Intelligent question and answer and recommendation integrated system and method
Through integrated system design, the question-answering and recommendation modules work collaboratively. By utilizing a shared data bus and multimodal data processing, a real-time feedback loop is constructed, solving the problem of the separation between question-answering and recommendation in existing technologies and improving the system's service accuracy and user experience.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-26
- Publication Date
- 2026-04-03
AI Technical Summary
Existing intelligent question-answering systems and personalized recommendation systems are often disconnected, lacking real-time collaboration, underutilizing multimodal information, and experiencing slow feedback loops. This results in inconsistent service experiences, recommendations that lack real-time relevance and contextual relevance, and an inability to achieve highly accurate user interaction.
By designing an integrated intelligent question-answering and recommendation system, a shared data bus is used to connect the question-answering and recommendation modules, integrate multimodal data processing, construct a real-time feedback closed loop, realize the collaborative work of question answering and recommendation, and use user feedback for online optimization.
It enables collaborative work between question answering and recommendation, improves the initiative and accuracy of services, enhances the utilization of multimodal data, shortens the feedback loop, and provides an end-to-end adaptive intelligent interaction solution.
Smart Images
Figure CN121786089A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of artificial intelligence, and in particular to an integrated intelligent question answering and recommendation system and method. Background Technology
[0002] Currently, intelligent question-answering systems and personalized recommendation systems are usually designed and deployed independently.
[0003] In intelligent question-answering scenarios, related technologies often employ retrieval-based QA or generative QA using pre-trained language models. The system receives the user's text question and provides an answer by retrieving relevant knowledge bases or directly generating text. For example, in agricultural machinery consultation, if a user enters "Why can't the tractor start?", the system will return a list of possible reasons or an explanatory text. These systems focus on providing accurate answers but typically do not delve deeper into the user's underlying needs or offer solutions.
[0004] In personalized recommendation scenarios, related technologies widely rely on algorithms such as collaborative filtering and content-based recommendation. These systems primarily analyze users' historical behavioral data (such as past searches, clicks, and purchase records) and item attributes to infer products or services that users might like. For example, an e-commerce platform might recommend similar agricultural machinery parts based on a user's past purchases. This type of recommendation is based on long-term interest profiles and lacks a direct, dynamic connection to the specific question the user is currently asking.
[0005] Furthermore, existing solutions often treat multimodal inputs (such as voice and images) as independent functional modules. For example, a system may have a separate speech-to-text service or image recognition service, but the recognized results are merely used as simple input replacements and are not deeply integrated into a unified framework that understands user intent and makes comprehensive decisions.
[0006] Therefore, how to achieve synergy between question answering and recommendation, and improve the accuracy of question answering and recommendation, has become an urgent problem to be solved. Summary of the Invention
[0007] The purpose of this application is to provide an integrated intelligent question-answering and recommendation system and method, which can realize the collaborative work of question answering and recommendation and improve the accuracy of question answering and recommendation.
[0008] To achieve the above objectives, this application provides the following solution: In a first aspect, this application provides an integrated intelligent question-answering and recommendation system, including: an interactive display UI function module and a data function module, a question-answering function module, a recommendation function module and a user feedback function module connected to the interactive display UI function module; Both the question-and-answer module and the recommendation module are connected to the data module via a shared data bus; the shared data bus is a data bus for sharing context and user profiles; the user feedback module is connected to the data module. The interactive UI module is used to input the user's questions. The data function module is used to process the data of the question and send the processed question data to the question-and-answer function module and the recommendation function module; The question-and-answer function module is used to analyze and answer the processed question data based on a large model, obtain the answer to the question, and send the answer to the data function module; The recommendation function module is used to generate recommendation suggestions based on the processed problem data using a machine learning model, and then send the recommendation suggestions to the data function module. The user feedback function module is used to generate feedback data based on the user's feedback actions, and send the feedback data to the data function module; The data function module is also used to process the feedback data and send the processed feedback data to the question-and-answer function module and the recommendation function module to optimize the answer to the question and the recommendation suggestions.
[0009] In one embodiment, the data function module integrates different types of recognition components; the data function module is used to perform multimodal data fusion and normalization processing on the question, and send the processed question data to the question-answering function module and the recommendation function module.
[0010] In one embodiment, the data function module integrates a speech recognition component and an image recognition component.
[0011] In one embodiment, the data function module is further used to determine whether the current user is online. If so, it waits for the user to input a new question or waits to receive new feedback data. If not, it organizes the received data.
[0012] In one embodiment, the feedback action includes: the user clicking a recommended item or clicking an ignore item in the interactive display UI function module; the recommended item is generated based on the recommendation suggestion.
[0013] Secondly, this application provides an integrated method for intelligent question answering and recommendation, including: Obtain processed problem data; processed problem data is obtained by processing the user's inquiry. Based on the large model, the processed problem data is analyzed and solved to obtain the answer to the problem; Based on machine learning models, recommendations are generated from the processed problem data. When processed feedback data is received, the answer to the question and the recommended suggestions are optimized based on the processed feedback data; the processed feedback data is obtained by processing the feedback data generated based on the user's feedback actions.
[0014] In one embodiment, the method for determining the processed problem data specifically includes: The user's inquiry is processed through multimodal data fusion and normalization to obtain processed inquiry data.
[0015] In one embodiment, after generating recommendation suggestions, the integrated intelligent question answering and recommendation method further includes: Determine if the user is currently online. If so, wait for the user to enter a new question or wait to receive new feedback data. If not, process the received data.
[0016] According to the specific embodiments provided in this application, this application has the following technical effects: This application provides an integrated intelligent question-answering and recommendation system and method. Both the question-answering and recommendation modules are connected to a data module via a shared data bus, which serves as a data bus for sharing context and user profiles. A user feedback module is also connected to the data module. The question-answering and recommendation modules receive question data processed by the data module in parallel. Both modules can make decisions based on a unified, real-time updated user state, thereby achieving logical complementarity and enhancement between answers and recommendations in terms of intent, realizing collaborative work between question answering and recommendation. The user feedback module generates feedback data based on user actions and feeds it back to the data module in real time. This feedback is then used by the question-answering and recommendation modules to immediately adjust model outputs or strategies, achieving a rapid closed loop of "perception-decision-feedback-optimization" and improving the accuracy of question answering and recommendation. Attached Figure Description
[0017] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0018] Figure 1 This is a schematic diagram of the structure of an integrated intelligent question-answering and recommendation system provided in an embodiment of this application; Figure 2 A flowchart illustrating the implementation process of an integrated intelligent question-answering and recommendation system provided in this application embodiment; Figure 3 A flowchart for adjusting model parameters using idle resources provided in this application embodiment; Figure 4 This is a flowchart illustrating an integrated intelligent question-answering and recommendation method provided in an embodiment of this application. Detailed Implementation
[0019] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0020] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.
[0021] In this application, all actions to obtain information or data are carried out in compliance with the relevant data protection laws and policies of the country where the application is located, and with the authorization of the owner of the relevant device.
[0022] Currently, the implementation methods for intelligent question answering and personalized recommendation scenarios have the following main drawbacks: (1) Service fragmentation and experience fragmentation.
[0023] The question-answering system is only responsible for answering questions, and the recommendation system is only responsible for recommending products; the two are independent in terms of business logic and data flow. This results in a disjointed user experience, where after obtaining an answer, users must manually search again or rely on a static recommendation list unrelated to the current consultation context if they need specific products or services related to the solution. The root cause is that the system architecture treats question answering and recommendation as two separate tasks, lacking a collaborative mechanism for sharing a real-time conversation context and a unified understanding of user intent.
[0024] (2) The recommendations lack real-time and scenario-specificity.
[0025] Related recommendation systems rely on historical behavioral data and cannot capture and respond to the latest and most urgent needs revealed in the user's current consultation session. For example, if a user asks, "What should I do if my rice gets leaf blight?", the system can provide an answer, but it cannot immediately and proactively recommend specific pesticides or spraying equipment for treating the disease. This is because the triggering and calculation logic of the recommendation module is not deeply integrated with the real-time output of the question-and-answer module, failing to use the "question" itself and its answer as the most direct and effective input signal for the recommendation system.
[0026] (3) Insufficient utilization of multimodal information.
[0027] While some systems support voice or image input, they typically only convert it to text and then hand it over to subsequent processes. This fails to fully leverage the rich information contained in multimodal data (such as visual features in images and emotional tone in speech) to enhance understanding of the user's context and improve the accuracy of question answering and recommendations. The reason for this is that the data processing module has a limited functionality and fails to deeply couple multimodal recognition technology with core semantic understanding and decision-making logic, lacking a unified data processing and fusion layer.
[0028] (4) The feedback loop is slow and the system has weak adaptive capability.
[0029] User feedback on recommendations (such as clicks or ignoring) is typically recorded for batch retraining of the model days or even weeks later, making it impossible to immediately optimize subsequent interactions within the current session. For example, even if a user ignores a recommendation, the system may still recommend similar items in the same conversation. This is because the system architecture lacks a real-time feedback loop, feedback data is decoupled from the decision-making process of the online service, and the model update cycle is long, hindering rapid online learning and behavior adjustment.
[0030] Based on the above problems, this application aims to solve the following key technical problems: (1) How to break down the barriers between question-answering and recommendation systems and achieve deep collaboration between the two in the context of real-time conversation, so that recommendations can be dynamically generated based on the user's current questions and intentions, thereby improving the initiative and accuracy of the service. (2) How to achieve a unified understanding and utilization of the user's multimodal input (voice, text, and images) and transform it into structured information that can be used to enhance the quality of question-answering and recommendation. (3) How to build an efficient real-time feedback loop that can capture user behavior feedback (such as clicks on recommended items) in real time and use this feedback data to dynamically adjust subsequent question-answering strategies and recommendation results, thereby achieving online self-optimization of the system. (4) How to integrate the above capabilities into one through an integrated system architecture, providing an end-to-end, adaptive, multimodal intelligent interaction solution to overcome the problems of functional dispersion and slow response in the existing technology.
[0031] In one exemplary embodiment, such as Figure 1 As shown, an integrated intelligent question-answering and recommendation system is provided, including: an interactive display UI function module and a data function module, a question-answering function module, a recommendation function module and a user feedback function module connected to the interactive display UI function module.
[0032] Both the question-and-answer module and the recommendation module are connected to the data module via a shared data bus; the shared data bus is a data bus for sharing context and user profiles; the user feedback module is connected to the data module.
[0033] The interactive display UI module is used to input the questions that users ask.
[0034] The data function module is used to process the question data and send the processed question data to the question-and-answer function module and the recommendation function module.
[0035] The question-and-answer function module is used to analyze and answer processed question data based on a large model, obtain the answer to the question, and send the answer to the data function module. The large model includes: deepseek, Doubao, LLaMA, Hunyuan, Pangu, and Tongyi Qianwen, etc.
[0036] The recommendation module generates recommendations based on processed problem data using a machine learning model, and then sends these recommendations to the data module. The recommendation module can use a large model to generate recommendations based on the processed problem data; alternatively, it can directly use a dedicated machine learning recommendation module. Specifically, the recommendation module uses a large model or a dedicated machine learning recommendation module to analyze user characteristics and the current user's problems and difficulties based on the processed problem data, and then recommends services or products to address these problems and difficulties, generating recommendation suggestions.
[0037] The user feedback function module is used to generate feedback data based on the user's feedback actions and send the feedback data to the data function module.
[0038] The data function module is also used to process the feedback data and send the processed feedback data to the question-and-answer function module and the recommendation function module to optimize the answer to the question and the recommendation suggestions.
[0039] The intelligent question-answering and recommendation integrated system described in the above embodiment forms a software system. The user first inputs the question to be asked in the interactive display UI function module; the data function module cleans and organizes the data of the question raised by the user, and sends the organized data to the question-answering function module and the recommendation function module; the question-answering function module analyzes and answers the question raised by the user.
[0040] Meanwhile, the Q&A module in the background analyzes and organizes the data of questions previously asked by users to prepare answers for the user's next question; the recommendation module also uses the previous question data to analyze user characteristics and the current problems and difficulties of users, and recommends some services or products in response to these problems and difficulties; the user feedback module uses whether users click on products as user feedback to generate new feedback data, which is sent to the data module for improvement of the recommendation algorithm.
[0041] In another exemplary embodiment of this application, the data function module integrates different types of recognition components; the data function module is used to perform multimodal data fusion and normalization processing on the question, and send the processed question data to the question-answering function module and the recommendation function module.
[0042] In another exemplary embodiment of this application, the data function module integrates a speech recognition component and an image recognition component.
[0043] In another exemplary embodiment of this application, the data function module is further used to determine whether the current user is online. If so, it waits for the user to input a new question or waits to receive new feedback data. If not, it organizes the received data.
[0044] In another exemplary embodiment of this application, the feedback action includes: the user clicking a recommended item or clicking an ignore item in the interactive display UI function module; the recommended item is generated based on the recommendation suggestion.
[0045] The following is combined with Figure 2 Taking a user's question-and-answer consultation task as an example, the implementation process of the above-mentioned integrated intelligent question-and-answer and recommendation system is described below: Step 1: Users first log in to this system and consult relevant matters through the system's output interface in the large model question-answering agent.
[0046] For example, if you are asking questions about agricultural machinery, you can enter your questions in various forms, such as text, voice, or images.
[0047] Step 2: The data function module will convert the input voice, text, and image information into standardized text data; the data function module will also use components such as voice recognition and image recognition to convert the input information into a certain standardized data format for analysis and processing by the question-answering function module and the recommendation function module.
[0048] Step 3: The data function module sends this text data to the question-and-answer function module and the recommendation function module respectively.
[0049] Step 4: The question-and-answer module and the recommendation module use user input data to provide corresponding answers or recommendations.
[0050] Step 5: The user provides feedback on the suggestions.
[0051] For example, if a user selects a recommended suggestion, it means that the user agrees with the suggestion; if a user does not select a suggestion, it means that the user does not agree with the suggestion.
[0052] Step 6: The user feedback function module obtains and generates feedback data.
[0053] The feedback module collects and analyzes user feedback information to form a relatively standardized user opinion feedback information.
[0054] Step 7: The data function module will convert the feedback data into standardized text data.
[0055] Simultaneously, the feedback data is transmitted to the question-and-answer module and the recommendation module. These modules will then use this feedback to improve the model and make adjustments accordingly.
[0056] Step 8: Determine if the current user is online.
[0057] After answering a question, the system will wait for the user to enter a new question or provide new feedback. If a user remains offline for an extended period, the system will assume the user has logged off and will close the service thread for that user.
[0058] Step 9: After the user logs off, the system will enter the background to organize and process the data.
[0059] At this point, the entire task is complete. When a new task arrives, steps 1 to 9 above will be repeated.
[0060] The intelligent question-answering and recommendation integrated system described in the above embodiments realizes a collaborative working mechanism between the question-answering and recommendation modules, a multimodal data fusion and unified processing flow, an online optimization mechanism based on real-time user feedback, and a dynamic resource management strategy based on user status.
[0061] (1) Collaborative working mechanism of question answering and recommendation modules: The question answering module and the recommendation module receive question data from the same user after processing by the data module in parallel, and analyze it using shared user history conversation data. The question answering module focuses on generating direct answers, while the recommendation module simultaneously analyzes user characteristics and the difficulties and needs implied in the current question to generate relevant service or product recommendations.
[0062] Specifically, by sharing contextual and user profile data, the two modules can make decisions based on a unified, real-time updated user state, thereby achieving complementarity and enhancement between answers and recommendations in terms of logic and intent.
[0063] (2) Multimodal data fusion and unified processing flow: The data function module integrates components such as speech recognition and image recognition, which transform different types of input information into standardized text data formats, providing a unified and high-quality data foundation for subsequent question answering and recommendation analysis.
[0064] Specifically, components such as speech recognition and image recognition are integrated into a front-end multimodal-to-text conversion layer. This conversion layer normalizes heterogeneous data, allowing the core question-answering and recommendation logic based on a large language model to focus on semantic understanding and generation. This simplifies the system architecture and improves the versatility of processing capabilities.
[0065] (3) Online optimization mechanism based on real-time user feedback: The user feedback function module transforms the user's explicit behavior (such as clicking on recommended items) into structured feedback data and feeds it back to the data function module in real time, which is then used by the question and answer and recommendation function modules to immediately adjust the model output or strategy.
[0066] Specifically, by utilizing feedback data as reward signals in reinforcement learning or training samples for online learning, the system can fine-tune its behavior in real time without complete retraining, achieving a rapid closed loop of "perception-decision-feedback-optimization." User feedback actions include: expressing dissatisfaction, ignoring recommendations, and clicking on recommendations. The user feedback module structures these actions, and the data module normalizes the structured feedback data, converting it into standardized text data. The user input and the standardized text data obtained after processing the feedback data are used as reward signals in reinforcement learning or training samples for online learning. This allows for fine-tuning of model parameters in the large models of the question-answering and recommendation modules, achieving online optimization.
[0067] (4) Dynamic resource management strategy based on user status: The system dynamically starts or stops the model training thread by judging the user's online status, and automatically triggers the background data processing and model calculation tasks after the user goes offline.
[0068] Specifically: By managing session lifecycle and recognizing session state, computing resources are allocated rationally to ensure response speed during online services, while idle resources are utilized for offline computing, thereby improving the overall resource utilization efficiency and scalability of the system. For example... Figure 3 As shown, the calculation process is as follows: The system determines the number of users online and the availability of resources. If users are online and there are many users (more than the set number) and the resource utilization rate is high (greater than the set percentage), then the system performs inference calculations using a large model to answer user questions and make recommendations, i.e., only inference without training. If the resource utilization rate is low (less than or equal to the set percentage) and the idle rate is high, then the system cleans and organizes the data, trains the model, and adjusts the model's parameters to optimize the answers to questions and the recommendations.
[0069] Based on the same inventive concept, this application also provides an integrated intelligent question-answering and recommendation method for implementation using the aforementioned integrated intelligent question-answering and recommendation system. The solution provided by this method is similar to the implementation scheme described in the above system; therefore, the specific limitations in one or more embodiments of the integrated intelligent question-answering and recommendation method provided below can be found in the limitations of the integrated intelligent question-answering and recommendation system described above, and will not be repeated here.
[0070] In one exemplary embodiment, such as Figure 4 As shown, an integrated intelligent question answering and recommendation method is provided, including: Step 401: Obtain the processed problem data; the processed problem data is obtained by processing the user's inquiry.
[0071] Step 402: Based on the large model, analyze and solve the processed problem data to obtain the answer to the problem.
[0072] Step 403: Based on the machine learning model, generate recommendation suggestions according to the processed problem data.
[0073] Step 404: When the processed feedback data is received, optimize the answer to the question and the recommended suggestions based on the processed feedback data.
[0074] The processed feedback data is obtained by processing the feedback data generated based on the user's feedback actions.
[0075] As an optional implementation method, the method for determining the processed problem data specifically includes: The user's inquiry is processed through multimodal data fusion and normalization to obtain processed inquiry data.
[0076] As an optional implementation, after generating recommendation suggestions, the integrated intelligent question-answering and recommendation method further includes: Step 405: Determine if the current user is online. If yes, wait for the user to input a new question or wait to receive new feedback data. If no, organize the received data.
[0077] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0078] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.
Claims
1. An integrated intelligent question-answering and recommendation system, characterized in that, The intelligent question-answering and recommendation integrated system includes: an interactive display UI function module and a data function module, a question-answering function module, a recommendation function module, and a user feedback function module connected to the interactive display UI function module; Both the question-and-answer module and the recommendation module are connected to the data module via a shared data bus; the shared data bus is a data bus for sharing context and user profiles; the user feedback module is connected to the data module. The interactive UI module is used to input the user's questions. The data function module is used to process the data of the question and send the processed question data to the question-and-answer function module and the recommendation function module; The question-and-answer function module is used to analyze and answer the processed question data based on a large model, obtain the answer to the question, and send the answer to the data function module; The recommendation function module is used to generate recommendation suggestions based on the processed problem data using a machine learning model, and then send the recommendation suggestions to the data function module. The user feedback function module is used to generate feedback data based on the user's feedback actions, and send the feedback data to the data function module; The data function module is also used to process the feedback data and send the processed feedback data to the question-and-answer function module and the recommendation function module to optimize the answer to the question and the recommendation suggestions.
2. The intelligent question-answering and recommendation integrated system according to claim 1, characterized in that, The data function module integrates different types of recognition components; the data function module is used to perform multimodal data fusion and normalization processing on the question, and send the processed question data to the question-answering function module and the recommendation function module.
3. The integrated intelligent question-answering and recommendation system according to claim 2, characterized in that, The data function module integrates speech recognition and image recognition components.
4. The intelligent question-answering and recommendation integrated system according to claim 1, characterized in that, The data function module is also used to determine whether the current user is online. If so, it waits for the user to input a new question or waits to receive new feedback data. If not, it organizes the received data.
5. The intelligent question-answering and recommendation integrated system according to claim 1, characterized in that, The feedback action includes: the user clicking on a recommended item or clicking on an ignored item in the interactive display UI function module; the recommended item is generated based on the recommendation suggestion.
6. A method for integrating intelligent question answering and recommendation, characterized in that, The integrated intelligent question answering and recommendation method includes: Obtain processed problem data; processed problem data is obtained by processing the user's inquiry. Based on the large model, the processed problem data is analyzed and solved to obtain the answer to the problem; Based on machine learning models, recommendations are generated from the processed problem data. When processed feedback data is received, the answer to the question and the recommended suggestions are optimized based on the processed feedback data; the processed feedback data is obtained by processing the feedback data generated based on the user's feedback actions.
7. The integrated intelligent question answering and recommendation method according to claim 6, characterized in that, The methods for determining properly processed problem data specifically include: The user's inquiry is processed through multimodal data fusion and normalization to obtain processed inquiry data.
8. The integrated intelligent question answering and recommendation method according to claim 6, characterized in that, After generating recommendations, the integrated intelligent question answering and recommendation method further includes: Determine if the user is currently online. If so, wait for the user to enter a new question or wait to receive new feedback data. If not, process the received data.