Response information output method and device, storage medium and electronic equipment

Abstract

The invention discloses a response information output method and device, a storage medium and electronic equipment, and relates to the technical field of smart home, and the response information output method comprises the following steps: inputting first interaction information input by a target object into an enhancement engine, so that the enhancement engine carries out intention recognition on the first interaction information, obtaining intention information corresponding to the first interaction information; controlling the enhancement engine to reconstruct the first interaction information according to the intention information and an interaction template so as to generate second interaction information; the second interaction information is input into the large model, the first response information corresponding to the second interaction information is output through the large model, and by adopting the technical scheme, the problem that the accuracy rate of the response information output by the large model is low is solved.

Description

Technical Field

[0001] The present application relates to the field of smart home technology, and more specifically, to a method and device for outputting response information, a storage medium, and an electronic device. Background Art

[0002] In current vertical industry large language model (LLM) applications, non-professional users often face a core challenge when asking questions: due to the lack of necessary background knowledge and professional terminology in a specific field, the questions they submit may not be precise or professional enough, which directly leads to the low accuracy of the first response information output by the large model. For example, in the scenario of home appliance fault diagnosis, users may use everyday language rather than professional terminology when describing the problem, such as "the washing machine is not working" instead of describing the specific fault phenomenon or component status, which makes it difficult for the large model to accurately understand the details of the problem, thereby reducing the pertinence and effectiveness of the solution it generates.

[0003] That is, the problems existing in the prior art are mainly reflected in that the accuracy of the response information output by the large model is low.

[0004] In view of the problem of low accuracy of response information output by large models in related technologies, no effective solution has been proposed yet. Summary of the invention

[0005] The embodiments of the present application provide a method and device for outputting response information, a storage medium, and an electronic device, so as to at least solve the problem of low accuracy of response information output by a large model in the related art.

[0006] According to an embodiment of the embodiments of the present application, a method for outputting response information is provided, including: inputting first interaction information input by a target object into an enhancement engine, so that the enhancement engine performs intent recognition on the first interaction information to obtain intent information corresponding to the first interaction information; controlling the enhancement engine to reconstruct the first interaction information according to the intent information and an interaction template to generate second interaction information; inputting the second interaction information into a large model to output first response information corresponding to the second interaction information through the large model.

[0007] In an exemplary embodiment, controlling the enhancement engine to reconstruct the first interaction information according to the intent information and the interaction template to generate second interaction information includes: parsing the first interaction information to generate a first parsing result; determining a target keyword of the first interaction information according to the first parsing result; determining type information corresponding to the first interaction information according to the intent information, and determining an interaction template corresponding to the type information; and generating the second interaction information according to the target keyword and the interaction template.

[0008] In an exemplary embodiment, determining the target keyword of the first interaction information based on the first parsing result includes: obtaining historical interaction information of the target object and parsing the historical interaction information to determine the second parsing result; identifying related keywords in the first parsing result and the second parsing result; and determining the target keyword of the first interaction information among the keywords based on the intention information.

[0009] In an exemplary embodiment, the second interaction information is generated according to the target keyword and the interaction template, including: writing the target keyword into the interaction template, and determining whether there is a missing field in the interaction template in which the target keyword is written; in the case that there is a missing field in the interaction template in which the target keyword is written, sending prompt information to the target object according to the missing field; and filling the missing field according to the second response information sent by the target object to generate the second interaction information.

[0010] In an exemplary embodiment, after the first response information corresponding to the second interaction information is outputted through the large model, the method further includes: receiving evaluation information input by the target object, and determining the target object's satisfaction with the first response information based on the evaluation information; using the first interaction information and the second interaction information as training samples, and determining the sample type of the training sample based on the satisfaction, wherein the sample type includes one of the following: positive sample, negative sample; training the enhancement engine based on the sample type and the training sample.

[0011] In an exemplary embodiment, determining the target object's satisfaction with the first response information according to the evaluation information includes: determining the target object's emotional information according to the evaluation information; and determining the satisfaction according to the emotional information.

[0012] In an exemplary embodiment, after determining the target object's satisfaction with the first response information based on the evaluation information, the method further includes: determining the relationship between the satisfaction and a preset satisfaction; when the relationship indicates that the satisfaction is less than the preset satisfaction, sending an inquiry message to the target object, wherein the inquiry message is used to indicate improvement suggestions for the target object; receiving third response information input by the target object, and adjusting the enhancement engine according to the third response information.

[0013] According to another embodiment of the embodiments of the present application, a response information output device is also provided, including: an enhancement engine, used to accept first interaction information input by a target object; perform intent recognition on the first interaction information to obtain intent information corresponding to the first interaction information; reconstruct the first interaction information according to the intent information and an interaction template to generate second interaction information; a large model, used to receive the second interaction information input into the large model, and output first response information corresponding to the second interaction information.

[0014] According to another aspect of the embodiments of the present application, a computer-readable storage medium is provided, in which a computer program is stored, wherein the computer program is configured to execute the above-mentioned method for outputting response information when running.

[0015] According to another aspect of an embodiment of the present application, there is also provided an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the method for outputting response information through the computer program.

[0016] In an embodiment of the present application, the first interactive information input by the target object is input into the enhancement engine so that the enhancement engine performs intent recognition on the first interactive information to obtain the intent information corresponding to the first interactive information; the enhancement engine is controlled to reconstruct the first interactive information according to the intent information and the interaction template to generate the second interactive information; the second interactive information is input into the large model so that the first response information corresponding to the second interactive information is output through the large model. That is, in an embodiment of the present application, by introducing the enhancement engine for intent recognition and interaction information reconstruction, the large model's understanding and response accuracy of the user input information is effectively improved, thereby improving the accuracy of the response information output by the large model; the above technical solution is adopted to solve the problem of low accuracy of the response information output by the large model. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0018] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0019] Figure 1 It is a hardware environment diagram of a method for outputting response information in an embodiment of the present application;

[0020] Figure 2 is a flowchart of a method for outputting response information according to an embodiment of the present application;

[0021] Figure 3 is a system architecture diagram of a method for outputting response information according to an embodiment of the present application;

[0022] Figure 4 is a structural block diagram of a device for outputting response information according to an embodiment of the present application (I);

[0023] Figure 5 This is a structural block diagram (II) of a response information output device according to an embodiment of the present application. DETAILED DESCRIPTION

[0024] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0025] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0026] According to one aspect of an embodiment of the present application, a method for outputting response information is provided. The method for outputting response information is widely used in smart home, smart home, smart home device ecology, smart residential (Intelligence House) ecology and other whole-house intelligent digital control application scenarios. Optionally, in this embodiment, the above-mentioned method for outputting response information can be applied to Figure 1 In the hardware environment composed of the terminal device 102 and the server 104 shown in FIG. Figure 1As shown, the server 104 is connected to the terminal device 102 via a network, and can be used to provide services (such as application services, etc.) for the terminal or a client installed on the terminal. A database can be set on the server or independently of the server to provide data storage services for the server 104. Cloud computing and / or edge computing services can be configured on the server or independently of the server to provide data computing services for the server 104.

[0027] The network may include but is not limited to at least one of the following: wired network, wireless network. The wired network may include but is not limited to at least one of the following: wide area network, metropolitan area network, local area network, and the wireless network may include but is not limited to at least one of the following: WIFI (Wireless Fidelity), Bluetooth. The terminal device 102 may be but is not limited to a PC, a mobile phone, a tablet computer, a smart air conditioner, a smart range hood, a smart refrigerator, a smart oven, a smart stove, a smart washing machine, a smart water heater, a smart washing device, a smart dishwasher, a smart projection device, a smart TV, a smart clothes drying rack, a smart curtain, a smart audio and video, a smart socket, a smart speaker, a smart fresh air device, a smart kitchen and bathroom device, a smart bathroom device, a smart sweeping robot, a smart window cleaning robot, a smart mopping robot, a smart air purification device, a smart steamer, a smart microwave oven, a smart kitchen treasure, a smart purifier, a smart water dispenser, a smart door lock, etc.

[0028] In this embodiment, a method for outputting response information is provided, which is applied to the above terminal device. Figure 2 : is a flowchart of a method for outputting response information according to an embodiment of the present application, and the process includes the following steps:

[0029] Step S202: inputting the first interaction information input by the target object into the enhancement engine, so that the enhancement engine performs intent recognition on the first interaction information to obtain intent information corresponding to the first interaction information;

[0030] The first interactive information input by the target object (usually the user) is fed into the enhancement engine. The "first interactive information" can be a natural language question or demand description raised by the user. The enhancement engine first performs intent recognition on this information and analyzes the real intention and demand behind the user input, such as fault description, product query, service demand, etc., to ensure that subsequent steps can be carried out based on accurate intent.

[0031] Step S204, controlling the enhancement engine to reconstruct the first interaction information according to the intention information and the interaction template to generate second interaction information;

[0032] After identifying the user's intention, the enhancement engine reconstructs the first interaction information based on the intention information and the preset interaction template to generate the second interaction information. The interaction template is designed based on industry knowledge, professional terms and common problems, and can guide the enhancement engine on how to transform the user's original input to make it clearer, more professional and structured. Through reconstruction, the second interaction information not only contains the user's original intention, but may also add necessary industry background, relevant terms or more detailed problem descriptions, thereby more accurately representing the user's needs and reducing the deviation in the understanding of the large model.

[0033] Step S206: input the second interaction information into the large model, so as to output the first response information corresponding to the second interaction information through the large model.

[0034] Finally, the second interactive information is fed into the big model, which uses the deep learning capabilities and industry knowledge base of the big model to generate first response information that matches the reconstructed question.

[0035] Through the above steps, the first interactive information input by the target object is input into the enhancement engine, so that the enhancement engine performs intent recognition on the first interactive information to obtain the intent information corresponding to the first interactive information; the enhancement engine is controlled to reconstruct the first interactive information according to the intent information and the interaction template to generate the second interactive information; the second interactive information is input into the large model to output the first response information corresponding to the second interactive information through the large model. That is, in the embodiment of the present application, by introducing the enhancement engine for intent recognition and interaction information reconstruction, the large model's understanding and response accuracy of the user input information is effectively improved, thereby improving the accuracy of the response information output by the large model; the above technical solution is adopted to solve the problem of low accuracy of the response information output by the large model.

[0036] Optionally, in order to better understand the above step S204, the above step S204 can be implemented in the following way: parse the first interaction information to generate a first parsing result; determine the target keyword of the first interaction information based on the first parsing result; determine the type information corresponding to the first interaction information based on the intention information, and determine the interaction template corresponding to the type information; generate the second interaction information based on the target keyword and the interaction template.

[0037] In the embodiment of the present application, the enhancement engine uses the interaction template to optimize and reconstruct the question description based on the user's original input (first interaction information) and the identified intention information, and generates more accurate, professional and structured second interaction information. Specifically:

[0038] When the user inputs the first interactive information, the enhancement engine first performs parsing to generate a first parsing result. This parsing process includes but is not limited to word segmentation, grammatical analysis, and semantic understanding of natural language, and can then convert the user input into structured information that can be processed by the computer.

[0039] The enhancement engine determines the target keywords of the first interactive information based on the first parsing result. The target keywords are the key to understanding the core of the problem, and they are usually directly related to the subject of the problem, such as "washing machine", "fault code", "repair service", etc. By determining the target keywords, the enhancement engine can more accurately identify user needs.

[0040] The enhancement engine determines the type information corresponding to the first interactive information based on the parsed intent information. The type information may include consultation, fault diagnosis, product query, etc. Each type of information has its own specific interaction template. By matching the correct interaction template, it can ensure that the generated second interactive information is in the correct format and clear structure, in line with industry standards and user expectations. For example, for fault diagnosis issues, the template may contain prompts for key information such as device model, fault phenomenon, and usage environment.

[0041] After the target keywords and interaction templates are determined, the enhancement engine will combine the target keywords and interaction templates to generate the second interaction information. For example, the target keywords will be embedded into the template, necessary details will be added, and the grammatical structure will be adjusted to ensure that the problem description is complete, professional, and easy to understand.

[0042] The embodiment of the present application, through parsing, keyword extraction, type identification and template application, not only optimizes the professionalism and clarity of user questions, but also ensures the structuring and standardization of question descriptions, providing a more accurate and easier to understand input for the big model. This not only improves the accuracy of the response information generated by the big model, but also improves the user experience, reduces the need for multiple rounds of interaction, and enables the system to provide users with the required information and services more quickly and efficiently.

[0043] Optionally, determining the target keyword of the first interaction information based on the first analysis result includes: obtaining historical interaction information of the target object, and parsing the historical interaction information to determine the second analysis result; identifying related keywords in the first analysis result and the second analysis result; and determining the target keyword of the first interaction information among the keywords based on the intention information.

[0044] The enhancement engine not only processes the user's current first interaction information, but also retrieves the historical interaction information of the target object. The analysis of the historical interaction information, i.e., the second analysis result, can provide background knowledge such as the context, preferences, and related questions asked by the user in the past. For example, if the user has previously asked questions about the maintenance of a washing machine, then "washing machine" and "maintenance" may become the target keywords in the historical interaction information.

[0045] After obtaining the parsing results of the user's current and historical interaction information, the enhancement engine will identify the relevant keywords in the two sets of parsing results. This identification process is based on text similarity calculation in natural language processing, and may use technologies such as word vectors, TF-IDF, and semantic similarity measurement. For example, if the user is currently asking about "washing machine fault code" and the historical record contains "washing machine" and "fault", then "washing machine", "fault code", and "fault" will be identified as relevant keywords.

[0046] The enhancement engine will further filter relevant keywords based on the user's intent information to determine the target keywords for the first interactive information. This determination process will give priority to words that are directly related to the current question intent to ensure that the reconstructed question description accurately reflects user needs. For example, if the user's intent is to understand the "washing machine fault code", then "washing machine" and "fault code" will be determined as target keywords, while "maintenance" may not be selected because it has a low correlation with the current question intent.

[0047] For example, suppose a user enters "My washing machine suddenly stopped working and displayed an error code." as the first interactive information on the home appliance consultation platform. The enhancement engine first identifies through analysis that the user's intention is to seek a solution to the washing machine failure, and parses the second analysis result from the historical interaction information (such as the user's previous inquiries about the use and maintenance knowledge of the washing machine). After combining the current and historical analysis results, the enhancement engine identifies relevant keywords including "washing machine", "stopped working", and "error code". Based on the user's intention, the enhancement engine determines "washing machine" and "error code" as target keywords, and may use industry templates to reconstruct the question as: "The washing machine model is XX, and the fault code YY is encountered. Request fault analysis and handling methods." Such second interactive information is not only clear and professional, but also contains the core information that users are concerned about, which helps the large model to accurately understand the problem and provide targeted answers.

[0048] The embodiment of the present application intelligently extracts key information from the user's historical and current interaction information to ensure that the question description is both professional and accurate, reduces the uncertainty of the large model when processing user questions, and improves the accuracy and quality of the final answer.

[0049] Optionally, the second interaction information is generated according to the target keyword and the interaction template, including: writing the target keyword into the interaction template, and determining whether there is a missing field in the interaction template in which the target keyword is written; in the case that there is a missing field in the interaction template in which the target keyword is written, sending prompt information to the target object according to the missing field; filling the missing field according to the second response information sent by the target object to generate the second interaction information.

[0050] The enhancement engine first fills the target keywords determined in the parsing results into the corresponding interaction template. The template contains a fixed structure and some preset fields to standardize the expression of the question. By writing the target keywords, the template can transform the question into a clearly structured and professionally expressed form.

[0051] After the target keywords are written into the interaction template, the enhancement engine checks whether there are any unfilled fields in the template. These missing fields may contain information that is critical for the large model to generate accurate responses, such as the specific model of the device, a detailed description of the fault phenomenon, the user's geographic location, etc.

[0052] If the enhancement engine finds that there are missing fields in the template, it will send a prompt message to the target object (user) to ask the user to provide the missing information. The design of the prompt message should take into account the user experience, ensure the simplicity and clarity of the information, and also guide the user to fill in these missing fields accurately and completely. The user provides a second response information based on the prompt message to supplement the missing key details.

[0053] After collecting the second response information provided by the user, the enhancement engine will use this information to fill in the blank fields in the template, and finally generate a complete, structured, and professionally expressed second interaction information. This information now contains all the details of the user's original intention, as well as the necessary industry background and professional terms, reflecting the user's problem more clearly and accurately.

[0054] For example, suppose a user enters "My washing machine doesn't work" on the home appliance fault consultation platform. The enhanced engine parses the target keywords "washing machine" and "not working", but may recognize that the problem description lacks key information such as the model of the device and the specific fault manifestation. Therefore, it will select a template that is most suitable for washing machine fault diagnosis from the preset interactive templates. This template may contain the following fields: "Device model: ____; Fault phenomenon: ____; Last time used: ____; Have you tried to restart: ____; Have you heard any abnormal sounds: ____; Have you seen an error code: ____;".

[0055] The enhancement engine writes "washing machine" and "not working" into the template, but there are still several empty fields in the template. Therefore, the enhancement engine sends a prompt message to the user, for example: "In order to solve your problem more accurately, please provide the following information: device model, last time used, and whether you see an error code or hear an abnormal sound." The user then replies: "The washing machine model is LG-321, and the last time it was used was yesterday. I didn't see an error code, but I heard a strange sound." Based on the second response information provided by the user, the enhancement engine fills in the empty fields in the template, and the generated second interactive information may be: "Device model: LG-321; Fault phenomenon: unable to start, accompanied by abnormal sounds; last time used: yesterday; Have you tried to restart: No; Have you seen an error code: No; Have you heard an abnormal sound: Yes".

[0056] The embodiment of the present application not only optimizes the problem description, but also ensures the completeness and accuracy of the problem description through user participation in filling in information, thereby improving the efficiency and effectiveness of subsequent processing.

[0057] Optionally, after outputting the first response information corresponding to the second interaction information through the large model, it also includes: receiving evaluation information input by the target object, and determining the target object's satisfaction with the first response information based on the evaluation information; using the first interaction information and the second interaction information as training samples, and determining the sample type of the training sample based on the satisfaction, wherein the sample type includes one of the following: positive sample, negative sample; training the enhancement engine according to the sample type and the training sample.

[0058] After the large model generates the first response information based on the second interaction information and displays it to the user, the system will invite the user to evaluate the answer. The user expresses the degree of approval of the first response information by entering evaluation information, such as "satisfied", "unsatisfied", "very satisfied", etc. The system determines the user's satisfaction with the answer based on this evaluation information.

[0059] The user's original first interaction information (i.e. the user's initial natural language question) and the second interaction information optimized by the enhanced engine will be saved and used as training samples. These samples contain the user's original intention expression and the optimized professional expression, and are an important data source for the enhanced engine's iterative learning.

[0060] Based on the satisfaction of user feedback, the system will mark the training samples as positive samples or negative samples. If the user is satisfied with the first response information, it is marked as a positive sample; if the user is dissatisfied, it is marked as a negative sample. Positive samples represent successful optimization cases, while negative samples indicate deficiencies in the optimization process and require further adjustments.

[0061] The enhancement engine uses these positive and negative samples for iterative learning to improve its problem optimization algorithm. Analysis of positive samples can help the enhancement engine understand which transformation strategies are effective, while analysis of negative samples reveals which areas need improvement. By adjusting algorithm parameters or optimizing template design, the enhancement engine can continuously evolve, improve its accuracy and effectiveness in the problem reconstruction process, and thus improve the user experience.

[0062] For example, a user asks a question on the home appliance fault consultation platform: "My air conditioner has been shutting down automatically recently. What could be the reason?" (first interactive information). After the enhanced engine optimization, the second interactive information generated may be: "Air conditioner type: wall-mounted; brand: Haier; fault phenomenon: automatic shutdown; usage environment: living room; last used time: yesterday; have you tried to restart: yes; have you seen the error code: no".

[0063] Based on the second interactive information, the large model generated possible fault causes and suggestions: "Automatic shutdown may be due to abnormal readings from the indoor sensor, causing the air conditioner to mistakenly believe that the set temperature has been reached and shut down. It is recommended to check whether the indoor sensor is clean and contact professional maintenance personnel if necessary."

[0064] The user is satisfied with the response and gives a positive comment ("very satisfied"). The system marks this case as a positive sample, saves the original first interaction information and the optimized second interaction information, as well as the user's high satisfaction with the answer. By analyzing this positive sample, the enhancement engine may learn that when dealing with the "air conditioner automatically shuts down" problem, the optimization strategy of adding fields such as "usage environment" and "whether you have tried to restart" is effective.

[0065] On the other hand, if the user is dissatisfied with the response of the large model (for example, "The air conditioner fault is not resolved and more specific suggestions are needed"), then this case will be marked as a negative sample. The enhancement engine will analyze the second interaction information and user feedback to find deficiencies in the optimization process, such as asking more detailed questions about the specific manifestations of the fault or providing more specific solution templates, so that a more satisfactory answer can be given the next time a similar problem is encountered.

[0066] In an exemplary embodiment, determining the target object's satisfaction with the first response information according to the evaluation information includes: determining the target object's emotional information according to the evaluation information; and determining the satisfaction according to the emotional information.

[0067] After the user provides the evaluation information, the enhancement engine will further analyze the information to determine the user's emotional inclination. The sentiment analysis here is usually based on natural language processing technology, which determines whether the user's attitude towards the answer is positive, negative or neutral by identifying elements such as emotional words, tone, emoticons, etc. in the text. For example, the user may say "This is very helpful", "Totally useless" or "Maybe", which reflect positive, negative and neutral emotions respectively.

[0068] Based on the user's emotional information, the system can more accurately evaluate the user's satisfaction with the first response information. Generally speaking, positive emotional information means that the user is satisfied with the answer, while negative emotions imply that the user is dissatisfied with the content or form of the answer. Neutral emotional information may require further interaction or information to clarify whether the user has truly found a solution.

[0069] Suppose a user asks a question on the home appliance fault consultation platform: "My refrigerator door doesn't close tightly, what should I do?" The big model generates the first response information based on the optimized question: "The refrigerator door may not close tightly because the sealing strip is aging or damaged. It is recommended to check whether the sealing strip of the refrigerator door is damaged. You can try to clean or replace the sealing strip."

[0070] The user's evaluation of this answer is: "Thank you, I'll try it, I hope it works." This evaluation contains positive sentiment words (such as "thank you") and a certain degree of uncertainty ("I hope it works"). The system's sentiment analysis module may classify this evaluation as slightly positive sentiment, but due to uncertainty, the satisfaction score may not be the highest.

[0071] Further analysis shows that if the user comments again after trying it: "After cleaning the sealing strip, the refrigerator door now closes tightly, which is great!" This comment contains strong positive sentiment words (such as "great") and positive feedback, the system will identify it as a high degree of positive sentiment, thereby determining that the user is very satisfied with the first response information.

[0072] On the contrary, if the user's evaluation is: "The problem still exists after checking the seal, what a waste of time." This evaluation contains strong negative sentiment words (such as "waste of time") and negative feedback, the system will identify it as highly negative sentiment, thereby determining that the user's satisfaction with the first response information is very low.

[0073] By determining user satisfaction through sentiment analysis, the system can understand the user's perception of answer quality in more detail, providing more comprehensive data support for subsequent optimization and learning. It helps the enhanced engine to continuously improve its question optimization strategy, ensuring that the generated second interactive information can more accurately reflect user needs, thereby improving the satisfaction and effectiveness of the answers output by the large model.

[0074] Optionally, after determining the target object's satisfaction with the first response information based on the evaluation information, the method further includes: determining the relationship between the satisfaction level and a preset satisfaction level; when the relationship indicates that the satisfaction level is less than the preset satisfaction level, sending an inquiry message to the target object, wherein the inquiry message is used to indicate improvement suggestions for the target object; receiving third response information input by the target object, and adjusting the enhancement engine according to the third response information.

[0075] The system first compares the user satisfaction determined based on the evaluation information with a preset satisfaction standard. The preset satisfaction can be a specific value, such as 80 points (out of 100 points), or a satisfaction level, such as "satisfied" or "very satisfied". This comparison step helps the system quantify the user's acceptance of the answer and determine whether further improvement measures are needed.

[0076] If the user's satisfaction is lower than the preset standard, the system will take active measures to send an inquiry message to the user, asking for specific suggestions for improvement. Here, the design of the inquiry message is crucial. It should be concise and clear, guiding the user to provide details for improvement, such as "We noticed that you are not satisfied with the previous answer. In what aspects do you want to improve?" or "What specific aspects of your previous answer are you dissatisfied with? How can we do better?"

[0077] The user provides the third response information based on the inquiry information, which is specific suggestions for improvement or details of dissatisfaction. The system needs to carefully collect and analyze these feedbacks to understand the user's expectations and standards for problem solving, and in what ways the system fails to meet these expectations.

[0078] Based on the third response information provided by the user, the system will make corresponding adjustments to the enhancement engine. This adjustment can be fine-tuning of algorithm parameters, optimization of template design, or adding specific industry knowledge to the model. For example, if the user points out that "the answer does not contain the specific meaning of the fault code", the system may need to optimize the model so that it automatically adds a detailed explanation of the fault code when generating an answer.

[0079] In order to better understand the process of the output method of the above-mentioned response information, the implementation method flow of the output of the above-mentioned response information is described below in combination with an optional embodiment, but it is not used to limit the technical solution of the embodiment of the present application.

[0080] In this embodiment, a method for outputting response information is provided. Figure 3 is a system architecture diagram of a method for outputting response information according to an embodiment of the present application, such as Figure 3 As shown, specifically including the following:

[0081] A. User interaction interface module:

[0082] As a bridge between users and the system, this module not only provides an intuitive graphical user interface (GUI), but also supports natural language input, ensuring that users can easily ask questions and provide feedback. It is used to receive and summarize user requirements and display system responses in a friendly and intuitive way, providing users with a seamless interactive experience.

[0083] B. Input preprocessing layer:

[0084] The input preprocessing layer not only receives user input, but also uses advanced data cleaning, format standardization and grammar correction techniques to ensure the accuracy and consistency of the input information.

[0085] C. Tips for enhanced optimization engine (small model):

[0086] The small model uses deep learning and natural language processing technology to perform semantic analysis and contextual understanding of pre-processed user input, and can intelligently identify industry terms, keywords and question structures, thereby reconstructing questions and improving their professionalism and accuracy. The small model's optimization engine can generate question descriptions that are more in line with industry standards based on the user's historical interactions and current intent.

[0087] D. User feedback and iterative learning module:

[0088] The user feedback collector collects all-round user satisfaction feedback on answers, including qualitative evaluation and quantitative scoring, providing direct feedback basis for the accuracy of answers and user satisfaction.

[0089] The accuracy scoring system conducts a comprehensive evaluation of feedback based on sentiment analysis and opinion mining technology, and archives high-scoring questions to provide excellent cases for subsequent small model training. The iterative learning controller dynamically adjusts the small model parameters based on the scoring results and user feedback, achieving self-learning and continuous optimization to ensure that the accuracy and relevance of the model continue to improve.

[0090] E. Data management module:

[0091] This module is used to manage user data, feedback information and model training data, and ensures data security and privacy through measures such as data encryption, privacy protection and integrity checking. At the same time, it also maintains the industry knowledge base, problem template library and algorithm library to ensure that there is rich industry knowledge and template support when the small model is iterated.

[0092] F. Large model processing engine:

[0093] The large model processing engine uses deep learning technology to deeply understand the optimized problems and generate comprehensive, professional and accurate answers. It integrates a large amount of industry knowledge and professional terms, analyzes the depth and breadth of the problems through advanced algorithms, and provides users with authoritative answers.

[0094] G. Output display layer:

[0095] The output display layer is used to display the answers generated by the large model to users in an intuitive and friendly way, while providing real-time display of user feedback and system optimization status, so that users can understand the source of the answers and the continuous improvement of the system. This layer also has the ability to dynamically adjust the output format, and can provide the best reading experience based on the user's device type and preferences, ensuring clear communication of information and user satisfaction.

[0096] The entire system architecture is modularly designed to ensure that all links work closely together, from receiving user needs, preprocessing and optimizing problems, collecting and analyzing feedback, managing data security, to generating professional answers and user-friendly presentations, to achieve an efficient, accurate and user-oriented industry large language model interaction process. This architecture not only improves the efficiency and quality of problem solving, but also promotes user participation and the self-evolution of the model, providing users with a seamless, intelligent and satisfying interaction experience.

[0097] Through the description of the above implementation methods, those skilled in the art can clearly understand that the method according to the above embodiment can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods of each embodiment of the present application.

[0098] Figure 4 is a structural block diagram of a device for outputting response information according to an embodiment of the present application (I); Figure 4 As shown, including:

[0099] The enhancement engine 42 is configured to receive first interaction information input by a target object; perform intent recognition on the first interaction information to obtain intent information corresponding to the first interaction information; and reconstruct the first interaction information according to the intent information and an interaction template to generate second interaction information;

[0100] The large model 44 is used to receive the second interaction information and input it into the large model, and output the first response information corresponding to the second interaction information.

[0101] Through the above-mentioned device, the first interactive information input by the target object is input into the enhancement engine, so that the enhancement engine performs intent recognition on the first interactive information to obtain the intent information corresponding to the first interactive information; the enhancement engine is controlled to reconstruct the first interactive information according to the intent information and the interaction template to generate the second interactive information; the second interactive information is input into the large model to output the first response information corresponding to the second interactive information through the large model. That is, in the embodiment of the present application, by introducing the enhancement engine for intent recognition and interaction information reconstruction, the large model's understanding and response accuracy of the user input information is effectively improved, thereby improving the accuracy of the response information output by the large model; the above-mentioned technical solution is adopted to solve the problem of low accuracy of the response information output by the large model.

[0102] In an exemplary embodiment, the enhancement engine 42 is used to parse the first interaction information to generate a first parsing result; determine a target keyword of the first interaction information based on the first parsing result; determine type information corresponding to the first interaction information based on the intention information, and determine an interaction template corresponding to the type information; and generate the second interaction information based on the target keyword and the interaction template.

[0103] In an exemplary embodiment, the enhancement engine 42 is used to obtain historical interaction information of the target object and parse the historical interaction information to determine a second parsing result; identify related keywords in the first parsing result and the second parsing result; and determine the target keyword of the first interaction information among the keywords based on the intention information.

[0104] In an exemplary embodiment, the enhancement engine 42 is used to write the target keyword into the interaction template, and determine whether there is a missing field in the interaction template in which the target keyword is written; if there is a missing field in the interaction template in which the target keyword is written, send prompt information to the target object according to the missing field; fill the missing field according to the second response information sent by the target object to generate the second interaction information.

[0105] In an exemplary embodiment, Figure 5As shown, the above-mentioned device also includes: a feedback module 52, which is used to receive the evaluation information input by the target object, and determine the satisfaction of the target object to the first response information based on the evaluation information; use the first interaction information and the second interaction information as training samples, and determine the sample type of the training sample based on the satisfaction, wherein the sample type includes one of the following: positive sample, negative sample; train the enhancement engine according to the sample type and the training sample.

[0106] In an exemplary embodiment, the feedback module 52 is used to determine the emotional information of the target object according to the evaluation information; and determine the satisfaction level according to the emotional information.

[0107] In an exemplary embodiment, the feedback module 52 is used to determine the relationship between the satisfaction level and a preset satisfaction level; when the relationship indicates that the satisfaction level is less than the preset satisfaction level, send an inquiry message to the target object, wherein the inquiry message is used to indicate improvement suggestions for the target object; receive third response information input by the target object, and adjust the enhancement engine according to the third response information.

[0108] An embodiment of the present application further provides a storage medium, which includes a stored program, wherein the program executes any of the above methods when it is run.

[0109] Optionally, in this embodiment, the storage medium may be configured to store program codes for executing the following steps:

[0110] S1, inputting first interaction information input by a target object into an enhancement engine, so that the enhancement engine performs intent recognition on the first interaction information to obtain intent information corresponding to the first interaction information;

[0111] S2, controlling the enhancement engine to reconstruct the first interaction information according to the intention information and the interaction template to generate second interaction information;

[0112] S3: input the second interaction information into the large model, so as to output the first response information corresponding to the second interaction information through the large model.

[0113] An embodiment of the present application further provides an electronic device, including a memory and a processor, wherein a computer program is stored in the memory, and the processor is configured to run the computer program to execute the steps in any one of the above method embodiments.

[0114] Optionally, the electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the processor, and the input / output device is connected to the processor.

[0115] Optionally, in this embodiment, the processor may be configured to perform the following steps through a computer program:

[0116] S1, inputting first interaction information input by a target object into an enhancement engine, so that the enhancement engine performs intent recognition on the first interaction information to obtain intent information corresponding to the first interaction information;

[0117] S2, controlling the enhancement engine to reconstruct the first interaction information according to the intention information and the interaction template to generate second interaction information;

[0118] S3: input the second interaction information into the large model, so as to output the first response information corresponding to the second interaction information through the large model.

[0119] Optionally, in this embodiment, the above-mentioned storage medium may include but is not limited to: a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk, and other media that can store program codes.

[0120] An embodiment of the present application further provides a computer program product, which includes a computer program. When the computer program is executed by a processor, the steps in any one of the above method embodiments are implemented.

[0121] An embodiment of the present application further provides another computer program product, including a non-volatile computer-readable storage medium, wherein the non-volatile computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps in any of the above method embodiments are implemented.

[0122] The embodiments of the present application also provide a computer program, which includes computer instructions, which are stored in a computer-readable storage medium; a processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device performs the steps in any one of the above method embodiments.

[0123] Optionally, the specific examples in this embodiment may refer to the examples described in the above embodiments and optional implementation modes, and this embodiment will not be described in detail here.

[0124] Obviously, those skilled in the art should understand that the above modules or steps of the present application can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, and optionally, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, and in some cases, the steps shown or described can be executed in a different order from that herein, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present application is not limited to any specific combination of hardware and software.

[0125] The above is only a preferred implementation of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A method for outputting response information, characterized in that: include: Inputting the first interaction information input by the target object into the enhancement engine, so that the enhancement engine performs intent recognition on the first interaction information to obtain intent information corresponding to the first interaction information; Controlling the enhancement engine to reconstruct the first interaction information according to the intention information and the interaction template to generate second interaction information; The second interaction information is input into the large model, so as to output first response information corresponding to the second interaction information through the large model.

2. The method for outputting response information according to claim 1, characterized in that: Controlling the enhancement engine to reconstruct the first interaction information according to the intention information and the interaction template to generate second interaction information includes: Parsing the first interaction information to generate a first parsing result; Determining a target keyword of the first interactive information according to the first parsing result; Determining type information corresponding to the first interaction information according to the intent information, and determining an interaction template corresponding to the type information; The second interaction information is generated according to the target keyword and the interaction template.

3. The method for outputting response information according to claim 2, characterized in that: Determining a target keyword of the first interactive information according to the first parsing result includes: Acquire historical interaction information of the target object, and parse the historical interaction information to determine a second parsing result; Identifying related keywords in the first parsing result and the second parsing result; A target keyword of the first interaction information is determined from among the keywords according to the intention information.

4. The method for outputting response information according to claim 2, characterized in that: Generating the second interaction information according to the target keyword and the interaction template includes: Writing the target keyword into the interaction template, and determining whether there is a vacant field in the interaction template in which the target keyword is written; When there is a blank field in the interactive template in which the target keyword is written, sending prompt information to the target object according to the blank field; The empty field is filled according to the second response information sent by the target object to generate the second interaction information.

5. The method for outputting response information according to claim 1, characterized in that: After outputting the first response information corresponding to the second interaction information through the large model, the method further includes: receiving evaluation information input by the target object, and determining the satisfaction of the target object with respect to the first response information according to the evaluation information; Using the first interaction information and the second interaction information as training samples, and determining a sample type of the training sample according to the satisfaction level, wherein the sample type includes one of the following: a positive sample and a negative sample; The enhancement engine is trained according to the sample type and the training sample.

6. The method for outputting response information according to claim 5, characterized in that: Determining the target object's satisfaction with the first response information according to the evaluation information includes: Determining the emotional information of the target object according to the evaluation information; The satisfaction level is determined based on the emotion information.

7. The method for outputting response information according to claim 5, characterized in that: After determining the target object's satisfaction with the first response information according to the evaluation information, the method further includes: Determine the relationship between the satisfaction level and a preset satisfaction level; In a case where the size relationship indicates that the satisfaction level is less than the preset satisfaction level, sending an inquiry message to the target object, wherein the inquiry message is used to indicate improvement suggestions for the target object; receiving third response information input by the target object, and adjusting the enhancement engine according to the third response information.

8. A device for outputting response information, characterized in that: include: An enhancement engine, configured to receive first interaction information input by a target object; performing intent recognition on the first interaction information to obtain intent information corresponding to the first interaction information; Reconstructing the first interaction information according to the intention information and the interaction template to generate second interaction information; The large model is used to receive the second interaction information and input it into the large model, and output the first response information corresponding to the second interaction information.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored program, wherein the program executes the method described in any one of claims 1 to 7 when executed.

10. An electronic device comprising a memory and a processor, characterized in that: A computer program is stored in the memory, and the processor is configured to execute the method according to any one of claims 1 to 7 through the computer program.

Images