive link is taken as the target link, and whether the scene needs to be switched is determined according to the vector of the target link interaction information of the user and the scene knowledge base;

[0011] If yes, switch to the new target scene and execute the step of calling the new target scene by the process engine;

[0012] If not, judging whether the link needs to be switched according to the vector of the target link interaction information and the link knowledge base, wherein the link knowledge base pre-stores vectors corresponding to multiple links;

[0013] If yes, switch to a new target link, and perform interaction with the user according to the new target link;

[0014] If not, continue with the target link and perform the interaction.

[0015] Furthermore, judging whether it is necessary to switch scenes according to the vector of the user's target link interaction information and the scene knowledge base further includes:

[0016] Performing vector retrieval from the scene knowledge base according to the vector of the target link interaction information to obtain the switching scene information with the highest matching degree and the approximation of the switching scene information;

[0017] Determining whether the similarity of the switching scene information exceeds a first scene switching threshold;

[0018] If so, the scene corresponding to the switching scene information is used as the new target scene, and all information in the current interaction process is cached, so that after switching to the new target scene, the process engine executes the step of calling the new target scene based on all cached information.

[0019] Furthermore, if the similarity of the switching scene information does not exceed the scene switching first threshold, the method further includes:

[0020] Determining whether the similarity of the switching scene information is between a second scene switching threshold and the first scene switching threshold;

[0021] If so, a prompt word is constructed according to the target link interaction information, the switching scene information, the specified prompt word and the specified output format, and the prompt word is input into a pre-trained large model for calculation to obtain a result of whether a scene switching is required.

[0022] Furthermore, if the similarity of the switching scene information does not exceed the second scene switching threshold, there is no need to perform scene switching.

[0023] Furthermore, after switching to the new target scene, the method further includes:

[0024] Determine whether the new target scene interaction link is a basic interaction link;

[0025] If yes, determining whether all the cached information includes the interaction information corresponding to the basic interaction link;

[0026] If yes, skip this basic interaction step.

[0027] Furthermore, judging whether a link needs to be switched according to the vector of the target link interaction information and the link knowledge base further includes:

[0028] Performing vector retrieval from the link knowledge base according to the vector of the target link interaction information to obtain the switching link information with the highest matching degree and the approximation of the switching link information;

[0029] Determine whether the similarity of the switching link information exceeds the link switching threshold;

[0030] If yes, it is determined whether the link corresponding to the switching link information is the same as the target link.

[0031] If so, there is no need to switch links;

[0032] If not, it is necessary to switch the link, and use the link corresponding to the switching link information as the new target link.

[0033] Furthermore, if the similarity of the switching link information does not exceed the link switching threshold, the method further includes:

[0034] Stop the interaction of the target link, generate the minimum interaction script by the pre-trained large model and execute the interaction.

[0035] Furthermore, if the similarity of the first-round scene information does not exceed the first-round scene threshold, the method further includes:

[0036] The pre-trained large model generates the minimum interactive script and executes the interaction.

[0037] On the other hand, the embodiment of this specification also provides a customer service system intention recognition device based on a large model, the device comprising:

[0038] An interaction information acquisition unit, used to acquire the first-round interaction information of the user, and vectorize the first-round interaction information to obtain a first-round interaction information vector;

[0039] A scene knowledge base retrieval unit, configured to perform vector retrieval from a scene knowledge base according to the first-round interaction information vector, to obtain the first-round scene information with the highest matching degree and the approximation of the first-round scene information, wherein the scene knowledge base pre-stores vectors corresponding to multiple scenes;

[0040] An interaction unit, configured to, when the similarity of the first-round scenario information exceeds a first-round scenario threshold, use the scenario corresponding to the first-round scenario information as a target scenario, and call a process engine, wherein the process engine calls up the target scenario, and starts to perform interaction with the user according to the steps corresponding to the target scenario;

[0041] Wherein, in each round of link interaction of the target scene, the current interactive link is taken as the target link, and whether the scene needs to be switched is determined according to the vector of the target link interaction information of the user and the scene knowledge base;

[0042] If yes, switch to the new target scene and execute the step of calling the new target scene by the process engine;

[0043] If not, judging whether the link needs to be switched according to the vector of the target link interaction information and the link knowledge base, wherein the link knowledge base pre-stores vectors corresponding to multiple links;

[0044] If yes, switch to a new target link, and perform interaction with the user according to the new target link;

[0045] If not, continue with the target link and perform the interaction.

[0046] On the other hand, an embodiment of the present specification further provides a computer device, comprising a memory, a processor, and a computer program stored in the memory, wherein the processor implements the above method when executing the computer program.

[0047] The customer service system intention recognition method based on a big model in the embodiment of this specification implements coverage of situations such as user regret, interruption, change of intention, etc. that are often encountered in the entire customer service business process; information can be synchronized conveniently and timely when the process links change, avoiding the need for redevelopment / training for business adjustments. In the field of intent recognition, big model technology fully explores and maximizes its recognition capabilities, effectively avoiding the risk of big model illusions. Compared with the existing technology, the main advantages are: flexibility, supporting flexible backtracking in links according to user intentions, or repeatedly jumping to new scenes. Business continuity, through the mechanism of caching information, key information can be retained during scene switching and link jumping. If you return to the original scene / original link again, the original information has been obtained and there is no need to repeatedly ask the user questions again, ensuring business continuity. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the embodiments of this specification or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the embodiments of this specification. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0049] Figure 1 The figure is a flow chart of a method for identifying customer service system intention based on a large model according to an embodiment of the present specification;

[0050] Figure 2 The figure is a schematic diagram of the process of scene switching in the embodiment of this specification;

[0051] Figure 3 The figure is a schematic diagram of the process of performing link switching in the embodiment of this specification;

[0052] Figure 4 The figure shows a schematic diagram of the structure of a customer service system intention recognition device based on a large model in an embodiment of this specification;

[0053] Figure 5 The figure is a schematic diagram of the structure of a computer device in an embodiment of the present specification.

[0054]

Description of reference numerals

[0055] 401. Interaction information acquisition unit;

[0056] 402. Scenario knowledge base retrieval unit;

[0057] 403, interactive unit;

[0058] 502. Computer equipment;

[0059] 504. Processing equipment;

[0060] 506. Storage resources;

[0061] 508. Drive system;

[0062] 510, input / output module;

[0063] 512. Input devices;

[0064] 514. Output device;

[0065] 516. Presentation equipment;

[0066] 518. Graphical user interface;

[0067] 520, network interface;

[0068] 522, communication link;

[0069] 524. Communication bus. DETAILED DESCRIPTION

[0070] The following will be combined with the drawings in the embodiments of this specification to clearly and completely describe the technical solutions in the embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this specification, not all of them. Based on the embodiments in the embodiments of this specification, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the embodiments of this specification.

[0071] It should be noted that the terms "first", "second", etc. in the description and claims of the embodiments of this specification and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the embodiments of this specification described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, device, product or equipment that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or equipment.

[0072] It should be noted that the acquisition, storage, use, and processing of data in the technical solutions of the embodiments of this specification comply with the relevant provisions of national laws and regulations.

[0073] It should be noted that in the embodiments of this specification, certain software, components, models and other existing solutions in the industry may be mentioned, and they should be regarded as exemplary. Their purpose is only to illustrate the feasibility of implementing the technical solution of this application, but it does not mean that the applicant has or will necessarily use the solution.

[0074] In view of the problems existing in the prior art, the embodiments of this specification provide a method, device and equipment for customer service system intention recognition based on a large model, which integrates multiple technologies such as large models and vector database retrieval to handle abnormal situations in conversations, thereby accurately identifying user intentions. In addition, a mechanism is designed to allow intelligent customer service to switch and jump freely in various business scenarios or links. Figure 1 The figure shows a flow chart of the customer service system intention recognition method based on a large model. The process of customer service system intention recognition is described in this figure. The order of steps listed in the embodiment is only one way of executing the steps among many steps, and does not represent the only execution order. When the actual system or device product is executed, it can be executed in the order of the method shown in the embodiment or the accompanying drawings or in parallel.

[0075] Specific as Figure 1 As shown, the method may include:

[0076] Step 101: obtaining the first round of interaction information of the user, and vectorizing the first round of interaction information to obtain a first round of interaction information vector;

[0077] Step 102: performing vector retrieval from a scene knowledge base according to the first-round interaction information vector to obtain the first-round scene information with the highest matching degree and the approximation of the first-round scene information;

[0078] In this step, vectors corresponding to multiple scenes are pre-stored in the scene knowledge base;

[0079] Step 103, determining whether the similarity of the first-round scene information exceeds a first-round scene threshold;

[0080] Step 104: if the similarity of the first-round scenario information exceeds the first-round scenario threshold, the scenario corresponding to the first-round scenario information is used as the target scenario, and the process engine is called, the process engine calls up the target scenario, and starts to interact with the user according to the steps corresponding to the target scenario;

[0081] Wherein, in each round of link interaction of the target scene, step 1041: taking the current interactive link as the target link, judging whether the scene needs to be switched according to the vector of the target link interaction information of the user and the scene knowledge base;

[0082] Step 1042: If yes, switch to the new target scene and execute the step of calling the new target scene by the process engine;

[0083] Step 1043: If not, determine whether the link needs to be switched based on the vector of the target link interaction information and the link knowledge base;

[0084] In this step, the link knowledge base pre-stores vectors corresponding to multiple links;

[0085] Step 10431: If yes, switch to a new target link, and perform interaction with the user according to the new target link;

[0086] Step 10432: If not, continue with the target link and perform the interaction.

[0087] In the embodiment of the present specification, the name and description of the scene are vectorized in advance, and these vectors are stored in the scene knowledge base.

[0088] The names and descriptions of the links are vectorized in advance, and these vectors are stored in the link knowledge base of the corresponding scenario partition.

[0089] When the user conducts the first round of interaction, the first round of interaction information is obtained and vectorized to obtain the first round of interaction information vector, and then vector retrieval is performed from the scene knowledge base based on the first round of interaction information vector to obtain the first round of scene information N with the highest matching degree and the approximation A of the first round of scene information N.

[0090] When the similarity A ≥ the first-round scenario threshold a1, it is determined that there is a matching scenario. According to the scenario information N, the process engine is called, the process engine calls up the relevant scenarios, and starts to execute interactions according to the links.

[0091] Furthermore, if Figure 1 As shown, after step 103 determines whether the similarity of the first-round scene information exceeds the first-round scene threshold, if the similarity of the first-round scene information does not exceed the first-round scene threshold, the following steps are performed:

[0092] Step 105: Generate the guaranteed interaction script using the pre-trained large model and execute the interaction.

[0093] In the embodiments of the present specification, the big model may adopt an existing general big model, input the user's first round of interaction information into the big model, generate a minimum interaction script and execute the interaction.

[0094] It should be noted that the first round of interaction information in the embodiments of this specification may be text information input by the user (for example, the user inputs the first round of interaction information by typing or voice conversation).

[0095] During each round of link interaction in the target scene, the current interactive link is taken as the target link, and whether the scene needs to be switched is determined based on the vector of the user's target link interaction information and the scene knowledge base.

[0096] Specifically, Figure 2 As shown, judging whether it is necessary to switch scenes according to the vector of the user's target link interaction information and the scene knowledge base further includes:

[0097] Step 201: performing vector retrieval from the scene knowledge base according to the vector of the target link interaction information to obtain the switching scene information with the highest matching degree and the approximation of the switching scene information;

[0098] Step 202: determining whether the similarity of the switching scene information exceeds a first scene switching threshold;

[0099] Step 203: If yes, the scene corresponding to the switching scene information is used as the new target scene, and all information in the current interaction process is cached, so that after switching to the new target scene, the process engine executes the step of calling the new target scene based on all cached information.

[0100] In each round of link interaction, the user's target link interaction information is Con. After the target link interaction information Con is vectorized, the scene knowledge base is used for vector retrieval to obtain the scene information N' and the approximation B with the highest matching degree.

[0101] When the similarity B ≥ the first scene switching threshold b1, it is determined that a scene switching is required. All information in the current interaction process is cached. According to the scene information N', the process engine is called, and the process engine calls up the relevant scene and starts to perform interaction according to the links in the new scene.

[0102] Further, step 204: if the similarity of the switching scene information does not exceed the first scene switching threshold, determining whether the similarity of the switching scene information is between the second scene switching threshold and the first scene switching threshold;

[0103] Step 2041: If yes, construct a prompt word according to the target link interaction information, the switching scene information, the specified prompt word and the specified output format, input the prompt word into the pre-trained large model for calculation, and obtain the result of whether the scene switching is required.

[0104] In the embodiments of this specification, when the second scene switching threshold b2 < B < b1, it is determined that scene reflection is required. Scene reflection means constructing a prompt word based on the target link interaction information, the switched scene information, a specified prompt, and a specified output format, and inputting the prompt word into a pre-trained large model for calculation to obtain a result indicating whether scene switching is required.

[0105] Exemplarily, input the target link interaction information Con>, <scene information N'>, the specified prompt <Whether the user's question belongs to the above description>, and the specified output format <Only output yes or no> into the large model (open-source general large model) to obtain the output result P. When P = "yes", it is determined that scene switching is required. Cache all information in the current interaction process. According to the scene information N', call the process engine, and the process engine invokes the relevant scene and starts to execute the interaction according to the links. When P = "no", it is determined that no scene switching is required.

[0106] Further, step 2042: If the approximation degree of the switched scene information is not between the second scene switching threshold and the first scene switching threshold, it is determined that the approximation degree of the switched scene information does not exceed the second scene switching threshold;

[0107] Step 20421: If so, no scene switching is required.

[0108] According to an embodiment of this specification, after switching to the new target scene in step 203, the method further includes:

[0109] Determine whether the link of the new target scene interaction is a basic interaction link;

[0110] If so, determine whether the cached all information includes the interaction information corresponding to the basic interaction link;

[0111] If so, skip this basic interaction link.

[0112] In the embodiments of this specification, the basic interaction link may be a link such as obtaining user identity information. Since the links in each scene are the complete links of that scene, and the links in different scenes may include the same basic interaction link. When performing the interaction of the links in the first round of the scene, the interaction of the basic interaction link has been executed, and user identity information, etc. has been obtained. Therefore, in the embodiments of this specification, after switching to the new target scene, it is determined whether the link of the new target scene is a basic interaction link. If so, it is skipped.

[0113] According to an embodiment of this specification, as Figure 3 shown, determining whether to switch links according to the vector of the target link interaction information and the link knowledge base further includes:

[0114] Step 301: Perform vector retrieval from the link knowledge base based on the vector of the target link interaction information to obtain the switching link information with the highest matching degree and the approximation degree of this switching link information;

[0115] Step 302: Determine whether the approximation degree of this switching link information exceeds the link switching threshold;

[0116] Step 303: If so, determine whether the link corresponding to the switching link information is the same as the target link;

[0117] Step 3031: If so, there is no need to switch links;

[0118] Step 3032: If not, link switching is required, and the link corresponding to the switching link information is used as the new target link.

[0119] Step 304: If the approximation degree of the switching link information does not exceed the link switching threshold, stop the interaction of the target link, and generate a fallback interaction script by a pre-trained large model and execute the interaction.

[0120] In the embodiments of this specification, the current link where the user is located is denoted as S. After vectorizing the target link interaction information Con, vector retrieval is performed using the link knowledge base to obtain the link information S' with the highest matching degree and the approximation degree C.

[0121] When the approximation degree C ≥ the link switching threshold c1, it is determined that the link can be switched. When S = S', continue the link and execute the interaction. When S ≠ S', cache all the information in the current interaction process, enter the specific link according to the link information S', and start executing the interaction according to the link.

[0122] When the approximation degree C < c1, it is determined that there is no matching link, and a fallback interaction script is generated by the large model and the interaction is executed. Feasibly, the process of step 105 can be referred to for generating the fallback interaction script, which will not be elaborated here.

[0123] It should be noted that the initial scene threshold, the first scene switching threshold, the second scene switching threshold, and the link switching threshold in the embodiments of this specification can be determined by the staff according to experience or tests, and this specification does not make any restrictions. The first scene switching threshold needs to be greater than the second scene switching threshold.

[0124] Exemplarily, preset scene vectors, {data package, international roaming, broadband appeal, general optional package} scenes, and their descriptions;

[0125] Preset link vectors, the links and their descriptions in the traffic package scenario: {mobile phone number extraction, customer information, customer's current package, new package selection, product subscription}; the links and their descriptions in the international roaming scenario: {mobile phone number extraction, customer information, international destination, package selection, product subscription}.

[0126] Set a1 = 70%; b1 = 80%, b2 = 60%; c1 = 70%.

[0127] 1) When the user interacts with "My traffic is not enough", according to the first-round scenario judgment, the scenario information N = "traffic package" and the approximation A = 79.23% are obtained. If A > a1, then enter the "traffic package" scenario.

[0128] 2) When the user interacts with "By the way, I'm going abroad next week", according to the scenario switching judgment, the scenario information N' = "international roaming" and the approximation B = 85.44% are obtained. If B ≥ b1, then enter the "international roaming" scenario.

[0129] 3) When the user interacts with "I want to go to Bali to see the sea", according to the scenario switching judgment, the scenario information N' = "international roaming" and the approximation B = 71.24% are obtained. If B ≥ b2 but B < b1, then enter the scenario reflection, and then confirm with the user whether it is the "international roaming" scenario. The prompt can be: "A service that allows the user to use the current mobile phone number to communicate with the device in other countries or regions after going abroad. Does the user's question conform to the current scenario description? Answer yes or no".

[0130] 4) After entering the subscription link, S = "product subscription". When the user interacts with "I chose the wrong traffic package before and don't need so much", according to the scenario switching judgment, the scenario information N' = "traffic package" and the approximation B = 55.92% are obtained. If B ≤ b2, then perform link backtracking.

[0131] 5) Vectorize the user interaction "I chose the wrong traffic package before and don't need so much", and then use the scenario knowledge base for vector retrieval to obtain the scenario information S' = "new package selection" with the highest matching degree and the approximation C = 83.73%.

[0132] 6) If S ≠ S', cache all the information in the current interaction process, enter the "new package selection" link, and start executing the interaction according to the link.

[0133] Based on the same inventive concept, an embodiment of this specification also provides an intent recognition device for a customer service system based on a large model, as Figure 4 shown. The device includes:

[0134] An interaction information acquisition unit 401, configured to acquire the first-round interaction information of the user, and vectorize the first-round interaction information to obtain a first-round interaction information vector.

[0135] A scene knowledge base retrieval unit 402 is used to perform vector retrieval from a scene knowledge base according to the first-round interaction information vector to obtain the first-round scene information with the highest matching degree and the approximation of the first-round scene information, wherein the scene knowledge base pre-stores vectors corresponding to multiple scenes;

[0136] The interaction unit 403 is used to, when the similarity of the first-round scenario information exceeds the first-round scenario threshold, use the scenario corresponding to the first-round scenario information as the target scenario, and call the process engine, so that the process engine calls up the target scenario and starts to perform interaction with the user according to the link corresponding to the target scenario;

[0137] Wherein, in each round of link interaction of the target scene, the current interactive link is taken as the target link, and whether the scene needs to be switched is determined according to the vector of the target link interaction information of the user and the scene knowledge base;

[0138] If yes, switch to the new target scene and execute the step of calling the new target scene by the process engine;

[0139] If not, judging whether the link needs to be switched according to the vector of the target link interaction information and the link knowledge base, wherein the link knowledge base pre-stores vectors corresponding to multiple links;

[0140] If yes, switch to a new target link, and perform interaction with the user according to the new target link;

[0141] If not, continue with the target link and perform the interaction.

[0142] Since the principle of solving the problem by the above device is similar to that of the above method, the implementation of the above device can refer to the implementation of the above method, and the repeated parts will not be repeated.

[0143] like Figure 5 The diagram is a schematic diagram of the structure of a computer device in an embodiment of the present specification. The apparatus in an embodiment of the present specification may be a computer device in an embodiment of the present specification, executing the method in the embodiment of the present specification described above.

[0144] The computer device 502 may include one or more processing devices 504, such as one or more central processing units (CPUs), each of which may implement one or more hardware threads. The computer device 502 may also include any storage resources 506 for storing any type of information such as code, settings, data, etc. Without limitation, for example, the storage resources 506 may include any one or more combinations of the following: any type of RAM, any type of ROM, a flash memory device, a hard disk, an optical disk, etc. More generally, any storage resource may use any technology to store information.

[0145] Further, any storage resource may provide volatile or non-volatile retention of information.

[0146] Further, any storage resource may represent a fixed or removable component of the computer device 502. In one case, when the processing device 504 executes the associated instructions stored in any storage resource or combination of storage resources, the computer device 502 may perform any operation of the associated instructions. The computer device 502 also includes one or more drive systems 508 for interacting with any storage resource, such as a hard disk drive system, an optical disk drive system, etc.

[0147] The computer device 502 may also include an input / output module 510 (I / O) for receiving various inputs (via input devices 512) and for providing various outputs (via output devices 514). A specific output mechanism may include a presentation device 516 and an associated graphical user interface (GUI) 518. In other embodiments, the input / output module 510 (I / O), the input device 512, and the output device 514 may not be included, and the computer device 502 may be used as a computer device in a network. The computer device 502 may also include one or more network interfaces 520 for exchanging data with other devices via one or more communication links 522. One or more communication buses 524 couple the components described above together.

[0148] The communication link 522 may be implemented in any manner, for example, through a local area network, a wide area network (e.g., the Internet), a point-to-point connection, etc., or any combination thereof. The communication link 522 may include any combination of hardwired links, wireless links, routers, gateway functions, name servers, etc., governed by any protocol or combination of protocols.

[0149] The embodiments of the present specification also provide a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and the computer program implements the above method when executed by a processor.

[0150] The embodiments of the present specification also provide a computer-readable instruction, wherein when a processor executes the instruction, the program therein causes the processor to execute the above method.

[0151] It should be understood that in the various embodiments of the present specification, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present specification.

[0152] It should also be understood that in the embodiments of this specification, the term "and / or" is only a description of the association relationship of the associated objects, indicating that three relationships can exist. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in the embodiments of this specification generally indicates that the associated objects before and after are in an "or" relationship.

[0153] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed in the embodiments of this specification can be implemented with electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are 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 exceed the scope of the embodiments of this specification.

[0154] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0155] In the several embodiments provided in the embodiments of this specification, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, or it can be an electrical, mechanical or other form of connection.

[0156] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the embodiments of this specification.

[0157] In addition, each functional unit in each embodiment of the present specification can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The above integrated unit can be implemented in the form of hardware or in the form of software functional units.

[0158] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiment of this specification is essentially or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the embodiment of this specification. 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 and other media that can store program code.

[0159] The embodiments of this specification use specific embodiments to illustrate the principles and implementation methods of the embodiments of this specification. The description of the above embodiments is only used to help understand the methods and core ideas of the embodiments of this specification. At the same time, for those skilled in the art, according to the ideas of the embodiments of this specification, there will be changes in the specific implementation methods and application scopes. In summary, the content of this specification should not be understood as a limitation on the embodiments of this specification.

Claims

1. A customer service system intention recognition method based on a large model, characterized in that: The method comprises: Acquire the user's first-round interaction information, and vectorize the first-round interaction information to obtain a first-round interaction information vector; Performing vector retrieval from a scene knowledge base according to the first-round interaction information vector to obtain the first-round scene information with the highest matching degree and the approximation of the first-round scene information, wherein the scene knowledge base pre-stores vectors corresponding to a plurality of scenes; If the similarity of the first-round scenario information exceeds the first-round scenario threshold, the scenario corresponding to the first-round scenario information is used as the target scenario, and the process engine is called, the process engine calls up the target scenario, and starts to interact with the user according to the steps corresponding to the target scenario; Wherein, in each round of link interaction of the target scene, the current interactive link is taken as the target link, and whether the scene needs to be switched is determined according to the vector of the target link interaction information of the user and the scene knowledge base; If yes, switch to the new target scene and execute the step of calling the new target scene by the process engine; If not, judging whether the link needs to be switched according to the vector of the target link interaction information and the link knowledge base, wherein the link knowledge base pre-stores vectors corresponding to multiple links; If yes, switch to a new target link, and perform interaction with the user according to the new target link; If not, continue with the target link and perform the interaction.

2. The method according to claim 1, characterized in that: Judging whether the scene needs to be switched according to the vector of the user's target link interaction information and the scene knowledge base further includes: Performing vector retrieval from the scene knowledge base according to the vector of the target link interaction information to obtain the switching scene information with the highest matching degree and the approximation of the switching scene information; Determining whether the similarity of the switching scene information exceeds a first scene switching threshold; If so, the scene corresponding to the switching scene information is used as the new target scene, and all information in the current interaction process is cached, so that after switching to the new target scene, the process engine executes the step of calling the new target scene based on all cached information.

3. The method according to claim 2, characterized in that If the similarity of the switching scene information does not exceed the scene switching first threshold, the method further includes: Determining whether the similarity of the switching scene information is between a second scene switching threshold and the first scene switching threshold; If so, a prompt word is constructed according to the target link interaction information, the switching scene information, the specified prompt word and the specified output format, and the prompt word is input into a pre-trained large model for calculation to obtain a result of whether a scene switching is required.

4. The method according to claim 3, characterized in that If the similarity of the switching scene information does not exceed the second scene switching threshold, there is no need to perform scene switching.

5. The method according to claim 2, characterized in that: After switching to the new target scene, the method further includes: Determine whether the new target scene interaction link is a basic interaction link; If yes, determining whether all the cached information includes the interaction information corresponding to the basic interaction link; If yes, skip this basic interaction step.

6. The method according to claim 1, characterized in that Judging whether a link needs to be switched according to the vector of the target link interaction information and the link knowledge base further includes: Performing vector retrieval from the link knowledge base according to the vector of the target link interaction information to obtain the switching link information with the highest matching degree and the approximation of the switching link information; Determine whether the similarity of the switching link information exceeds the link switching threshold; If yes, it is determined whether the link corresponding to the switching link information is the same as the target link. If so, there is no need to switch links; If not, it is necessary to switch the link, and use the link corresponding to the switching link information as the new target link.

7. The method according to claim 6, characterized in that If the similarity of the switching link information does not exceed the link switching threshold, the method further includes: Stop the interaction of the target link, generate the minimum interaction script by the pre-trained large model and execute the interaction.

8. The method according to claim 1, characterized in that If the similarity of the first-round scene information does not exceed the first-round scene threshold, the method further includes: The pre-trained large model generates the minimum interactive script and executes the interaction.

9. A customer service system intention recognition device based on a large model, characterized in that: The device comprises: An interaction information acquisition unit, used to acquire the first-round interaction information of the user, and vectorize the first-round interaction information to obtain a first-round interaction information vector; A scene knowledge base retrieval unit, configured to perform vector retrieval from a scene knowledge base according to the first-round interaction information vector, to obtain the first-round scene information with the highest matching degree and the approximation of the first-round scene information, wherein the scene knowledge base pre-stores vectors corresponding to multiple scenes; An interaction unit, configured to, when the similarity of the first-round scenario information exceeds a first-round scenario threshold, use the scenario corresponding to the first-round scenario information as a target scenario, and call a process engine, wherein the process engine calls up the target scenario, and starts to perform interaction with the user according to the steps corresponding to the target scenario; Wherein, in each round of link interaction of the target scene, the current interactive link is taken as the target link, and whether the scene needs to be switched is determined according to the vector of the target link interaction information of the user and the scene knowledge base; If yes, switch to the new target scene and execute the step of calling the new target scene by the process engine; If not, judging whether the link needs to be switched according to the vector of the target link interaction information and the link knowledge base, wherein the link knowledge base pre-stores vectors corresponding to multiple links; If yes, switch to a new target link, and perform interaction with the user according to the new target link; If not, continue with the target link and perform the interaction.

10. A computer device comprising a memory, a processor, and a computer program stored in the memory, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 8 is implemented.

Images