Data processing method, device, electronic device and storage medium

Abstract

An embodiment of the present application proposes a data processing method, device, electronic device and computer storage medium, the method comprising: obtaining an intention feedback data set of at least one user, the at least one user including a first user, the intention feedback data set comprising at least one intention feedback data, each intention feedback data in the at least one intention feedback data representing: user reply information to a response result of a previous user request; identifying a first intention of the first user, the first intention representing an intention corresponding to a current user request; obtaining target intention feedback data related to the first intention from the intention feedback data set; and generating response information for the current user request, the response information comprising the target intention feedback data.

Description

Technical Field

[0001] The present application relates to the field of natural language processing, and in particular to a data processing method, device, electronic device and computer storage medium. Background Art

[0002] The Natural Language Understanding (NLU) module is the core module in the intelligent response system. The main function of the NLU module is to identify user intentions and obtain NLU recognition results. Other modules provide guidance or responses based on the NLU recognition results.

[0003] In the related art, when a user interacts with an intelligent answering system, the intelligent answering system usually divides the answer types into two types based on the NLU recognition results: the first type is a normal answer with precise intent, and the second type is a clarifying question with ambiguous intent; wherein, the clarifying question is due to the fact that the user's intent is not clearly expressed, resulting in ambiguity in the NLU recognition intent, and then, based on each intent in the possible intent list, a clarifying question is asked. However, for the two different answer types, the system answer given by the intelligent answering system based on the NLU recognition results is fixed; this is because the recognition effect of the recognition model in the intelligent answering system is fixed after training; therefore, the intelligent answering system answers and asks questions based on the recognition results output by the recognition model, so that the effects of the answers and questions are also relatively fixed, and it is impossible to dynamically and specifically adjust the answer and question strategies according to user feedback. Summary of the invention

[0004] The present application provides a data processing method, device, electronic device and computer storage medium.

[0005] The technical solution of this application is implemented as follows:

[0006] The present application provides a data processing method, the method comprising:

[0007] Acquire an intention feedback data set of at least one user, the at least one user including a first user, the intention feedback data set including at least one intention feedback data, each intention feedback data in the at least one intention feedback data representing: user reply information for a response result of a previous user request;

[0008] identifying a first intention of the first user, where the first intention indicates an intention corresponding to a current user request;

[0009] Acquire target intent feedback data related to the first intent from the intent feedback data set;

[0010] Generate response information for the current user request, wherein the response information includes the target intention feedback data.

[0011] In some embodiments, obtaining the intention feedback dataset of at least one user includes:

[0012] Acquire the intention data of each user of the at least one user and the corresponding user intention; the intention data includes: the current user request and the last user request;

[0013] When the user intention corresponding to the current user request and the user intention corresponding to the previous user request are both accurate, and when there is an intention switch between the user intention corresponding to the current user request and the user intention corresponding to the previous user request, an intention feedback data set for the at least one user is generated.

[0014] In some embodiments, generating the at least one user's intention feedback dataset includes:

[0015] Determine the feedback type of the user intent corresponding to the current user request;

[0016] According to the feedback type, the user intent corresponding to the current user request is placed in the intent list of the user intent corresponding to the previous user request, and the heat value of the user intent corresponding to the current user request is controlled to increase;

[0017] Generate an intention feedback data set of the at least one user based on the intention list and the heat value.

[0018] In some embodiments, the intent list includes: a clarifying rhetorical question intent list, a refined guidance intent list or a progressive guidance intent list; wherein, the clarifying rhetorical question intent list includes the intent after correction of the user intent corresponding to the last user request; the refined guidance intent list includes the intent after supplementation of the user intent corresponding to the last user request; the progressive guidance intent list includes the intent after progressive guidance of the user intent corresponding to the last user request.

[0019] In some embodiments, placing the user intent corresponding to the current user request into an intent list of the user intent corresponding to the previous user request according to the feedback type includes:

[0020] In the case that the previous user request does not have a business word, and the current user request has a business word, the user intent corresponding to the current user request is put into a clarification question intent list;

[0021] In the case where the service word requested by the last user is at least partially the same as the service word requested by the current user, the user intent corresponding to the current user request is put into a refined guidance intent list;

[0022] In the case that the service word requested by the user last time is completely different from the service word requested by the user this time, the user intent corresponding to the user request this time is put into the progressive guidance intent list.

[0023] In some embodiments, generating response information of the current user request includes:

[0024] Determining a recognition result type of the first intention;

[0025] Acquire an intent list of the first intent according to the recognition result type;

[0026] Generate response information for the current user request based on the current user request and the intent list of the first intent.

[0027] In some embodiments, generating response information of the current user request according to the current user request and the intent list of the first intent includes:

[0028] When the recognition result type is accurate intent, obtaining a heat value of each intent in the intent list of the first intent;

[0029] According to the heat value and the intention similarity, the response information of the current user request is generated; the intention similarity represents the similarity between each intention in the intention list and the current user request.

[0030] In some embodiments, obtaining the heat value of each intent in the intent list of the first intent includes:

[0031] When the clarification and questioning intention list of the first intention satisfies a first set condition, obtaining a heat value of each intention in the clarification and questioning intention list;

[0032] When the clarification question intention list of the first intention does not meet the first set condition and the refined guidance intention list meets the second set condition, obtaining the heat value of each intention in the refined guidance intention list;

[0033] When the refined guidance intention list of the first intention does not satisfy the second set condition and the progressive guidance intention list satisfies the third set condition, the heat value of each intention in the progressive guidance intention list is obtained.

[0034] In some embodiments, generating response information of the current user request according to the heat value and the intention similarity includes:

[0035] Calculate the weight of each intent in the intent list according to the heat value and the intent similarity;

[0036] A first target intent whose weight is greater than a set threshold is obtained, and response information of the current user request is generated according to the first target intent.

[0037] In some embodiments, generating response information of the current user request according to the current user request and the intent list of the first intent includes:

[0038] In the case where the recognition result type is intent ambiguity, traversing each approximate intent in a possible intent list; the possible intent list includes a predetermined approximate intent of the first intent;

[0039] Obtain the heat value of each intent in the clarification and questioning intent list of each similar intent;

[0040] Response information of the current user request is generated according to the heat value and the intention similarity, wherein the intention similarity indicates the similarity between each intention in the clarification question intention list and the current user request.

[0041] In some embodiments, generating response information of the current user request according to the heat value and the intention similarity includes:

[0042] Calculate the weight of each intent in the clarification question intent list according to the heat value and intent similarity;

[0043] A second target intent whose weight is greater than a set threshold is obtained, and response information of the current user request is generated according to the second target intent.

[0044] The embodiment of the present application further provides a data processing device, which includes a first acquisition module, an identification module, a second acquisition module and a generation module, wherein:

[0045] A first acquisition module is used to acquire an intention feedback data set of at least one user, wherein the at least one user includes a first user, and the intention feedback data set includes at least one intention feedback data, wherein each intention feedback data in the at least one intention feedback data represents: user reply information for a response result of a previous user request;

[0046] an identification module, configured to identify a first intention of the first user, where the first intention indicates an intention corresponding to a current user request;

[0047] A second acquisition module, configured to acquire target intent feedback data related to the first intent from the intent feedback data set;

[0048] A generation module is used to generate response information for the current user request, wherein the response information includes the target intention feedback data.

[0049] An embodiment of the present application provides an electronic device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the data processing method provided by one or more of the aforementioned technical solutions when executing the program.

[0050] An embodiment of the present application provides a computer storage medium, which stores a computer program; after the computer program is executed, it can implement the data processing method provided by one or more of the above-mentioned technical solutions.

[0051] The embodiment of the present application proposes a data processing method, device, electronic device and computer storage medium, the method comprising: obtaining an intention feedback data set of at least one user, the at least one user including a first user, the intention feedback data set including at least one intention feedback data, each intention feedback data in the at least one intention feedback data represents: user reply information for the response result of the last user request; identifying the first intention of the first user, the first intention represents the intention corresponding to the current user request; obtaining target intention feedback data related to the first intention from the intention feedback data set; generating response information for the current user request, the response information including the target intention feedback data. In this way, without modifying the recognition model, an intention feedback data set is generated by collecting the intentions of user feedback, and the response and guidance strategies are dynamically adjusted based on the intention feedback data set to guide the user to obtain a more refined response or a progressive response. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 is a flow chart of a data processing method in an embodiment of the present application;

[0053] Figure 2 A schematic diagram of the structure of a data processing system according to an embodiment of the present application;

[0054] Figure 3 A flowchart of another data processing method in an embodiment of the present application;

[0055] Figure 4 A schematic diagram of the structure of a data processing device according to an embodiment of the present application;

[0056] Figure 5 A schematic diagram of the structure of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0057] The present application will be further described in detail below in conjunction with the accompanying drawings and Examples. It should be understood that the embodiments provided herein are only used to explain the present application and are not intended to limit the present application. In addition, the embodiments provided below are partial embodiments for implementing the present application, rather than providing all embodiments for implementing the present application. In the absence of conflict, the technical solutions recorded in the embodiments of the present application can be implemented in any combination.

[0058] It should be noted that, in the embodiments of the present application, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a method or apparatus including a series of elements includes not only the elements explicitly recorded, but also includes other elements not explicitly listed, or also includes elements inherent to the implementation of the method or apparatus. In the absence of further restrictions, an element defined by the sentence "includes a ..." does not exclude the presence of other related elements (such as steps in a method or units in a device, for example, a unit may be a portion of a circuit, a portion of a processor, a portion of a program or software, etc.) in the method or apparatus including the element.

[0059] The term "and / or" herein is only a description of the association relationship of the associated objects, indicating that there may be three relationships. For example, I and / or J can represent the three situations that I exists alone, I and J exist at the same time, and J exists alone. In addition, the term "at least one" herein represents any combination of at least two of any one or more of a plurality of. For example, including at least one of I, J, and R can represent including any one or more elements selected from the set consisting of I, J, and R.

[0060] For example, the data processing method provided in the embodiment of the present application includes a series of steps, but the data processing method provided in the embodiment of the present application is not limited to the recorded steps. Similarly, the data processing device provided in the embodiment of the present application includes a series of modules, but the data processing device provided in the embodiment of the present application is not limited to including the modules explicitly recorded, and may also include modules required to obtain relevant time series data or perform processing based on time series data.

[0061] The embodiments of the present application can be applied to a computer system composed of a terminal device and a server, and can operate with many other general or special computing system environments or configurations. Here, the terminal device can be a thin client, a thick client, a handheld or laptop device, a microprocessor-based system, a set-top box, a programmable consumer electronic product, a network personal computer, a small computer system, etc., and the server can be a server computer system, a small computer system, a large computer system, and a distributed cloud computing technology environment including any of the above systems, etc.

[0062] Electronic devices such as terminal devices and servers can be described in the general context of computer system executable instructions (such as program modules) executed by a computer system. In general, program modules can include routines, programs, object programs, components, logic, data structures, etc., which perform specific tasks or implement specific abstract data types. Computer systems / servers can be implemented in a distributed cloud computing environment, where tasks are performed by remote processing devices linked through a communication network. In a distributed cloud computing environment, program modules can be located on local or remote computing system storage media including storage devices.

[0063] The intelligent response system is one of the core technologies of human-computer interaction, and its common application scenario is mainly intelligent customer service. By building an intelligent response system, the robot can quickly find answers that match user questions and automatically answer them, thereby greatly reducing labor costs; further, in addition to providing professional question-and-answer functions, the intelligent response system can also provide answering services such as chatting and personal assistants.

[0064]

[0065] Table 1

[0066] In the related art, when a user interacts with an intelligent response system, the response types of the intelligent response system are generally divided into two types: one is a normal response with precise intention, and the other is a clarifying question with ambiguous intention; as shown in Table 1, for the user question "I want to modify the order", the response type is a normal response with precise intention; for the user question "I spent 75 yuan, how can I get a refund of 69 yuan", the response type is a clarifying question with ambiguous intention.

[0067] In the related art, the interaction between users and intelligent response systems mainly has the following disadvantages:

[0068] 1) In the related art, when the user's intention is unclear, once the NLU recognition result is accurate, the clarification question will not be triggered; that is, in some cases, the user did not express a clear business, but the recognition result is accurate, and a fixed system response is given directly; however, the system response is inaccurate, which will lead to a poor user experience. Here, Table 2 illustrates this situation. It can be seen that for the user question "My address is wrong, I need to change it", the NLU recognition result does not distinguish whether the order address is wrong, the invoice address is wrong, or the account address is wrong; if the user does not give feedback or supplement the business afterwards, it means that the system response is correct, but if the user gives feedback or supplements the business after the response, it means that the system response is inaccurate, which will lead to a poor user experience.

[0069]

[0070] Table 2

[0071] 2) In the related art, the clarification questions given by the intelligent answering system are not targeted; that is, they can only be asked clarification questions for each intent in the possible intent list obtained by the recognition model. Since the possible intent list is only a collection of multiple approximate intents, and the possible intent list is basically unchanged, the clarification questions cannot be adjusted in a targeted manner; here, Table 3 is used to illustrate this situation. It can be seen that for the user question "My address is wrong", there is no targeted question about the "address" related intent in the possible intent list obtained by the recognition model; and this possible intent list is relatively fixed. If the user feedbacks other services, the question cannot be adjusted dynamically.

[0072]

[0073]

[0074] Table 3

[0075] 3) In the related art, the intelligent answering system cannot dynamically guide the user to obtain a more detailed answer or a progressive answer after answering; and the guidance on the answer is mainly based on a manually configured answer knowledge base, which is also relatively fixed; that is, it cannot be dynamically adjusted based on user feedback, and the improvement in user experience is limited.

[0076] In the related technologies, if the above technical problems are to be solved, human intervention is required again; for example, manually annotating the corpus and adjusting the recognition model and retraining it; for example, manually adjusting the answer knowledge base; in this way, although the above technical problems can be fixed to a certain extent; however, this repair measure has a long repair cycle, limited effect and high cost.

[0077] In view of the above technical problems, the following embodiments are proposed.

[0078] In some embodiments of the present application, the data processing method can be implemented using a processor in a data processing device, and the above-mentioned processor can be at least one of an application specific integrated circuit (ASIC), a digital signal processor (DSP), a digital signal processing device (DSPD), a programmable logic device (PLD), a field programmable gate array (FPGA), a central processing unit (CPU), a controller, a microcontroller, and a microprocessor.

[0079] Figure 1 is a flow chart of a data processing method in an embodiment of the present application, and the data processing method can be applied to an intelligent response system, such as Figure 1 As shown, the method comprises the following steps:

[0080] Step 100: Obtain an intention feedback data set of at least one user, the at least one user including a first user, the intention feedback data set including at least one intention feedback data, each intention feedback data in the at least one intention feedback data represents: user reply information for the response result of the last user request.

[0081] In an embodiment of the present application, at least one user includes each user who consults a question through the intelligent answering system; wherein the intelligent answering system can display the interface in the form of a dialog box, and the user can directly enter the question to be consulted in the dialog box, or click on some links displayed in the dialog box to consult the question; here, there is no limitation on the application field of the intelligent answering system.

[0082] In some embodiments, there is no restriction on the type of questions asked by users; assuming an e-commerce customer service intelligent response system as an example, the questions asked by users may be questions related to product information or questions related to logistics information.

[0083] In an embodiment of the present application, the intent feedback data set includes the intent feedback data of each user; wherein the intent feedback data represents the current consultation question that the user provides feedback on based on the answer result of the previous consultation question.

[0084] In one embodiment, the question consultation input by each user is collected in real time, and it is determined whether the question consultation input by each user meets the set conditions. If the set conditions are met, an intention feedback data set of at least one user is obtained.

[0085] In some embodiments, obtaining an intent feedback data set for at least one user may include: obtaining intent data and corresponding user intent of each user among the at least one user; the intent data includes: this user request and the last user request; when the user intent corresponding to the current user request and the user intent corresponding to the last user request are both accurate, and there is an intent switch between the user intent corresponding to the current user request and the user intent corresponding to the last user request, generating an intent feedback data set for at least one user.

[0086] In an embodiment of the present application, the current user request represents the current user's request for consulting a question through the intelligent answering system this time; the last user request represents the current user's request for consulting a question through the intelligent answering system last time; by obtaining the intention data and the corresponding user intention of each user in turn, and determining that when the intention data and the corresponding user intention of each user satisfy the above conditions, the intention feedback data set is updated, and then the corresponding intention feedback data set is generated.

[0087] In some embodiments, the user intent corresponding to the intent data of each user can be identified through the NLU in the intelligent answering system; after the NLU identification is completed, it is determined whether the current user's question consultation is the first input. If the current user's question consultation is the first input, that is, when the intent data only exists for this user request but not for the last user request, it means that this user request is not the user's intent feedback data; therefore, the current user's intent feedback data set cannot be generated, and the update process of the intent feedback data set is terminated.

[0088] In some embodiments, if user 1's current user request is "I want to modify the order address"; and user 1's last user request was "I want to modify the order"; it can be seen that this user request "I want to modify the order address" is a further feedback service to the last user request "I want to modify the order"; that is, user 1's current user request is not the first input; if user 1 has not consulted on other business issues before this user request, it means that user 1's current user request is the first input and is not a further feedback service from user 1; in this case, user 1's feedback intention cannot be obtained, and therefore, the update process of the intent feedback data set is terminated.

[0089] In some embodiments, it is determined whether the user intent corresponding to the current user request and the user intent corresponding to the previous user request are both accurate; if any of the user intent corresponding to the current user request and the user intent corresponding to the previous user request is inaccurate, the update process of the intent feedback data set is terminated to ensure the accuracy of the intent feedback data set.

[0090] In some embodiments, if the current user request of user 2 is "I want to modify the order address", and the last user request of user 2 is "I want to modify the address"; wherein the intention of the current user request of user 2 is "modify the order address", the recognition result type is precise intent; the intention of the last user request of user 2 may be "modify the order address" or "modify the account address", and the recognition result type is fuzzy intent; since the intention of the last user request of user 2 is not precise, the accuracy of the intent feedback data set cannot be guaranteed, and the update process of the intent feedback data set is terminated.

[0091] In some embodiments, it is determined whether there is an intention switch between the user intent corresponding to the current user request and the user intent corresponding to the previous user request; if the intention switches, the intention corresponding to the subsequent user input request can be guided and questioned based on the switched intention; if the intention does not switch, it means that the user intent corresponding to the current user request is the same as the user intent corresponding to the previous user request, and it is meaningless to guide and question the intention corresponding to the subsequent user input request based on the same intention, and the update process of the intent feedback data set is ended.

[0092] In some embodiments, if the current user request of user 3 is "I want to modify the order product information", and the last user request of user 3 is "I want to modify the order information"; wherein, the intention of the current user request of user 3 is "modify the order product"; the intention of the last user request of user 3 is "modify the order"; it can be seen that the intention of the current user request of user 3 and the last user request has switched. In this case, the intention feedback data set is updated, and the intention of other users inputting related requests can be guided and questioned based on the switched intention in the intention feedback data set.

[0093] In some embodiments, generating an intent feedback data set for at least one user may include: determining the feedback type of the user intent corresponding to the current user request; according to the feedback type, placing the user intent corresponding to the current user request into an intent list of the user intent corresponding to the previous user request, and controlling the increase in the heat value of the user intent corresponding to the current user request; generating an intent feedback data set for at least one user based on the intent list and the heat value.

[0094] In the embodiment of the present application, the feedback type of the user intent corresponding to the current user request includes: correction of the user intent corresponding to the previous user request, supplement of the user intent corresponding to the previous user request, or progressive guidance of the user intent corresponding to the previous user request. The feedback type of the user intent corresponding to the current user request can be determined based on the relationship between the user intent corresponding to the current user request and the user intent corresponding to the previous user request.

[0095] In some embodiments, a business word list can be obtained in advance; the business word list mainly includes keywords that can distinguish the business scope to which the intent corresponding to the user's problem consultation request belongs; here, the business word list can be determined by manual combing or by other means, and the embodiments of the present application are not limited.

[0096] In some embodiments, the business words in the business word list may include: order, invoice, order address, invoice address, account address, order delivery time, order goods, after-sales, return time, balance, billing record, etc.

[0097] In some embodiments, the current user request and the previous user request may be segmented using a segmentation tool or a segmentation algorithm, the segmentation results of the current user request and the previous user request may be obtained, and the business words may be obtained from the segmentation results.

[0098] In some embodiments, when the business words of the current user request and the previous user request are not obtained in the word segmentation results, it will be impossible to determine the feedback type of the user intent corresponding to the current user request, and the update process of the intent feedback data set will be terminated.

[0099] In some embodiments, the word segmentation tool or word segmentation algorithm used can be set according to the actual application scenario, and the embodiments of the present application are not limited thereto; for example, it can be the pkuseg word segmentation tool, the Jieba word segmentation algorithm, etc.

[0100] In some embodiments, based on the feedback type, the user intent corresponding to the current user request is placed in the intent list of the user intent corresponding to the last user request, which may include: when the last user request did not have a business word, but the current user request has a business word, the user intent corresponding to the current user request is placed in the clarification and questioning intention list (Rule 1); when the business word of the last user request is at least partially the same as the business word of the current user request, the user intent corresponding to the current user request is placed in the refined guidance intention list (Rule 2); when the business word of the last user request is completely different from the business word of the current user request, the user intent corresponding to the current user request is placed in the progressive guidance intention list (Rule 3).

[0101] In some embodiments, the intent list may include: a clarifying rhetorical question intent list, a refined guiding intent list, or a progressive guiding intent list; wherein the clarifying rhetorical question intent list includes the intent after correction of the user intent corresponding to the last user request; the refined guiding intent list includes the intent after supplementation of the user intent corresponding to the last user request; and the progressive guiding intent list includes the intent after progressive guidance of the user intent corresponding to the last user request.

[0102] In some embodiments, if the intent of the last user request was "modify order address", and the intent of this user request is "modify invoice address"; since "order address" and "invoice address" respectively represent different addresses, that is, the intent of this user request is different from the intent of this user request, it can be seen that the intent of "modify invoice address" is the intention after correcting the intent of "modify order address"; then, the intent of this user request "modify invoice address" is put into the clarification and questioning intention list of the intent of the last user request "modify order address".

[0103] In some embodiments, if the intent of the last user request was "modify order", and the intent of this user request is "modify order delivery time"; since an order may include information such as order address, order items, order delivery time, etc., it can be seen that "order delivery time" is a further supplement to "order", and thus, the intent of this user request "modify order delivery time" is placed in the refined guidance intention list of the intent of the last user request "modify order".

[0104] In some embodiments, if the intent of the last user request was "after-sales processing progress and cycle", the intent of this user request is "after-sales return time"; since "after-sales return time" is other intents that are progressively guided based on "after-sales processing progress and cycle"; then, the intent of this user request, "after-sales return time", is placed in the progressively guided intent list of the intent of the last user request, "after-sales processing progress and cycle".

[0105] In some embodiments, each user intent has a corresponding intent list; wherein the heat value of the user intent corresponding to the current user request corresponds to a counter; when the user intent corresponding to the current user request is placed in the intent list of the user intent corresponding to the previous user request, the count of the corresponding counter is controlled to increase accordingly; based on the intent list and the heat value, a user intent feedback data set is generated. The process of generating a user intent feedback data set is described below through Table 4.

[0106] As can be seen from Table 4, when the last user request "My address is wrong, I need to change it" has no business words, and the current user request "I mean the invoice address is wrong, I need to change the invoice" has the business words "invoice address" and "invoice", Rule 1 is triggered, and the user intent "modify invoice address" corresponding to the current user request is put into the clarification question intent list of the user intent "modify order address" corresponding to the last user request, and the heat value of the user intent "modify invoice address" corresponding to the current user request is controlled to add 1 to the original value N; where N is an integer greater than or equal to 0.

[0107] When the last user request "I want to change the order information" contains the business word "order", and the current user request "How to change the product information in the order" contains the business words "order" and "product", rule 2 is triggered, and the user intent "modify order products" corresponding to the current user request is placed in the refined guidance intention list of the user intent "modify order" corresponding to the last user request, and the heat value of the user intent "modify order products" corresponding to the current user request is controlled to add 1 to the original value N; where N is an integer greater than or equal to 0.

[0108] When the last user request "I want to query the after-sales processing progress" contains the business terms "after-sales" and "processing progress", and the current user request "No one comes to pick up the after-sales goods" contains the business term "after-sales goods", rule 3 is triggered, and the user intent "after-sales return time" corresponding to the current user request is placed in the progressive guidance intent list of the user intent "after-sales processing progress and cycle" corresponding to the last user request, and the heat value of the user intent "after-sales return time" corresponding to the current user request is controlled to add 1 to the original value N; where N is an integer greater than or equal to 0.

[0109]

[0110]

[0111] Table 4

[0112] In the intent feedback dataset, each user intent is classified and put into three types of lists corresponding to the intent list: clarification and questioning intent list, refinement and guidance intent list, and progressive guidance intent list; one or more of the three types of lists for each intent classification may be empty; each intent in the list has a real-time statistical heat value, that is, the number of times the feedback intent is triggered; here, the heat value within a fixed period can be retained, for example, the heat value within two weeks or one week can be retained, so as to ensure the timeliness of the feedback dataset. The following table 5 describes the final generated intent feedback dataset.

[0113]

[0114]

[0115] Table 5

[0116] As can be seen from Table 5, for the user intention of "modify order address", there are two intentions, "modify invoice address" and "modify account address" in the clarification question intention list; and the heat value corresponding to the intention of "modify invoice address" is 630, and the heat value corresponding to the intention of "modify account address" is 272; it is empty in the refined guidance intention list; in the progressive guidance intention list, there is the intention of "cancel order"; and the heat value corresponding to this intention is 1133.

[0117] For the user intention of "modify order", it is empty in the clarification question intention list; in the refined guidance intention list, there are three intentions: "modify order address", "modify order items" and "modify order delivery time"; and the heat value corresponding to the intention of "modify order address" is 2201, the heat value corresponding to the intention of "modify order items" is 821, and the heat value corresponding to the intention of "modify order delivery time" is 2157; in the progressive guidance intention list, there is the intention of "cancel order"; and the heat value corresponding to this intention is 601.

[0118] For the user intent "after-sales processing progress and cycle", it is empty in the clarification question intent list; it is empty in the refined guidance intent list; in the progressive guidance intent list, there is the intent of "after-sales return time"; and the heat value corresponding to this intent is 195.

[0119] Step 101: Identify a first intention of a first user, where the first intention indicates an intention corresponding to a current user request.

[0120] In the embodiment of the present application, the first user represents the user who is currently consulting a question through the intelligent answering system, and the NLU module can be used to identify the user intention of the current user to consult the question and obtain the identification result.

[0121] Among them, NLU intent recognition is for known training corpus; for example, the corpus format is a tuple list in the (x, y) format, where x is the training corpus and y is the expected output category or intent; a recognition model is constructed using the selected algorithm, and then the unknown text is classified based on the constructed recognition model; here, the unknown text refers to the data text of the current user's question consultation.

[0122] In some embodiments, the processing flow of NLU intent recognition is as follows: obtain training data and perform word segmentation in a fixed format; extract required features to form feature vectors; the extracted feature vectors and corresponding expected outputs are used to train the recognition model; the same feature extraction is performed on the newly input data to obtain a feature vector for recognition; use the trained recognition model to recognize the features extracted from the newly input data and return the recognition results.

[0123] Step 102: Obtain target intent feedback data related to the first intent from the intent feedback data set.

[0124] In an embodiment of the present application, after identifying the first intention of the first user, if the generated intent feedback data set stores an intent list of the first intent, the intent list of the first intent is obtained in the intent feedback data set; wherein, the target intent feedback data represents the intent data in the intent list of the first intent.

[0125] In some embodiments, for the first intent, the corresponding user request in the intent feedback data set may not be exactly the same as the content of the current user request; for example, the current user request is "My address is wrong and needs to be changed"; and the user request corresponding to the first intent in the intent feedback data set may be "I want to modify the order address" or "How to modify the order address"; and the user intents corresponding to these user requests are all "Modify the order address".

[0126] In an embodiment of the present application, when the recognition result type of the first intention is precise intent, the intention list of the first intention includes: a clarification and rhetorical question intention list of the first intention, a refinement and guidance intention list of the first intention, and a progressive guidance intention list of the first intention; wherein, one or two of the three lists included in the first intention may be empty; the acquisition of target intent feedback data is in the following order from high to low priority: a clarification and rhetorical question intention list of the first intention, a refinement and guidance intention list of the first intention, and a progressive guidance intention list of the first intention.

[0127] In one embodiment, for the first intention M, if the clarification question intention list includes intention M01 and intention M02; the refinement guidance intention list is empty, and the progressive guidance intention list includes three intentions: intention M21, intention M22, and intention M23; then first obtain the two intentions of intention M01 and intention M02, and then obtain the three intentions of intention M21, intention M22, and intention M23.

[0128] In the embodiment of the present application, when the recognition result type of the first intention is ambiguous intention, the intention list of the first intention is a clarification and interrogation intention list of the first intention.

[0129] In some embodiments, the embodiment of the present application does not limit the method for obtaining the target intention feedback data, and it can be obtained through manual input or other methods.

[0130] Step 103: Generate response information for the current user request, where the response information includes target intention feedback data.

[0131] In some embodiments, the response information represents the response result obtained based on the intention feedback data set, and a corresponding rhetorical question list can be generated according to the response result; the rhetorical question list can supplement the response result given by the intelligent response system itself.

[0132] In some embodiments, generating response information for the current user request may include: determining the recognition result type of the first intent; obtaining an intent list of the first intent based on the recognition result type; and generating response information for the current user request based on the current user request and the intent list of the first intent.

[0133] In an embodiment of the present application, the recognition result types of the first intention are divided into two types: the first is precise intention, and the second is ambiguous intention; wherein, in the case of precise intention, the first intention represents a clear intention; in the case of ambiguous intention, the first intention represents a set of at least one intention similar to the user intention of the first user to consult a question.

[0134] In one embodiment, when the user question is "My address is wrong", if the recognition result is that the intent is accurate, the first intent is a clear intent of "modify the order address"; if the recognition result is that the intent is vague, the first intent is a set of three intents of "modify the order address", "how to cancel the order" and "service order modification".

[0135] In some embodiments, generating response information for the current user request based on the current user request and the intent list of the first intent may include: when the recognition result type is intent accuracy, obtaining the heat value of each intent in the intent list of the first intent; generating response information for the current user request based on the heat value and intent similarity; the intent similarity represents the similarity between each intent in the intent list and the current user request.

[0136] In an embodiment of the present application, when the recognition result type of the first intention is accurate intent, an intention list of the first intention is obtained; here, a clarification and questioning intention list, a refined guidance intention list, or a progressive guidance intention list of the first intention can be obtained in order of priority.

[0137] In some embodiments, obtaining the heat value of each intention in the intention list of the first intention may include: when the clarification and rhetorical question intention list of the first intention satisfies the first set condition, obtaining the heat value of each intention in the clarification and rhetorical question intention list; when the clarification and rhetorical question intention list of the first intention does not satisfy the first set condition and the refined guidance intention list satisfies the second set condition, obtaining the heat value of each intention in the refined guidance intention list; when the refined guidance intention list of the first intention does not satisfy the second set condition and the progressive guidance intention list satisfies the third set condition, obtaining the heat value of each intention in the progressive guidance intention list.

[0138] In an embodiment of the present application, the first setting condition indicates that the clarification and questioning intention list of the first intention is not empty, and there is no business word in the current user request; the second setting condition indicates that the detailed guidance intention list of the first intention is not empty, and the third setting condition indicates that the progressive guidance intention list of the first intention is not empty.

[0139] In some embodiments, generating response information for the current user request based on the popularity value and intent similarity may include: calculating the weight of each intent in the intent list based on the popularity value and intent similarity; obtaining the first target intent whose weight is greater than a set threshold, and generating response information for the current user request based on the first target intent.

[0140] In the embodiment of the present application, the weight of each intent in the intent list is calculated according to formula (1), and the current user requests greater than the set threshold are filtered and obtained.

[0141] f(A)=lg(count(A))*similar(A,T) (1)

[0142] Here, A represents each intent in the intent list, T represents the current user request, count(A) represents the heat value corresponding to intent A, and similar(A,T) represents the similarity between intent A and the current user request T.

[0143] In some embodiments, a similarity model can be called to calculate the similarity between each intent in the intent list and the current user request; here, different similarity models can be called according to the actual application scenario, for example, the Deep Structured Semantic Models (DSSM) can be called to calculate the similarity, or other matching models can be called to calculate the similarity.

[0144] In some embodiments, the intent similarity between each intent in the clarification and rhetorical question intent list of the first intent is first determined, and the weight of each intent in the clarification and rhetorical question intent list is calculated based on the heat value and intent similarity of each intent in the clarification and rhetorical question intent list to obtain the first target intent greater than the set threshold. Here, the setting of the set threshold can be set according to the actual application scenario, and the embodiment of the present application is not limited.

[0145] In some embodiments, if the first target intent is obtained in the clarification and rhetorical question intention list, the corresponding answer is matched in the pre-configured answer knowledge base according to the first target intent, and the response information requested by the current user is generated. If the first target intent is not obtained in the clarification and rhetorical question intention list, the weight of each intent in the refined guidance intention list is calculated to obtain the first target intent that is greater than the set threshold; the response information requested by the current user is generated according to the first target intent. If the first target intent is not obtained in the refined guidance intention list, the weight of each intent in the progressive guidance intention list is calculated to obtain the first target intent that is greater than the set threshold; the response information requested by the current user is generated according to the first target intent.

[0146] The following table 6 illustrates the generation of response information for the current user request based on the first target intent.

[0147]

[0148] Table 6

[0149] It can be seen from Table 6 that since the weights of the intentions "modify invoice address" and "modify account address" in the clarification question intention list of the user's intention "modify order address" are both greater than the set threshold 1, the response information "how to modify invoice address" and "how to modify personal account address" can be generated based on these two intentions; the final pushed system response is a combination of the original push result and the response information.

[0150] In some embodiments, generating response information for the current user request based on the current user request and the intent list of the first intent may include: when the recognition result type is intent ambiguity, traversing each approximate intent in the possible intent list; the possible intent list includes a predetermined approximate intent of the first intent; obtaining the heat value of each intent in the clarification and rhetorical intention list of each approximate intent; generating response information for the current user request based on the heat value and the intent similarity, the intent similarity indicating the similarity between each intent in the clarification and rhetorical intention list and the current user request.

[0151] In some embodiments, generating response information for the current user request based on the popularity value and intent similarity may include: calculating the weight of each intent in the clarification question intent list based on the popularity value and intent similarity; obtaining a second target intent whose weight is greater than a set threshold, and generating response information for the current user request based on the second target intent.

[0152] In the embodiment of the present application, when the clarification question intention list of each approximate intention is not empty and the current user request does not have a business word, the weight of each approximate intention in the possible intention list is calculated according to formula (1), and the second target intention with a weight greater than the set threshold is obtained; the response information of the current user request is generated according to the second target intention. The generation of the response information of the current user request according to the second target intention is described below in Table 7.

[0153]

[0154] Table 7

[0155] As can be seen from Table 7, for the possible intent list including the approximate intents "modify order address", "how to cancel order" and "service order modification", among them, the weights of the intents "modify invoice address" and "modify account address" in the clarification and questioning intent list of the approximate intent "modify order address" are both greater than the set threshold 1, and the response information "how to modify invoice address" and "how to modify personal account address" can be generated according to these two intentions; while the approximate intents "how to cancel order" and "service order modification" have no clarification and questioning intent list, the response information "can cancel order" and "change exchange to return" are directly generated according to the approximate intent; the system response finally pushed is the combination of the original push result and the response information.

[0156] The embodiment of the present application proposes a data processing method, device, electronic device and computer storage medium, the method comprising: obtaining an intention feedback data set of at least one user, the at least one user including a first user, the intention feedback data set including at least one intention feedback data, each intention feedback data in the at least one intention feedback data represents: user reply information for the response result of the last user request; identifying the first intention of the first user, the first intention represents the intention corresponding to the current user request; obtaining target intention feedback data related to the first intention from the intention feedback data set; generating response information of the current user request, the response information including the target intention feedback data. In this way, without modifying the recognition model, the intention feedback data set is generated by collecting the intention of user feedback, and the response and guidance strategy is dynamically adjusted based on the intention feedback data set to ensure that the user obtains a more refined response or progressive response, thereby improving the user experience.

[0157] In order to better reflect the purpose of the present application, further examples are given based on the above embodiments of the present application.

[0158] Figure 2 Schematic diagram of the structure of the data processing system of the embodiment of the present application, such as Figure 2 As shown, the system includes two modules and a data set: a feedback collection module, an intention feedback data set and a dynamic supplementary response module; wherein the feedback collection module is used to collect the intention data of each user and the corresponding user intention, and then generate an intention feedback data set; the intention feedback data set includes each user's intention feedback data such as correction intention, supplementary intention, progressive guidance intention, etc. for the last intention feedback; the dynamic supplementary response module is used to call the target intention data related to the corresponding intention of the current user's problem consultation in the intention feedback data set, and dynamically adjust the system response and guidance strategy of the current user according to the target intention data.

[0159] Figure 3 FIG. 1 is a flow chart of another data processing method in an embodiment of the present application. Figure 3 As shown, the intelligent answering system in the related technology includes four parts: user input, NLU module, answer knowledge base and system response; its processing flow is: the user inputs the question consultation request to the NLU module, the NLU module outputs the intention recognition result of the question consultation request, matches the answer corresponding to the intention recognition result in the answer knowledge base, and generates a system response to the user's question consultation request.

[0160] On this basis, the embodiment of the present application adds a feedback collection module, an intention feedback data set and a dynamic supplementary response module; wherein the feedback collection module can collect data at any stage in the above processing flow; first, obtain the user's current question consultation request, and determine whether the current question consultation request is the first input. If the judgment result is yes, exit the process; if the judgment result is no, obtain the previous question consultation request of the current question consultation request, and determine whether the intentions of the current question consultation request and the previous question consultation request are accurate. If the judgment result is no, exit the process; if the judgment result is yes, determine whether there is an intention switch between the current question consultation request and the previous question consultation request. If the judgment result is no, exit the process; if the judgment result is yes, update the intention feedback data set.

[0161] The dynamic supplementary response module first obtains the recognition result type of the current question consultation request and the corresponding intent; when the recognition result type is accurate intent, the intent feedback dataset is called to generate a guided rhetorical question list; when the recognition result type is ambiguous intent, the intent feedback dataset is called to generate a supplementary guided rhetorical question list; response information is generated according to the intent in the rhetorical question list, and the final pushed system response is a combination of the original push result and the response information.

[0162] Figure 4 Schematic diagram of the structure of the data processing device according to the embodiment of the present application. Figure 4 As shown, the device includes: a first acquisition module 400, an identification module 401, a second acquisition module 402 and a generation module 403, wherein:

[0163] A first acquisition module 400 is used to acquire an intention feedback data set of at least one user, the at least one user including the first user, the intention feedback data set including at least one intention feedback data, each intention feedback data in the at least one intention feedback data represents: user reply information for a response result of a previous user request;

[0164] An identification module 401 is used to identify a first intention of a first user, where the first intention indicates an intention corresponding to a current user request;

[0165] A second acquisition module 402 is used to acquire target intent feedback data related to the first intent from the intent feedback data set;

[0166] The generating module 403 is used to generate the response information of the current user request, and the response information includes the target intention feedback data.

[0167] In some embodiments, the first acquisition module 400 is used to acquire an intention feedback dataset of at least one user, including:

[0168] Obtaining intent data and corresponding user intent of each user of at least one user; the intent data includes: current user request and last user request;

[0169] When the user intent corresponding to the current user request and the user intent corresponding to the previous user request are both accurate, and when there is an intention switch between the user intent corresponding to the current user request and the user intent corresponding to the previous user request, generate at least one user intent feedback data set.

[0170] In some embodiments, the first acquisition module 400 is used to generate at least one user's intention feedback data set, including:

[0171] Determine the feedback type of the user intent corresponding to this user request;

[0172] According to the feedback type, the user intent corresponding to the current user request is placed in the intent list of the user intent corresponding to the previous user request, and the heat value of the user intent corresponding to the current user request is controlled to increase;

[0173] Generate at least one user's intent feedback dataset based on the intent list and the heat value.

[0174] In some embodiments, the intent list includes: a clarifying rhetorical question intent list, a refined guiding intent list or a progressive guiding intent list; wherein, the clarifying rhetorical question intent list includes the intent after correction of the user intent corresponding to the last user request; the refined guiding intent list includes the intent after supplementation of the user intent corresponding to the last user request; the progressive guiding intent list includes the intent after progressive guidance of the user intent corresponding to the last user request.

[0175] In some embodiments, the first acquisition module 400 is used to put the user intent corresponding to the current user request into the intent list of the user intent corresponding to the previous user request according to the feedback type, including:

[0176] If the last user request did not have a business word, but the current user request has a business word, the user intent corresponding to the current user request is put into a clarification question intent list;

[0177] If the service word requested by the last user is at least partially the same as the service word requested by the current user, the user intent corresponding to the current user request is put into the refined guidance intent list;

[0178] When the service word requested by the user last time is completely different from the service word requested by the user this time, the user intent corresponding to the user request this time is put into the progressive guidance intent list.

[0179] In some embodiments, the generating module 403 is used to generate the response information of the current user request, including:

[0180] Determine the recognition result type of the first intention;

[0181] According to the recognition result type, obtain the intent list of the first intent;

[0182] Generate response information for the current user request based on the current user request and the intent list of the first intent.

[0183] In some embodiments, the generating module 403 is used to generate the response information of the current user request according to the current user request and the intent list of the first intent, including:

[0184] When the recognition result type is precise intent, obtain the heat value of each intent in the intent list of the first intent;

[0185] Generate response information for the current user request based on the heat value and intent similarity; intent similarity indicates the similarity between each intent in the intent list and the current user request.

[0186] In some embodiments, the generation module 403 is used to obtain the heat value of each intent in the intent list of the first intent, including:

[0187] When the clarification and rhetorical question intention list of the first intention satisfies the first set condition, obtaining the heat value of each intention in the clarification and rhetorical question intention list;

[0188] When the clarification question intention list of the first intention does not meet the first set condition, and the refined guidance intention list meets the second set condition, obtaining the heat value of each intention in the refined guidance intention list;

[0189] When the refined guidance intention list of the first intention does not satisfy the second set condition and the progressive guidance intention list satisfies the third set condition, the heat value of each intention in the progressive guidance intention list is obtained.

[0190] In some embodiments, the generation module 403 is used to generate the response information of the current user request according to the heat value and the intention similarity, including:

[0191] Calculate the weight of each intent in the intent list based on the heat value and intent similarity;

[0192] A first target intent whose weight is greater than a set threshold is obtained, and response information of the current user request is generated according to the first target intent.

[0193] In some embodiments, the generating module 403 is used to generate the response information of the current user request according to the current user request and the intent list of the first intent, including:

[0194] When the recognition result type is intent ambiguity, traverse each approximate intent in the possible intent list; the possible intent list includes the approximate intent of the predetermined first intent;

[0195] Get the heat value of each intent in the clarification and questioning intent list of each similar intent;

[0196] According to the heat value and intent similarity, the response information of the current user request is generated. The intent similarity indicates the similarity between each intent in the clarification question intent list and the current user request.

[0197] In some embodiments, the generation module 403 is used to generate the response information of the current user request according to the heat value and the intention similarity, including:

[0198] Calculate the weight of each intent in the clarification question intent list based on the heat value and intent similarity;

[0199] A second target intent whose weight is greater than a set threshold is obtained, and response information of the current user request is generated according to the second target intent.

[0200] In practical applications, the above-mentioned first acquisition module 400, identification module 401, second acquisition module 402 and generation module 403 can all be implemented by a processor located in an electronic device, and the processor can be at least one of ASIC, DSP, DSPD, PLD, FPGA, CPU, controller, microcontroller, and microprocessor.

[0201] In addition, each functional module in this embodiment 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 software functional modules.

[0202] If the integrated unit is implemented in the form of a software function module and is not 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 this embodiment is essentially or the part that contributes to the relevant technology or the whole or 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 for a computer device (which can be a personal computer, server, or network device, etc.) or a processor to perform all or part of the steps of the method of this embodiment. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc. Various media that can store program codes.

[0203] Specifically, a computer program instruction corresponding to a data processing method in this embodiment can be stored on a storage medium such as a CD, a hard disk, or a USB flash drive. When the computer program instruction corresponding to a data processing method in the storage medium is read or executed by an electronic device, any data processing method in the aforementioned embodiments is implemented.

[0204] Based on the same technical concept as the above embodiments, see Figure 5 , which shows an electronic device 500 provided by the present application, which may include: a memory 501 and a processor 502; wherein,

[0205] Memory 501, used for storing computer programs and data;

[0206] The processor 502 is used to execute the computer program stored in the memory to implement any one of the data processing methods in the foregoing embodiments.

[0207] In practical applications, the memory 501 may be a volatile memory, such as RAM; or a non-volatile memory, such as ROM, flash memory, hard disk drive (HDD) or solid-state drive (SSD); or a combination of the above types of memory, and provide instructions and data to the processor 502.

[0208] The processor 502 may be at least one of ASIC, DSP, DSPD, PLD, FPGA, CPU, controller, microcontroller, and microprocessor. It is understandable that for different data processing devices, the electronic device used to implement the processor function may also be other, which is not specifically limited in the embodiments of the present application.

[0209] In some embodiments, the functions or modules included in the apparatus provided in the embodiments of the present application can be used to execute the method described in the above method embodiment. The specific implementation can refer to the description of the above method embodiment. For the sake of brevity, it will not be repeated here.

[0210] The above description of the various embodiments tends to emphasize the differences between the various embodiments. The same or similar aspects can be referenced to each other. For the sake of brevity, this article will not repeat them.

[0211] The methods disclosed in the various method embodiments provided in this application can be arbitrarily combined without conflict to obtain new method embodiments.

[0212] The features disclosed in the various product embodiments provided in this application can be arbitrarily combined without conflict to obtain new product embodiments.

[0213] The features disclosed in the various method or device embodiments provided in this application can be arbitrarily combined without conflict to obtain new method embodiments or device embodiments.

[0214] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of hardware embodiments, software embodiments, or embodiments in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage and optical storage, etc.) that contain computer-usable program code.

[0215] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0216] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0217] The above are only preferred embodiments of the present application and are not intended to limit the protection scope of the present application.

Claims

1. A data processing method, characterized in that: The method comprises: Acquire an intention feedback data set of at least one user, the at least one user including a first user, the intention feedback data set including at least one intention feedback data, each intention feedback data in the at least one intention feedback data representing: user reply information for a response result of a previous user request; identifying a first intention of the first user, where the first intention indicates an intention corresponding to a current user request; Acquire target intent feedback data related to the first intent from the intent feedback data set; Generate response information for the current user request, wherein the response information includes the target intention feedback data.

2. The method according to claim 1, characterized in that The obtaining of at least one user's intention feedback dataset includes: Acquire the intention data of each user of the at least one user and the corresponding user intention; the intention data includes: the current user request and the last user request; When the user intention corresponding to the current user request and the user intention corresponding to the previous user request are both accurate, and when there is an intention switch between the user intention corresponding to the current user request and the user intention corresponding to the previous user request, an intention feedback data set for the at least one user is generated.

3. The method according to claim 2, characterized in that The generating the intention feedback dataset of the at least one user comprises: Determine the feedback type of the user intent corresponding to the current user request; According to the feedback type, the user intent corresponding to the current user request is placed in the intent list of the user intent corresponding to the previous user request, and the heat value of the user intent corresponding to the current user request is controlled to increase; Generate an intention feedback data set of the at least one user based on the intention list and the heat value.

4. The method according to claim 3, characterized in that The intention list includes: a clarifying rhetorical question intention list, a refined guidance intention list or a progressive guidance intention list; wherein, the clarifying rhetorical question intention list includes the intention after correction of the user intention corresponding to the last user request; the refined guidance intention list includes the intention after supplementation of the user intention corresponding to the last user request; the progressive guidance intention list includes the intention after progressive guidance of the user intention corresponding to the last user request.

5. The method according to claim 4, characterized in that The step of placing the user intent corresponding to the current user request into an intent list of the user intent corresponding to the previous user request according to the feedback type includes: In the case that the previous user request does not have a business word, and the current user request has a business word, the user intent corresponding to the current user request is put into a clarification question intent list; In the case where the service word requested by the last user is at least partially the same as the service word requested by the current user, the user intent corresponding to the current user request is put into a refined guidance intent list; In the case that the service word requested by the user last time is completely different from the service word requested by the user this time, the user intent corresponding to the user request this time is put into the progressive guidance intent list.

6. The method according to claim 1, characterized in that The generating the response information of the current user request includes: Determining a recognition result type of the first intention; According to the recognition result type, obtaining an intent list of the first intent; Generate response information for the current user request based on the current user request and the intent list of the first intent.

7. The method according to claim 6, characterized in that The generating, according to the current user request and the intent list of the first intent, response information of the current user request includes: When the recognition result type is accurate intent, obtaining a heat value of each intent in the intent list of the first intent; According to the heat value and the intention similarity, the response information of the current user request is generated; the intention similarity represents the similarity between each intention in the intention list and the current user request.

8. The method according to claim 7, characterized in that The obtaining the heat value of each intent in the intent list of the first intent includes: When the clarification and rhetorical question intention list of the first intention satisfies a first set condition, obtaining a heat value of each intention in the clarification and rhetorical question intention list; When the clarification and questioning intention list of the first intention does not meet the first set condition, and the refined guidance intention list meets the second set condition, obtaining the heat value of each intention in the refined guidance intention list; When the clarification and questioning intention list of the first intention does not meet the first set condition, the refined guidance intention list of the first intention does not meet the second set condition, and the progressive guidance intention list meets the third set condition, the heat value of each intention in the progressive guidance intention list is obtained.

9. The method according to claim 7, characterized in that: The step of generating response information for the current user request according to the heat value and the intention similarity includes: Calculate the weight of each intent in the intent list according to the heat value and the intent similarity; A first target intent whose weight is greater than a set threshold is obtained, and response information of the current user request is generated according to the first target intent.

10. The method according to claim 6, characterized in that The generating, according to the current user request and the intent list of the first intent, response information of the current user request includes: In the case where the recognition result type is intent ambiguity, traversing each approximate intent in a possible intent list; the possible intent list includes a predetermined approximate intent of the first intent; Obtain the heat value of each intent in the clarification and questioning intent list of each similar intent; Response information of the current user request is generated according to the heat value and the intention similarity, wherein the intention similarity indicates the similarity between each intention in the clarification question intention list and the current user request.

11. The method according to claim 10, characterized in that The step of generating response information for the current user request according to the heat value and the intention similarity includes: Calculate the weight of each intent in the clarification question intent list according to the heat value and intent similarity; A second target intent whose weight is greater than a set threshold is obtained, and response information of the current user request is generated according to the second target intent.

12. A data processing device, characterized in that: The device comprises: A first acquisition module is used to acquire an intention feedback data set of at least one user, wherein the at least one user includes a first user, and the intention feedback data set includes at least one intention feedback data, wherein each intention feedback data in the at least one intention feedback data represents: user reply information for a response result of a previous user request; an identification module, configured to identify a first intention of the first user, where the first intention indicates an intention corresponding to a current user request; A second acquisition module, configured to acquire target intent feedback data related to the first intent from the intent feedback data set; A generation module is used to generate response information for the current user request, wherein the response information includes the target intention feedback data.

13. An electronic device, characterized in that: The device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, and the processor implements the method according to any one of claims 1 to 11 when executing the program.

14. A computer storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 11 is implemented.

Images