User answering method, device and equipment based on intelligent interaction tool and storage medium

Abstract

The invention discloses a user answering method and device based on an intelligent interaction tool, equipment and a storage medium, and relates to the technical field of artificial intelligence operation, and the method comprises the steps: obtaining a product question sent by a user side for a target product based on the intelligent interaction tool, and generating an initial answer for the product question through a preset large language model; analyzing the product problem based on a preset intention classifier and a preset emotion adjective matcher to obtain a target two-tuple corresponding to the product problem; analyzing an operation behavior of the user side for the initial answer based on a local timer and a preset event burying point to obtain a behavior tag; and if the target two-tuple and the behavior tag form negative feedback information, calling a natural language generation tool to generate an initial question answer, and obtaining a modification instruction of a preset operation and maintenance party for the initial question answer based on a preset interaction interface to determine a target question answer. In this way, on the premise that the model does not need to be trained again, the dialogue answering result can be optimized.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence operation technology, and in particular to a user response method, apparatus, device, and storage medium based on intelligent interactive tools. Background Technology

[0002] In the current wave of enterprise digital transformation, dialogue systems, as the core hub of human-computer interaction, directly impact user experience and business performance. However, existing system optimization methods have significant limitations, making it difficult to promptly identify and address "hidden" experience issues. On one hand, the industry generally relies on semantic analysis, evaluating response quality through indicators such as intent recognition accuracy or sentiment scores. This method not only requires pre-setting highly subjective judgment thresholds, but its sentiment labels are often limited to a binary "positive / negative" division. This crude qualitative judgment cannot reveal the specific reasons for user dissatisfaction and fails to provide operations personnel with intuitive and actionable guidance for modifying dialogue, causing optimization work to often remain superficial. On the other hand, behavioral analysis collects raw log data such as user click behavior and page dwell time, and then performs subsequent statistical processing to obtain macro indicators such as session churn rate. This method has significant lag, long calculation cycles, and overly coarse data granularity, making it impossible to accurately trace back to the specific dialogue round that caused the user experience breakdown. Due to the aforementioned technical shortcomings, enterprises generally find themselves in a dilemma: although the system can "answer" questions correctly—that is, it is factually accurate—users experience poorly in actual interactions. Discovering these hidden bad cases of "answering correctly but having a poor experience" is extremely difficult, and because the root cause is unclear, the subsequent problem localization and repair cycle is also greatly extended, which seriously restricts the continuous improvement of the service quality of the dialogue system and the full realization of its business value.

[0003] Therefore, improving the service quality of dialogue in intelligent interaction tools is a problem that needs to be addressed. Summary of the Invention

[0004] In view of this, the purpose of this invention is to provide a user response method, apparatus, device, and storage medium based on intelligent interactive tools, which can improve the service quality of dialogue in intelligent interactive tools. The specific solution is as follows:

[0005] Firstly, this application discloses a user response method based on intelligent interaction tools, including:

[0006] The system uses intelligent interaction tools to obtain product questions sent by users regarding the target product, and uses a preset large language model to generate initial answers to the product questions.

[0007] The product question is analyzed based on a preset intent classifier and a preset sentiment adjective matcher to obtain the target tuple corresponding to the product question; the target tuple includes an intent tag and a sentiment adjective.

[0008] The user's actions in response to the initial answer are analyzed based on local timers and preset event tracking points to obtain behavior tags;

[0009] If the target tuple and the behavior label constitute negative feedback information, then a natural language generation tool is invoked to generate an initial question answer, and a preset modification instruction from the operation and maintenance party for the initial question answer is obtained based on a preset interactive interface, so as to determine the target question answer for the product problem based on the modification instruction and the initial question answer.

[0010] Optionally, the step of analyzing the product question based on a preset intent classifier and a preset sentiment adjective matcher to obtain the target tuple corresponding to the product question includes:

[0011] The problem generation scenario corresponding to the product problem is analyzed using a preset intent classifier to generate intent tags corresponding to the product problem;

[0012] A preset sentiment adjective matcher performs semantic analysis on the product problem and selects the corresponding sentiment adjective from the preset sentiment adjective library based on the analysis results.

[0013] Construct the target tuple corresponding to the product question based on the intent tag and the sentiment adjective.

[0014] Optionally, before analyzing the user's operational behavior in response to the initial answer based on a local timer and preset event tracking points, the method further includes:

[0015] Track all user behaviors targeting the target product and generate event names and associated attributes corresponding to those user behaviors;

[0016] Based on the event naming and associated attributes, target event tracking instructions are injected into the corresponding code logic of the local client to obtain preset event tracking points.

[0017] Optionally, the analysis of the user's operational behavior in response to the initial answer based on a local timer and preset event tracking points to obtain behavioral tags includes:

[0018] A local timer is used to determine whether the user terminal takes any action in response to the initial response within a preset time period.

[0019] If the user takes an action in response to the initial answer within the preset time period, the action is analyzed using the front-end SDK and preset event tracking points to obtain action tags.

[0020] Optionally, the step of calling a natural language generation tool to generate an initial question answer and obtaining a preset modification instruction from the operations and maintenance party regarding the initial question answer based on a preset interactive interface includes:

[0021] The natural language generation tool is invoked, and a target answer is selected from the historical data repository based on the target tuple to obtain the initial question answer;

[0022] The initial question answer is displayed to the preset operation and maintenance party through a preset visual interface, and the preset operation and maintenance party's modification instructions for the initial question answer are obtained through a preset interactive interface.

[0023] Optionally, after determining the target answer to the product question based on the modification instruction and the initial question answer, the method further includes:

[0024] The target tuple and the answer to the target question are used to generate a response to be stored, and the response to be stored is updated in the historical data repository.

[0025] Optionally, determining the target question answer for the product question based on the modification instruction and the initial question answer includes:

[0026] Target key-value pairs are generated based on the target tuple and the target question answer, where the key in the target key-value pair is the target tuple, and the value in the target key-value pair is the target question answer.

[0027] Update the target key-value pair to the historical data repository.

[0028] Secondly, this application discloses a user response device based on an intelligent interaction tool, comprising:

[0029] The initial response generation module is used to obtain product questions sent by users for the target product based on intelligent interaction tools, and generate initial responses to the product questions using a preset large language model.

[0030] The binary tuple generation module is used to analyze the product question based on a preset intent classifier and a preset sentiment adjective matcher to obtain the target binary tuple corresponding to the product question; the target binary tuple includes an intent tag and a sentiment adjective.

[0031] The behavior analysis module is used to analyze the user's operational behavior in response to the initial answer based on a local timer and preset event tracking points, so as to obtain behavior tags;

[0032] The target answer generation module is used to call a natural language generation tool to generate an initial question answer if the target tuple and the behavior label constitute negative feedback information, and to obtain a preset modification instruction from the operation and maintenance party for the initial question answer based on a preset interactive interface, so as to determine the target question answer for the product problem based on the modification instruction and the initial question answer.

[0033] Thirdly, this application discloses an electronic device, including:

[0034] Memory, used to store computer programs;

[0035] A processor for executing the computer program to implement the aforementioned user response method based on intelligent interactive tools.

[0036] Fourthly, this application discloses a computer-readable storage medium for storing a computer program, which, when executed by a processor, implements the aforementioned user response method based on an intelligent interactive tool.

[0037] As can be seen, the process involves: acquiring product questions from users regarding the target product using intelligent interaction tools; generating initial answers to these questions using a pre-defined large language model; analyzing the product questions using a pre-defined intent classifier and a pre-defined sentiment adjective matcher to obtain target tuples corresponding to the product questions; these target tuples including intent tags and sentiment adjectives; analyzing user actions in response to the initial answers using local timers and pre-defined event tracking to obtain behavior tags; if the target tuples and behavior tags constitute negative feedback, then a natural language generation tool is invoked to generate an initial question answer, and pre-defined modification instructions from the operations and maintenance team regarding the initial question answer are obtained based on a pre-defined interactive interface, so as to determine the target question answer for the product problem based on the modification instructions and the initial question answer. In this way, by analyzing the initial questions sent by users, a tuple containing intent tags and sentiment adjectives is obtained; then, the corresponding actions taken by users in response to the initial answers are collected; the tuples and corresponding actions are comprehensively analyzed; and the corresponding initial question answer is generated based on the analysis results; the initial question answer is adaptively modified based on the modification instructions, ultimately yielding the target question answer for the product problem. In this way, without modifying the preset large language model, targeted answers can be generated based on user feedback, which can improve the efficiency of answering questions and make real-time improvements to the quality of the dialogue. Attached Figure Description

[0038] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0039] Figure 1 This application discloses a flowchart of a user response method based on an intelligent interactive tool.

[0040] Figure 2 This is a schematic diagram of a user response device based on an intelligent interactive tool disclosed in this application;

[0041] Figure 3 This is a structural diagram of an electronic device disclosed in this application. Detailed Implementation

[0042] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0043] Existing dialogue system optimization faces a dual bottleneck: on the semantic side, it relies on preset thresholds for intent recognition or coarse-grained emotion binary labels, which cannot guide specific improvements in dialogue; on the behavioral side, it relies on lagging statistical indicators (such as churn rate), making it difficult to pinpoint the specific dialogue rounds that cause user experience breakdowns. This makes it difficult to discover and fix hidden bad cases of "correct answers but poor experience," severely restricting system service improvement. This application will specifically introduce a user response method based on intelligent interaction tools that can solve the above problems.

[0044] See Figure 1 As shown in the figure, this application discloses a user response method based on an intelligent interaction tool, including:

[0045] Step S11: Obtain product questions sent by users regarding the target product based on intelligent interaction tools, and generate initial answers to the product questions using a preset large language model.

[0046] In this embodiment, the intelligent interaction tool is a customer service robot, an APP (Application) built-in assistant, an intelligent voice terminal, or other products that need to continuously optimize the user experience after going online. Therefore, when a customer (i.e., the user end) sends a corresponding question about the target product, the intelligent interaction tool can obtain the product question sent by the user end about the target product, and then call the preset large language model set in the intelligent interaction tool to generate an initial answer to the product question.

[0047] Step S12: Analyze the product problem based on a preset intent classifier and a preset sentiment adjective matcher to obtain the target tuple corresponding to the product problem; the target tuple includes intent tags and sentiment adjectives.

[0048] In this embodiment, before generating the initial answer, the product question is analyzed based on a preset intent classifier and a preset sentiment adjective matcher to obtain the target tuple corresponding to the product question. This includes: analyzing the question generation scenario corresponding to the product question using the preset intent classifier to generate an intent tag for the product question; performing semantic analysis on the product question using the preset sentiment adjective matcher and selecting the corresponding sentiment adjective from a preset sentiment adjective library based on the analysis results; and constructing the target tuple corresponding to the product question based on the intent tag and the sentiment adjective. For each round of user input, an "intent tag + sentiment adjective" tuple is output, where the adjective comes from an open lexicon (such as "confused," "anxious," "satisfied"), which can be added or deleted in the backend at any time. In other words, sentiment adjectives do not require numerical mapping, the lexicon supports real-time addition and deletion, and the system only performs string matching. The intent tag uses the existing classifier; the sentiment adjective is matched through a preset lexicon + synonym expansion method, requiring no additional model training and can be used immediately upon initial deployment. Specifically, when using a preset intent classifier to analyze the problem generation scenario corresponding to the product problem to generate the intent tag for the product problem, the first step is to obtain the actual scenario corresponding to the user issuing the product problem. For example, whether the user is in a pre-sales or after-sales state; whether the user initiated the product problem through the pre-sales channel or the after-sales channel.

[0049] Step S13: Analyze the user's operational behavior in response to the initial answer based on the local timer and preset event tracking points to obtain behavior tags.

[0050] In this embodiment, before analyzing the user's operational behavior regarding the initial response based on a local timer and preset event tracking points, the method further includes: tracking all user behaviors related to the target product and generating event names and associated attributes corresponding to the user behaviors; and embedding target tracking point instructions into the code logic corresponding to the local client based on the event names and associated attributes to obtain preset event tracking points. Specifically, the SDK (Software Development Kit) first completes initialization, generates a unique identifier, and collects global context information such as device and user information, establishing the foundation for data collection. Subsequently, every key user action within the target product application, whether it's clicking a button, browsing a new page, or completing a payment, will trigger the pre-embedded tracking point code. This code will immediately call the SDK's interface, submitting a standardized event object containing event type and custom attributes to the SDK to obtain preset event tracking points.

[0051] In this embodiment, the analysis of the user's actions on the initial response based on a local timer and preset event tracking points to obtain behavior tags includes: using a local timer to perform timing operations to determine whether the user has taken any action on the initial response within a preset time period; if the user has taken any action on the initial response within the preset time period, the front-end SDK and preset event tracking points are used to analyze the action to obtain behavior tags. Specifically, for actions that require timing, such as page dwell time, the SDK accurately records the start and end timestamps locally and calculates the duration, treating it as a complete duration event attribute. Then, the front-end SDK and preset event tracking points are used to analyze the action to obtain behavior tags. Silence timing is completed by a local countdown on the client side, and network jitter does not affect timing accuracy. In this way, after two rounds of dialogue and actions, each round of user input can be appended with "intent tag + sentiment adjective" to form a semantic wheel, where the sentiment adjective comes from an open lexicon that can be manually maintained; adjacent rounds are appended with three types of behavior tags: "silence / repetition / jump," to form a behavior wheel. Among them, monitoring user actions between adjacent rounds only records three types of markers: silence (>6 seconds without input), repetition (sending the same question again), and exit (leaving the session or transferring to human assistant), without relying on thresholds.

[0052] Step S14: If the target tuple and the behavior label constitute negative feedback information, then the natural language generation tool is invoked to generate an initial question answer, and a preset modification instruction from the operation and maintenance party for the initial question answer is obtained based on the preset interactive interface, so as to determine the target question answer for the product problem based on the modification instruction and the initial question answer.

[0053] In this embodiment, the step of calling a natural language generation tool to generate an initial question answer and obtaining modification instructions from a preset operations and maintenance (O&M) party based on a preset interactive interface includes: calling the natural language generation tool and selecting a target answer from a historical data repository based on the target tuple to obtain the initial question answer; displaying the initial question answer to the preset O&M party through a preset visualization interface, and obtaining modification instructions from the preset O&M party based on the preset interactive interface. Specifically, if the same user uses negative adjectives in two consecutive rounds, or if any round is accompanied by one of the three types of behavioral markers, it is determined to be a "low experience." At this time, a target answer can be selected from the historical data repository to obtain the initial question answer. Then, the operations personnel directly modify the script in natural language on the Web console (i.e., the preset interactive interface), and after clicking "Publish," the system immediately writes the new script into the online reply database, and the original dialogue flow takes effect one minute later.

[0054] In this embodiment, after determining the target question answer for the product question based on the modification instruction and the initial question answer, the method further includes: generating a stored answer from the target tuple and the target question answer, and updating the stored answer to the historical data repository. Specifically, a target key-value pair can be generated based on the target tuple and the target question answer, where the key in the target key-value pair is the target tuple; the value in the target key-value pair is the target question answer; and the target key-value pair is updated to the historical data repository. That is, the historical data repository is hot-updated in key-value format, where the key is a combination of intent tags and behavior tags, and the value is the natural language text edited by operations personnel.

[0055] In addition, this application allows for the determination of a fixed time interval to update the preset large language model using the answers to each question stored in the historical data repository, thereby obtaining a new preset large language model. In this way, the model is continuously updated and evolved within the intelligent interaction tool, which can better improve the user experience.

[0056] In the first specific implementation (e-commerce customer service), the dialogue flow is as follows: ① User: "What's the difference between this one and the previous one?" We can deduce: Intent = Price comparison; Adjective = Confused; Behavior = None. ② Robot: "See above for parameters of both models." We can deduce: Adjective = None. ③ User: "The above is too messy, I don't understand." We can deduce: Intent = Price comparison; Adjective = Dissatisfaction; Behavior = Repetition. The final system action: After ③, if "Confused + Dissatisfied" appears for two consecutive rounds, a micro-script is immediately generated and pushed to the backend. Then, the operations team modifies the response: changing the original reply to a comparison table + a short summary sentence, and clicking publish. The technical effect is: the rate of repeated questions with the same intent the next day decreased from 14% to 4%, without needing to retrain the model.

[0057] In the second specific implementation (APP built-in assistant), the dialogue flow is as follows: ① User: "How do I turn off automatic renewal?" This yields: Intent = Turn off renewal; Adjective = Urgent. ② Robot: "Please click Settings - Membership - Automatic Renewal Switch." This yields: Adjective = None; ③ User's 8-second silence yields: Behavior = Silence. System action: Silence triggers a low point, extracting ①-② to generate a micro-script. Operational modification: Supplement with a graphic guidance GIF (Graphics Interchange Format), effective immediately upon release. Result: The median silence duration at this node decreased from 9 seconds to 3 seconds, and the self-service resolution rate increased by 12 percentage points.

[0058] As can be seen, in this embodiment, a semantic wheel is formed by combining "intent label + sentiment adjective" for each round of user input and output, with the adjectives sourced from an open lexicon that can be maintained manually. A behavior wheel is formed by capturing three types of behavioral markers in adjacent rounds: "silence / repetition / bounce." When negative adjectives appear consecutively in the semantic wheel or any marker appears in the behavior wheel, a low-experience point is immediately identified, and a natural language micro-script is automatically generated. Operators can directly modify the micro-script in natural language on the Web (World Wide Web) console, and it takes effect online within one minute after a one-click release, without needing to retrain the model. The embodiment shows that the system can locate low-experience points within one minute, and the rate of repeated questions with the same intent decreases by more than 10 percentage points the next day. This is suitable for optimizing the online experience of dialogue products such as customer service robots and built-in assistants in apps. In this way, by analyzing the user's initial question, a tuple containing intent label and sentiment adjective is obtained. Then, the user's corresponding actions in response to the initial answer are collected, and the tuple and corresponding actions are comprehensively analyzed. Based on the analysis results, a corresponding initial question answer is generated. The initial question answer is adaptively modified based on modification instructions, ultimately yielding the target question answer for the product issue. In this way, without modifying the preset large language model, targeted answers can be generated based on user feedback, which can improve the efficiency of answering questions and make real-time improvements to the quality of the dialogue.

[0059] refer to Figure 2 The present application also discloses a user response device based on an intelligent interaction tool, comprising:

[0060] The initial answer generation module 11 is used to obtain product questions sent by users for the target product based on intelligent interaction tools, and generate an initial answer for the product questions using a preset large language model.

[0061] The binary tuple generation module 12 is used to analyze the product problem based on a preset intent classifier and a preset sentiment adjective matcher to obtain the target binary tuple corresponding to the product problem; the target binary tuple includes an intent label and a sentiment adjective.

[0062] The behavior analysis module 13 is used to analyze the user's operational behavior in response to the initial answer based on a local timer and preset event tracking points, so as to obtain behavior tags;

[0063] The target answer generation module 14 is used to call a natural language generation tool to generate an initial question answer if the target tuple and the behavior label constitute negative feedback information, and to obtain a preset modification instruction from the operation and maintenance party for the initial question answer based on a preset interactive interface, so as to determine the target question answer for the product problem based on the modification instruction and the initial question answer.

[0064] By analyzing the initial questions posed by users, a pair of tuples containing intent tags and sentiment adjectives is obtained. Then, the user's actions in response to the initial answer are collected. A comprehensive analysis is performed combining the tuples and actions, and a corresponding initial question answer is generated based on the analysis results. The initial question answer is then adaptively modified based on modification instructions to finally obtain the target question answer for the product issue. In this way, without modifying the pre-set large language model, targeted answers can be generated based on user feedback, improving the efficiency of question answering and enabling real-time improvement of dialogue quality.

[0065] In some specific embodiments, the tuple generation module 12 may specifically include:

[0066] The intent recognition unit is used to analyze the problem generation scenario corresponding to the product problem using a preset intent classifier, so as to generate intent tags corresponding to the product problem.

[0067] The sentiment adjective determination unit is used to perform semantic analysis on the product problem by a preset sentiment adjective matcher, and select the sentiment adjective corresponding to the product problem from a preset sentiment adjective library based on the analysis results.

[0068] The binary tuple determination unit is used to construct the target binary tuple corresponding to the product question based on the intent tag and the sentiment adjective.

[0069] In some specific embodiments, the user response device based on intelligent interaction tools may further include:

[0070] The behavior tracking module is used to track all user behaviors targeting the target product and generate event names and associated attributes corresponding to the user behaviors.

[0071] The event tracking module is used to inject target event tracking instructions into the code logic corresponding to the local client based on the event name and associated attributes, so as to obtain preset event tracking points.

[0072] In some specific embodiments, the behavior analysis module 13 may specifically include:

[0073] An operation timing unit is used to perform timing operations using a local timer to determine whether the user terminal has taken any action in response to the initial answer within a preset time period;

[0074] The behavior analysis unit is used to analyze the user's behavior in response to the initial answer within the preset time period using the front-end SDK and preset event tracking points to obtain behavior tags.

[0075] In some specific embodiments, the target answer generation module 14 may specifically include:

[0076] The answer selection unit is used to invoke a natural language generation tool and select a target answer from a historical data repository based on the target tuple to obtain an initial question answer;

[0077] The answer modification unit is used to display the initial question answer to the preset operation and maintenance party based on a preset visual interface, and to obtain the preset operation and maintenance party's modification instructions for the initial question answer based on a preset interactive interface.

[0078] In some specific embodiments, the user response device based on intelligent interaction tools may further include:

[0079] The answer storage module is used to generate a to-be-stored answer from the target tuple and the answer to the target question, and update the to-be-stored answer to the historical data storage repository.

[0080] In some specific embodiments, the target answer generation module 14 may specifically include:

[0081] A key-value pair generation unit is configured to generate target key-value pairs based on the target tuple and the target question answer, wherein the key in the target key-value pair is the target tuple, and the value in the target key-value pair is the target question answer.

[0082] A key-value pair storage unit is used to update the target key-value pair to the historical data repository.

[0083] Furthermore, embodiments of this application also disclose an electronic device, Figure 3 This is a structural diagram of an electronic device 20 according to an exemplary embodiment. The content of the diagram should not be construed as limiting the scope of this application.

[0084] Figure 3This is a schematic diagram of the structure of an electronic device 20 provided in an embodiment of this application. Specifically, the electronic device 20 may include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26. The memory 22 stores a computer program, which is loaded and executed by the processor 21 to implement the relevant steps in the user response method based on intelligent interactive tools disclosed in any of the foregoing embodiments. Alternatively, the electronic device 20 in this embodiment may specifically be an electronic computer.

[0085] In this embodiment, the power supply 23 is used to provide operating voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and external devices, and the communication protocol it follows can be any communication protocol applicable to the technical solution of this application, and is not specifically limited here; the input / output interface 25 is used to acquire external input data or output data to the outside world, and its specific interface type can be selected according to specific application needs, and is not specifically limited here.

[0086] In addition, the memory 22, as a carrier for resource storage, can be a read-only memory, random access memory, disk or optical disk, etc. The resources stored thereon can include operating system 221, computer program 222, etc., and the storage method can be temporary storage or permanent storage.

[0087] The operating system 221 is used to manage and control the various hardware devices on the electronic device 20 and the computer program 222, which may be Windows Server, Netware, Unix, Linux, etc. In addition to including a computer program capable of performing the user response method based on the intelligent interactive tool executed by the electronic device 20 as disclosed in any of the foregoing embodiments, the computer program 222 may further include computer programs capable of performing other specific tasks.

[0088] Furthermore, this application also discloses a computer-readable storage medium for storing a computer program; wherein, when the computer program is executed by a processor, it implements the aforementioned user response method based on an intelligent interactive tool. Specific steps of this method can be found in the corresponding content disclosed in the foregoing embodiments, and will not be repeated here.

[0089] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to in the method section.

[0090] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0091] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented directly by hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0092] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0093] The technical solutions provided in this application have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A user answering method based on an intelligent interaction tool, characterized by, The method comprises the following steps: obtaining a product question sent by a user terminal for a target product based on an intelligent interaction tool, and generating an initial answer to the product question by using a preset large language model; analyzing the product question based on a preset intent classifier and a preset sentiment adjective matcher to obtain a target two-tuple corresponding to the product question; the target two-tuple comprises an intent label and a sentiment adjective; analyzing the operation behavior of the user terminal for the initial answer based on a local timer and a preset event burying point to obtain a behavior label; if the target two-tuple and the behavior label constitute negative feedback information, then calling a natural language generation tool to generate an initial question answer, and obtaining a modification instruction of a preset operation and maintenance party for the initial question answer based on a preset interaction interface, so as to determine a target question answer of the product question based on the modification instruction and the initial question answer. 2.The smart interactive tool based user answering method according to claim 1, wherein, The method comprises the following steps: analyzing the question generation scene corresponding to the product question by using a preset intent classifier to generate an intent label corresponding to the product question; the preset sentiment adjective matcher performs semantic analysis on the product question, and selects a sentiment adjective corresponding to the product question from a preset sentiment adjective library based on the corresponding analysis result; constructing a target two-tuple corresponding to the product question based on the intent label and the sentiment adjective. 3.The smart interactive tool based user answering method according to claim 1, wherein, Before analyzing the operation behavior of the user terminal for the initial answer based on the local timer and the preset event burying point, the method further comprises the following steps: tracking all user behaviors for a target product, and generating event naming and attached attributes corresponding to the user behaviors; implanting a target burying point instruction in the code logic corresponding to the local client based on the event naming and attached attributes to obtain a preset event burying point. 4.The smart interactive tool based user answering method according to claim 3, wherein, The method comprises the following steps: using a local timer to perform timing operation to determine whether the user terminal takes operation behavior for the initial answer within a preset time period; if the user terminal takes operation behavior for the initial answer within the preset time period, then analyzing the operation behavior by using a front-end SDK and a preset event burying point to obtain a behavior label. 5.The smart interactive tool based user answering method according to claim 1, wherein, The method comprises the following steps: calling a natural language generation tool and selecting a target answer from a historical data storage based on the target two-tuple to obtain an initial question answer; displaying the initial question answer to a preset operation and maintenance party based on a preset visual interface, and obtaining a modification instruction of the preset operation and maintenance party for the initial question answer based on a preset interaction interface. 6.The smart interactive tool based user answering method according to claim 5, wherein, After determining the target question answer of the product question based on the modification instruction and the initial question answer, the method further comprises the following steps: The target pair and the target question answer are generated into a to-be-stored answer, and the to-be-stored answer is updated into the historical data repository. 7.The smart interactive tool based user answering method according to claim 5, wherein, The target question answer of the product question is determined based on the modification instruction and the initial question answer, including: A target key-value pair is generated based on the target pair and the target question answer, Key in the target key-value pair is a target pair; and Value in the target key-value pair is the target question answer; The target key-value pair is updated into the historical data repository.

8. A user answering apparatus based on an intelligent interactive tool, characterized by, Including: An initial answer generation module is configured to acquire a product question sent by a user terminal for a target product based on an intelligent interaction tool, and generate an initial answer for the product question by using a preset large language model; A pair generation module is configured to analyze the product question based on a preset intent classifier and a preset sentiment adjective matcher to obtain a target pair corresponding to the product question; The target pair includes an intent label and a sentiment adjective; A behavior analysis module is configured to analyze an operation behavior of the user terminal for the initial answer based on a local timer and a preset event burying point to obtain a behavior label; A target answer generation module is configured to, if the target pair and the behavior label constitute negative feedback information, call a natural language generation tool to generate an initial question answer, and acquire a modification instruction of a preset operation and maintenance party for the initial question answer based on a preset interaction interface, so as to determine a target question answer of the product question based on the modification instruction and the initial question answer.

9. An electronic device, comprising: Including: A memory is configured to save a computer program; A processor is configured to execute the computer program to implement the user answer method based on the intelligent interaction tool according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, A computer program is saved, and the computer program is executed by a processor to implement the user answer method based on the intelligent interaction tool according to any one of claims 1 to 7.

Images