Artificial intelligence-based consultation method, apparatus and device, and storage medium

Abstract

The invention provides a consultation method and device based on artificial intelligence, equipment and a storage medium, and relates to the technical field of artificial intelligence. The method comprises the steps of obtaining a consultation strategy text; the consultation strategy text is input into a large model, a plurality of executable rule entry texts are obtained, and each rule entry text comprises a visitor state and an abstract action which is suggested to be executed and corresponds to the visitor state. The method has the technical effects that manual coding is avoided, and manpower consumption is reduced.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, specifically to a consultation method, apparatus, device, and storage medium based on artificial intelligence. Background Technology

[0002] With the development of artificial intelligence technology, intelligent question answering and human-computer interaction consultation have gradually become research hotspots. Achieving intelligent and natural human-computer language interaction has become one of the important goals in the field of artificial intelligence.

[0003] However, most existing human-computer interaction systems still rely on manually coded rules. Programmers edit logic code based on texts summarized by experts to achieve intelligent question answering and human-computer interactive consultation. Although the above method achieves intelligent question answering and human-computer interactive consultation, it involves a large amount of coding and requires a lot of manpower. Summary of the Invention

[0004] This application provides an artificial intelligence-based consultation method, apparatus, device, and storage medium to avoid manual coding and reduce manpower consumption.

[0005] In a first aspect, this application provides an artificial intelligence-based consultation method, the method comprising: acquiring consultation strategy text; inputting the consultation strategy text into a large model to obtain multiple executable rule entry texts, wherein each rule entry text includes a visitor state and an abstract action to be performed corresponding to the visitor state.

[0006] By adopting the above technical solution, this method fully leverages human advantages in language expression and logical thinking by acquiring consultation strategy texts, forming a large-scale text knowledge base. Compared to manually writing rules, directly outputting language text can more fully express professional knowledge. Simultaneously, utilizing the semantic analysis capabilities of large-scale models, the text can be automatically and efficiently parsed to extract structured knowledge such as states and actions, generating multiple large-scale rule entries, avoiding the complexity of manual knowledge conversion and encoding.

[0007] Optionally, after inputting the consultation strategy text into the large model to obtain multiple executable rule entries, the method further includes: obtaining the visitor's statement; determining the current visitor state based on the visitor's statement, and selecting multiple rule entries corresponding to the current visitor state; prioritizing the multiple rule entries; and executing the abstract actions suggested for the current visitor state corresponding to the highest priority, non-contradictory rule entries in the priority ranking through the large model to generate a specific response.

[0008] By adopting the above technical solution, after constructing a state- and action-based knowledge base, this method also designs a systematic application process to achieve intelligent and personalized response generation. When a visitor's statement is input, semantic analysis technology is first used to determine their current state. Then, multiple possible response rules corresponding to that state are found in the knowledge base, and the rule with the highest priority and best match is selected. A complete application solution from knowledge acquisition to response generation is constructed, achieving systematic intelligent optimization in knowledge representation, semantic understanding, and personalized generation.

[0009] Optionally, determining the current visitor state based on the visitor's statement and selecting multiple rule entries corresponding to the current visitor state includes: performing semantic analysis on the visitor's statement to analyze the state reflected in the visitor's statement; matching a first visitor state in the database corresponding to the state reflected in the visitor's statement, where the first visitor state is the current visitor state; and searching the database for multiple rule entries that match the current visitor state.

[0010] By adopting the above technical solution, after obtaining the visitor's statement, natural language processing technology is first used to perform semantic analysis on the statement to determine the current state most likely reflected by the visitor's statement. Once the current visitor state is determined, the system can quickly retrieve multiple selectable rule entries related to that state from the rule knowledge base. This alternative approach avoids the need to manually establish a state labeling system and recognition logic, thus completely eliminating manual annotation and reducing the complexity of implementing the requirements.

[0011] Optionally, prioritizing the plurality of rule entries includes: scoring the plurality of rule entries to obtain a score for each rule entry; and prioritizing the plurality of rule entries according to the scores from highest to lowest priority.

[0012] By employing the above technical solution, the scoring model can comprehensively analyze the characteristics of rule entries and assign scores. These characteristics can include the rule's historical usage effectiveness and logical rationality. The model can be trained through machine learning to achieve automated scoring. After obtaining the quality scores of the rule entries, this solution simply sorts the rules according to their scores; the higher the score, the higher its selection priority during response generation. Using this scoring and sorting scheme effectively filters and optimizes the ranking of multiple candidate rules, making it more likely that the most suitable and appropriate rule entry for the current situation will be selected when generating a response. Compared to random selection, this significantly improves response quality and satisfaction.

[0013] Optionally, the step of scoring the plurality of rule entries to obtain a score for each rule entry includes: establishing a preliminary scoring model; training the preliminary scoring model with labeled data to obtain a scoring model; and sending the plurality of rule entries to the scoring model to obtain a score for each rule entry.

[0014] By adopting the above technical solution and using machine learning to train the scoring model, compared to manual scoring, automation and objectivity can be achieved, and the scoring strategy can be continuously learned and optimized. Using rich consultation data for training allows the scoring results to more closely resemble real-world scenarios, thereby generating rule entries of optimal quality.

[0015] Optionally, the step of executing the suggested abstract actions for the current visitor's state corresponding to the highest priority, non-contradictory rule entries in the priority ranking through the large model to generate a specific response includes: inputting the abstract actions from the highest priority, non-contradictory rule entries into the large model; and the large model generating a specific response corresponding to the abstract actions based on the abstract actions and related background knowledge.

[0016] By adopting the above technical solution, the abstract actions from the optimal rules are directly input into the large model. The large model possesses a deep understanding of background semantics, can parse the intent within the actions, and, combined with the current context, utilizes rich linguistic knowledge to generate a complete and fluent response text. This integrated approach to semantic generation from the large model ensures that the response simultaneously adheres to expert rules and meets the user's personalized needs. It strictly follows the defined logical flow of the dialogue while also possessing powerful language organization and expression capabilities.

[0017] Optionally, after generating a specific response by executing the abstract action suggested by the current visitor state corresponding to the highest priority non-contradictory rule entries in the priority ranking through the large model, the method further includes: receiving feedback information sent by the user terminal; and sending the feedback information to the business personnel's terminal device so that the business personnel can adjust and optimize the rule entries based on the feedback information.

[0018] By adopting the above technical solution and forming a closed loop through real user interaction, the quality of the rule base can be continuously improved, enabling the knowledge base to evolve and significantly increasing the satisfaction of system responses. This solution allows the system to autonomously and intelligently iterate and optimize knowledge, promoting the interpretability and controllability of knowledge-driven dialogue systems.

[0019] Secondly, this application provides an artificial intelligence-based consultation device, the device comprising: an acquisition module and an output module; wherein, the acquisition module is used to acquire consultation strategy text; the output module is used to input the consultation strategy text into a large model to obtain multiple executable rule entry texts, wherein each rule entry text includes a visitor state and an abstract action to be performed corresponding to the visitor state.

[0020] By employing the aforementioned technical solution, this method leverages the strengths of humans in language expression and logical thinking by acquiring rich consultation strategy texts written by human experts, forming a large-scale textual knowledge base. Compared to manually writing rules, experts directly outputting linguistic text can more fully express their professional knowledge. Simultaneously, utilizing the semantic analysis capabilities of large-scale models, the text can be automatically and efficiently parsed, extracting structured knowledge such as state-action pairs, generating multiple large-scale rule entries, and forming a knowledge base, thus avoiding the complexity of manual knowledge conversion and encoding.

[0021] Thirdly, this application provides an electronic device that adopts the following technical solution: including a processor, a memory, a user interface, and a network interface, wherein the memory is used to store instructions, the user interface and the network interface are used to communicate with other devices, and the processor is used to execute the instructions stored in the memory to enable the electronic device to execute a computer program based on any of the above-mentioned artificial intelligence-based consultation methods.

[0022] Fourthly, this application provides a computer-readable storage medium that stores a computer program capable of being loaded by a processor and executing any of the aforementioned artificial intelligence-based consultation methods.

[0023] In summary, this application includes at least one of the following beneficial technical effects: 1. It avoids manual coding, reducing manpower consumption; 2. This integrated approach to semantic generation ensures that responses simultaneously adhere to expert rules and meet individual user needs. It rigorously follows a predefined logical flow while also possessing strong language organization and expression capabilities. Attached Figure Description

[0024] Figure 1 This is a flowchart illustrating an artificial intelligence-based consultation method provided in an embodiment of this application; Figure 2 This is a schematic diagram of the structure of an artificial intelligence-based consulting device provided in an embodiment of this application; Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0025] Explanation of reference numerals in the attached figures: 1000, electronic device; 1001, processor; 1002, communication bus; 1003, user interface; 1004, network interface; 1005, memory. Detailed Implementation

[0026] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification 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.

[0027] In the description of the embodiments in this application, words such as "illustrative," "for example," or "for example" are used to indicate examples, illustrations, or explanations. Any embodiment or design described as "illustrative," "for example," or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or designs. Rather, the use of words such as "illustrative," "for example," or "for example" is intended to present the relevant concepts in a specific manner.

[0028] Figure 1 This is a flowchart illustrating an artificial intelligence-based consultation method provided in an embodiment of this application. It should be understood that, although... Figure 1 The steps in the flowchart are shown sequentially as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows; unless explicitly stated herein, there is no strict order requirement for the execution of these steps, and they can be executed in other orders; and Figure 1 At least some of the steps in the process may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least some of the sub-steps or stages of other steps.

[0029] This application discloses an artificial intelligence-based consultation method, such as... Figure 1 As shown, the method includes S101-S102.

[0030] S101, Obtain the consulting strategy text.

[0031] In one example, this invention requires acquiring consultation strategy texts written by human experts as a knowledge source for constructing the consultation system. This is because high-quality consultation responses need to fully utilize the expertise and rich experience of human experts. Compared to machine-generated content, consultation strategy texts written by human experts are often more accurate, comprehensive, and logically sound, and can serve as the foundation of the consultation system's knowledge base.

[0032] In practice, the first step is to select human experts with extensive domain knowledge and strong communication skills. Then, based on different consultation scenarios, these experts will write systematic consultation strategy texts addressing various potential client questions or states. For example, for an investment or financial management issue, the expert can provide detailed suggested steps to solve the problem; for a user's emotional state, the expert can offer coping strategies. The texts written by the experts will cover all possible scenarios.

[0033] After obtaining the text written by experts, it undergoes preprocessing, including format conversion and content extraction, to transform it into structured knowledge. Finally, this text content is input into a knowledge base as the basis for generating subsequent responses. This method maximizes the use of human expert knowledge to build a high-quality, comprehensive consultation knowledge base. Based on this knowledge base, responses generated using a large model can achieve a high level of professionalism and logical coherence, thus realizing efficient and intelligent consultation interaction.

[0034] S102, input the consultation strategy text into the large model to obtain multiple executable rule entry texts, where each rule entry text includes the visitor's state and the suggested abstract action to be performed for the corresponding visitor state.

[0035] In one example, the large model in this approach refers to a large-scale pre-trained language model such as GPT-4. This type of model, pre-trained on large-scale text datasets, achieves powerful natural language understanding and generation capabilities. Large models can perform deep semantic analysis on text, automatically extract key information, and generate new text. The advantage of using large models is the ability to quickly extract structured information from unstructured text.

[0036] Rule entries are knowledge structures extracted from consultation strategy texts that link visitor states with corresponding suggested actions. Each rule entry consists of two parts: Visitor State: This refers to a situation or state that the visitor may experience during the consultation process, such as "housing anxiety." Corresponding Action: This is the abstract action the dialogue system should take when the visitor exhibits this state. For example, "guided description, step-by-step breakdown." Linking these two parts together forms a rule entry, which constitutes the basic unit of the dialogue system's knowledge base. Rule entries not only include judgments about different visitor states but also provide corresponding action suggestions. Building a knowledge base based on rule entries can guide the dialogue system to generate high-quality responses.

[0037] After acquiring consultation strategy texts written by human experts, these texts are input into a pre-trained large model for processing. Because the large model has powerful natural language processing capabilities, it can analyze the semantics of long texts, extract key information, and generate corresponding structured outputs.

[0038] The specific implementation involves inputting the collected expert consultation strategy texts into a large-scale model. The model automatically performs deep semantic analysis on the text, detecting the visitor's state description and the corresponding expert advice or response methods. For example, from the text "If the visitor expresses anxiety about buying a house, you can first guide the visitor to describe the source of anxiety, and then give suggestions to break down the problem step by step," the large-scale model can extract the visitor's state "anxiety about buying a house" and the corresponding abstract action "guide description, break down the problem step by step."

[0039] Large-scale models can quickly and efficiently extract rule entries corresponding to states and actions from massive amounts of consultation strategy text. Finally, the extracted rule entries are saved in textual form to a knowledge base database as a reference for subsequent response generation. This method leverages the powerful semantic processing capabilities of large-scale models to rapidly obtain clear rule knowledge from freely written text, enriching the consultation system's knowledge base. Subsequently, only the visitor's state needs to be input to directly retrieve the suggested actions corresponding to the matching rules, thereby generating high-quality responses and achieving intelligent and automated consultation services.

[0040] S101-S102 constitute the preparation phase, which involves converting consultation strategy texts written by human experts into multiple rule entries. These rule entries are then processed to construct a knowledge base. This knowledge base provides the foundational resources for subsequent applications. When a visitor inputs a statement, the corresponding rule can be directly retrieved from the knowledge base to generate a response, thus achieving intelligent human-computer consultation services. For specific applications, please refer to the following examples.

[0041] Obtain the visitor's statement; based on the visitor's statement, determine the current visitor's state and select multiple rule entries corresponding to the current visitor's state; prioritize the multiple rule entries; through the large model, execute the abstract action suggested by the highest priority rule entry corresponding to the current visitor's state to generate a specific response.

[0042] In one example, the current visitor state refers to the state or situation the visitor is currently in, determined by analyzing the statements input by the visitor during the interaction between the human-computer dialogue system and the visitor. For example, the visitor might express states such as "feeling anxious," "uncertain about the outcome," or "lacking exercise." Determining the current visitor state is crucial for subsequently determining the appropriate strategy the system should adopt.

[0043] Suggested abstract actions: After determining the visitor's current state, the rules in the knowledge base will provide abstract actions that the system should perform in that state. An abstract action represents the dialogue system's next response strategy; it is an abstract verbal description, such as "provide guidance" or "offer advice." These abstract actions also need to be combined with specific scenarios and converted into concrete, semantically coherent response sentences by subsequent generation modules. Designing actions as abstract representations increases the applicability of the knowledge base rules. Associating visitor states with corresponding abstract actions as rule entries forms the foundation of the knowledge base, guiding the dialogue system to generate highly targeted responses.

[0044] During the consultation dialogue, each time the client's statement is received, it needs to be processed and a response provided to achieve intelligent and natural human-computer interaction. Specifically, when the client's problem description, "I feel very anxious when buying a house, afraid that I haven't considered things carefully enough and will cause problems," is obtained, semantic analysis technology is first used to analyze the statement to determine that the client's current emotional state is "housing anxiety." Then, the system searches its built knowledge base database for multiple rule entries related to the state of "housing anxiety." For example, the following two matching rules are found: Rule 1: When the state is "housing anxiety," provide "guided description of the source of anxiety." Rule 2: When the state is "housing anxiety," provide "suggestions for breaking down complex problems."

[0045] Next, the system will call the scoring model to score and rank the two rules mentioned above, assuming rule 1 has higher priority. The system will then select the abstract action corresponding to rule 1, "guided description of the source of anxiety," and input it into the larger model for execution. The larger model will generate a semantically appropriate response based on the action and background knowledge: "Hello, I understand that buying a house can indeed be anxiety-inducing. Could you tell me what problems you are most worried about encountering when buying a house?" This response will then be returned to the visitor.

[0046] This process allows the system to automatically analyze visitor needs and generate logical responses based on knowledge base rules, achieving intelligent human-computer interaction and improving problem-solving efficiency.

[0047] Suppose that through the initial text extraction, the knowledge base contains the following two rule entries: Rule 1: When the visitor's state is "lack of exercise, high life stress," the action is "suggest the visitor engage in yoga." Rule 2: When the visitor's state is "lack of exercise, high life stress," the action is "provide gym coupons." Now, the visitor's statement is: "I've been extremely busy at work lately, often working overtime until very late. When I get home, I just want to lie down and rest immediately. I have absolutely no time to exercise, my health is getting worse, and I'm under a lot of stress. I really want to relax." Processing flow: Perform semantic analysis on the visitor's statement to determine that the visitor's current state is "lack of exercise, high life stress." Search the knowledge base for the rule corresponding to this state to obtain Rule 1 and Rule 2. Score the two rules using a scoring model, assuming Rule 1 scores higher. Select Rule 1 and extract its abstract action "suggest the visitor engage in yoga."

[0048] The abstract action is input into the large model, which generates a specific response based on the action and background knowledge: "Fitness can indeed help us relieve stress. I suggest you try yoga; it can soothe the mind and body and is also very beneficial for improving flexibility. I have some beginner yoga video tutorials here; you can follow the videos to do simple yoga exercises at home." The final generated response is then returned to the visitor. In this way, a logical and personalized response is generated through knowledge base rules and the large model.

[0049] Based on the visitor's statement, determine the current visitor state, and select multiple rule entries corresponding to the current visitor state, including: performing semantic analysis on the visitor's statement to analyze the state reflected in the visitor's statement; matching the first visitor state in the database that corresponds to the state reflected in the visitor's statement, with the first visitor state being the current visitor state; and searching the database for multiple rule entries that match the current visitor state.

[0050] In one example, to enable the system to analyze the intent and underlying state expressed in a visitor's statement, semantic analysis of the input statement is required to identify the visitor's state reflected in the current statement. For instance, when a visitor inputs, "I've been feeling extremely anxious and confused lately because of a job transfer, and I can't sleep at night," the system will use natural language processing technology to determine that this most likely reflects the visitor's state as "career planning anxiety." After obtaining the current visitor's state, the system will search the knowledge base rule library for all rule entries corresponding to that state label. For example, the following matching rules might be found: Rule 1: When the status is "career planning anxiety", guide the description of the source of anxiety.

[0051] Rule 2: When the status is "career planning anxiety", provide career planning advice.

[0052] Through this process, the system can automatically identify the visitor's state using semantic analysis technology, and then find the corresponding rules based on the state, avoiding subjective human judgment and making subsequent response generation more intelligent and automated. Responses generated using this rule-based method are also more logical and better aligned with the visitor's needs.

[0053] Prioritizing multiple rule entries involves: scoring each rule entry to obtain a score; and then sorting the rule entries by priority from highest to lowest based on their scores.

[0054] In one example, after obtaining multiple possible rule entries related to the visitor's current state, these rule entries need to be scored and then sorted according to their scores. The reason for scoring and sorting is that for the same visitor state, the knowledge base may contain multiple different rule entries, some of which may be more applicable and need to be given higher priority so that they can be used preferentially when generating responses.

[0055] In practice, a scoring model can be built to predict the quality score of each rule entry. This scoring model can be trained using machine learning methods to comprehensively evaluate various features within the rule entry. Alternatively, historical data on rule application can be used for statistical analysis, assigning higher weights to these features. After scoring all candidate rule entries, the system can simply rank the rules based on their scores; higher scores result in higher priority during response generation.

[0056] In this way, when determining the response content, the system will first select the rule with the highest score and the strongest applicability, thus maximizing the generation of the most appropriate and logical response to the visitor's current state. Furthermore, with continuous application, the scoring model for the rule entries can be continuously optimized, making the ranking results more accurate. Using this method can significantly improve the effectiveness of human-computer interaction.

[0057] The process of scoring multiple rule items to obtain a score for each rule item includes: establishing a preliminary scoring model; training the preliminary scoring model with labeled data to obtain a scoring model; and sending multiple rule items to the scoring model to obtain a score for each rule item.

[0058] In one example, to automatically score and rank multiple rule entries, a scoring model needs to be built. Specifically, first, an initial basic scoring model is designed based on the characteristics of the rule entries. This model can pre-code scores for certain salient features. Then, real-world rule entry application records are collected as labeled data. This data annotates the effectiveness evaluation of each rule entry after its use, such as the quality of the generated responses. Using this labeled data, a preliminary scoring model can be trained using machine learning algorithms to further optimize its scoring strategy.

[0059] After training the scoring model, when multiple candidate rule entries are obtained, these rule entries can be directly fed into the scoring model. The scoring model will automatically analyze their features and give a quality score for each rule entry based on the trained scoring strategy. Finally, these scores are sorted, with rule entries with higher scores having higher selection priority during response generation.

[0060] The large model executes abstract actions suggested for the current visitor's state based on the highest priority, non-contradictory rule entries in the priority ranking, and generates specific responses. This includes: inputting the abstract actions from the highest priority, non-contradictory rule entries into the large model; and generating specific responses corresponding to the abstract actions based on the abstract actions and relevant background knowledge.

[0061] In one example, after evaluating the quality and prioritizing multiple candidate rule entries using a scoring model, a larger model is needed to generate a specific response tailored to the current visitor's state based on the optimal rule entry. The reason for this larger model generation is that the rule entries only provide abstract actions, which need to be transformed into coherent semantic expressions, and the larger model possesses powerful semantic understanding and generation capabilities.

[0062] Specifically, the system directly selects the best-ranked rule entry and extracts the corresponding abstract action description text from it, such as "provide career planning advice." Then, the system inputs this abstract action text into a pre-trained large model. The model comprehensively analyzes the current visitor's state and the logic within the context, and generates a semantically consistent and complete response text based on the abstract action instruction. For example, "Hello, changing jobs can indeed cause anxiety. I suggest you first list the advantages of your current job and the growth you've achieved, which can help maintain a positive mindset; at the same time, try to broaden your horizons, reflect on your professional development direction, and formulate a reasonable career planning roadmap." After generating a specific response by executing the abstract action suggested by the highest priority, non-contradictory rule entries in the large model for the current visitor's state, the process also includes: receiving feedback information sent by the user; and sending the feedback information to the business personnel's terminal devices so that the business personnel can adjust and optimize the rule entries based on the feedback information.

[0063] In one example, to continuously improve the quality of the system's responses, after generating and returning specific responses to visitors, the system also needs to obtain feedback from visitors regarding their satisfaction with the responses. The purpose of obtaining feedback is to verify the effectiveness of the current rule entries and response generation strategies, and to identify any existing problems.

[0064] Specific methods for obtaining feedback include: providing a simple rating mechanism after the response, or obtaining visitor evaluations of the response through questionnaires. The system will receive and record this feedback information. Subsequently, the feedback information will be sent to customer service personnel periodically, where professionals will analyze the feedback to check whether the rule entries and their ratings are reasonable, or whether the response generation strategy needs adjustment.

[0065] Service personnel can modify or delete ineffective rules based on feedback, and also add more reasonable rules. The scoring model can be incrementally tuned and continuously optimized. By constantly adjusting and iterating the system's knowledge base and generation strategies using real visitor feedback, the quality and satisfaction of responses can be continuously improved, making the system's responses more aligned with visitor needs.

[0066] Based on the above method, this application also discloses an artificial intelligence-based consulting device, such as... Figure 2 As shown, Figure 2 This is a schematic diagram of the structure of an artificial intelligence-based consulting device provided in an embodiment of this application.

[0067] An artificial intelligence-based consultation device includes: an acquisition module and an output module; wherein the acquisition module is used to acquire consultation strategy text; the output module is used to input the consultation strategy text into a large model to obtain multiple executable rule entry texts, wherein each rule entry text includes the visitor's state and the suggested abstract action to be performed for the corresponding visitor state.

[0068] In one example, the above-mentioned device is also used to obtain the visitor's statement; determine the current visitor state based on the visitor's statement, select multiple rule entries corresponding to the current visitor state; prioritize the multiple rule entries; and generate a specific response by executing the abstract actions suggested by the highest priority, non-contradictory rule entries corresponding to the current visitor state through the large model.

[0069] In one example, the device is also used to perform semantic analysis on the visitor's statement, analyze the state reflected in the visitor's statement; match the first visitor state in the database corresponding to the state reflected in the visitor's statement, the first visitor state being the current visitor state; and search the database for multiple rule entries that match the current visitor state.

[0070] In one example, the above-mentioned device is also used to score multiple rule entries to obtain a score for each rule entry; and to sort the multiple rule entries in descending order of priority according to the score.

[0071] In one example, the above-mentioned apparatus is also used to establish a preliminary scoring model; train the preliminary scoring model with labeled data to obtain a scoring model; and send multiple rule entries to the scoring model to obtain the score of each rule entry.

[0072] In one example, the aforementioned device is also used to input abstract actions from the highest priority, non-contradictory rule entries into a large model; the large model generates specific responses to the corresponding abstract actions based on the abstract actions and related background knowledge.

[0073] In one example, the aforementioned device is also used to receive feedback information sent by the user terminal; and to send the feedback information to the terminal device of the business personnel so that the business personnel can adjust and optimize the rule entries based on the feedback information.

[0074] It should be noted that the above embodiments of the apparatus are only illustrated by the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the apparatus and method embodiments provided in the above embodiments belong to the same concept, and the specific implementation process can be found in the method embodiments, which will not be repeated here.

[0075] Please see Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Figure 3 As shown, the electronic device 1000 may include: at least one processor 1001, at least one network interface 1004, a user interface 1003, a memory 1005, and at least one communication bus 1002.

[0076] The communication bus 1002 is used to realize the connection and communication between these components.

[0077] The user interface 1003 may include a display screen and a camera. Optionally, the user interface 1003 may also include a standard wired interface and a wireless interface.

[0078] The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface).

[0079] The processor 1001 may include one or more processing cores. The processor 1001 connects to various parts of the server using various interfaces and lines, and performs various server functions and processes data by running or executing instructions, programs, code sets, or instruction sets stored in the memory 1005, and by calling data stored in the memory 1005. Optionally, the processor 1001 may be implemented using at least one hardware form of Digital Signal Processing (DSP), Field-Programmable Gate Array (FPGA), or Programmable Logic Array (PLA). The processor 1001 may integrate one or a combination of several of the following: Central Processing Unit (CPU), Graphics Processing Unit (GPU), and modem. The CPU primarily handles the operating system, user interface, and applications; the GPU is responsible for rendering and drawing the content to be displayed on the screen; and the modem handles wireless communication. It is understood that the modem may also not be integrated into the processor 1001 and may be implemented as a separate chip.

[0080] The memory 1005 may include random access memory (RAM) or read-only memory. Optionally, the memory 1005 may include a non-transitory computer-readable storage medium. The memory 1005 can be used to store instructions, programs, code, code sets, or instruction sets. The memory 1005 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as touch function, sound playback function, image playback function, etc.), instructions for implementing the above-described method embodiments, etc.; the data storage area may store data involved in the above-described method embodiments, etc. Optionally, the memory 1005 may also be at least one storage device located remotely from the aforementioned processor 1001. Figure 3 As shown, the memory 1005, which serves as a computer storage medium, may include an operating system, a network communication module, a user interface module, and an application program based on an artificial intelligence-based consulting method.

[0081] exist Figure 3 In the electronic device 1000 shown, the user interface 1003 is mainly used to provide an input interface for the user and to obtain the user input data; while the processor 1001 can be used to call an application program stored in the memory 1005 that is an artificial intelligence-based consultation method. When executed by one or more processors, the electronic device performs one or more of the methods described in the above embodiments.

[0082] An electronic device readable storage medium stores instructions that, when executed by one or more processors, cause the electronic device to perform one or more of the methods described in the above embodiments.

[0083] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0084] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0085] In the several embodiments provided in this application, it should be understood that the disclosed apparatus can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual couplings, direct couplings, or communication connections may be through some service interfaces; indirect couplings or communication connections between devices or units may be electrical or other forms.

[0086] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0087] Furthermore, the functional units in the various embodiments of this application 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 integrated unit can be implemented in hardware or as a software functional unit.

[0088] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage device (CMD). Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned memory includes various media capable of storing program code, such as USB flash drives, portable hard drives, magnetic disks, or optical disks.

[0089] The foregoing description is merely an exemplary embodiment of this disclosure and should not be construed as limiting the scope of this disclosure. Any equivalent changes and modifications made in accordance with the teachings of this disclosure shall still fall within the scope of this disclosure. Other embodiments of this disclosure will be readily apparent to those skilled in the art upon consideration of the specification and practice of the disclosure herein. This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not described herein. The specification and embodiments are to be considered exemplary only, and the scope and spirit of this disclosure are defined by the claims.

Claims

1. A consultation method based on artificial intelligence, characterized in that, The method includes: Obtain the consulting strategy text; The consultation strategy text is input into the large model to obtain multiple executable rule entries, wherein each rule entry includes the visitor's state and the suggested abstract action to be performed corresponding to the visitor's state.

2. The AI-based consultation method according to claim 1, characterized in that, After inputting the consultation strategy text into the large model to obtain multiple executable rule entries, the process further includes: Obtain visitor statements; Based on the visitor's statement, determine the current visitor state and select multiple rule entries corresponding to the current visitor state; The multiple rule entries are sorted by priority. The large model executes the abstract actions suggested for the current visitor's state, corresponding to the highest priority, non-contradictory rule entries in the priority ranking, to generate a specific response.

3. The AI-based consultation method according to claim 2, characterized in that, The step of determining the current visitor state based on the visitor's statement and selecting multiple rule entries corresponding to the current visitor state includes: Perform semantic analysis on the visitor's statements to analyze the state reflected in the visitor's statements; Match the first visitor state in the database to the state reflected in the visitor's statement, where the first visitor state is the current visitor state; Search the database for multiple rule entries that match the current visitor's status.

4. The AI-based consultation method according to claim 2, characterized in that, The step of prioritizing the multiple rule entries includes: The multiple rule entries are scored to obtain a score for each rule entry; Based on the scores, the multiple rule entries are sorted in descending order of priority.

5. The AI-based consultation method according to claim 4, characterized in that, The scoring of the multiple rule entries to obtain a score for each rule entry includes: Establish a preliminary scoring model; The initial scoring model is trained using labeled data to obtain the final scoring model. The multiple rule entries are sent to the scoring model to obtain the score for each rule entry.

6. The AI-based consultation method according to claim 2, characterized in that, The abstract action of executing the suggestions for the current visitor's state corresponding to the highest priority, non-contradictory rule entries in the priority ranking through the large model, and generating a specific response, includes: The abstract actions from the highest priority, non-contradictory rule entries are input into the large model; The large model generates a specific response corresponding to the abstract action based on the abstract action and related background knowledge.

7. The AI-based consultation method according to claim 2, characterized in that, After generating a specific response by executing the abstract action of suggesting the current visitor's state based on the highest priority, non-contradictory rule entries in the priority ranking through the large model, the process further includes: Receive feedback information sent by the user terminal; The feedback information is sent to the terminal device of the business personnel so that the business personnel can adjust and optimize the rule entries based on the feedback information.

8. A consultation device based on artificial intelligence, characterized in that, The device includes: an acquisition module and an output module; wherein... The acquisition module is used to acquire the consultation strategy text; The output module is used to input the consultation strategy text into the large model to obtain multiple executable rule entry texts, wherein each rule entry text includes the visitor's state and the suggested abstract action to be performed corresponding to the visitor's state.

9. An electronic device, characterized in that, The device includes a processor, a memory, a user interface, and a network interface. The memory is used to store instructions, the user interface and the network interface are used to communicate with other devices, and the processor is used to execute the instructions stored in the memory to cause the electronic device to perform the method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer program is stored that can be loaded by a processor and executed as described in any one of claims 1-7.

Images