Auxiliary questioning display system for large language model

Abstract

The invention relates to the technical field of natural language processing and intelligent interaction, in particular to an auxiliary questioning display system for a large language model (LLM), and aims to provide intelligent all-around questioning assistance and personalized information display for a user by integrating powerful language understanding and generating capabilities of the large language model. According to the information displayed in front of the user, the user can refer to or directly retell the question and can quickly obtain related information when asking a question for another core large language model, the experience and efficiency of the user in the information interaction process are improved, and the user experience is improved. The system is widely applied to multiple fields of psychological consultation, family doctors, education, scientific research, commercial consultation, intelligent customer service and the like.

Description

Technical Field

[0001] This invention relates to the fields of natural language processing and intelligent interaction technology, specifically to an auxiliary question display system for large language models. It aims to provide users with efficient, intelligent, and personalized question assistance and display functions, helping users to ask questions correctly and efficiently when using professional large language models. It is widely applicable in various fields such as psychological counseling, family medicine, education, academic research, business consulting, and intelligent customer service. Background Technology

[0002] In today's information-saturated era, people's demand for acquiring and processing information is constantly increasing. When seeking advice, exploring knowledge, or making decisions, users need to ask precise and effective questions to obtain the necessary information. However, because users do not understand the underlying logic and operating principles of large language models, they often face many difficulties when asking questions. These include not knowing how to ask questions, difficulty in accurately expressing their needs, lack of questioning skills, inability to quickly organize the logic of the question, and even forgetting or temporarily forgetting how to ask questions, leading to low efficiency in information acquisition. For example, we have found that many users, when using a psychological counseling intelligent robot (patent number ZL2019105928955), often fail to fully demonstrate the robot's effectiveness and achieve the expected results due to incorrect questioning methods.

[0003] Therefore, developing an auxiliary questioning system display device that can comprehensively assist users in asking questions and intelligently display relevant information about the questions has significant practical implications. Summary of the Invention

[0004] Purpose of the invention:

[0005] The purpose of this invention is to provide an assisted question display system based on a large language model. By integrating the powerful language understanding and generation capabilities of the large language model, it provides users with comprehensive question assistance and personalized information display, helping users to ask high-quality questions more efficiently and display them to the user. Users can refer to or directly repeat the questions. When asking another core large language model, relevant information can be quickly obtained, improving the user's experience and efficiency in the information interaction process.

[0006] Technical solution:

[0007] The auxiliary questioning and display system of this large language model mainly includes a content generation system and a content display system.

[0008] The content generation system is used to generate guiding prompts when users use large language models.

[0009] The content display system is used to display the generated prompts separately, in addition to the display system of the core large language model.

[0010] The content generation system mainly includes the following core modules:

[0011] User Input Module: This module receives initial information from the user, including but not limited to text, voice, and images. For voice input, the system is equipped with advanced speech recognition technology, which can accurately convert the user's speech into text; for image input, image recognition algorithms are used to extract key information from the image and convert it into a text description. The user input module has good compatibility and adaptability, supporting access from various devices and platforms, allowing users to operate anytime, anywhere.

[0012] Large Language Model Processing Module: This module is the core of the system, employing an advanced large language model architecture with powerful language understanding and generation capabilities. Upon receiving information from the user input module, the large language model processing module first performs in-depth analysis to understand the user's intent, needs, and context. Based on the analysis results, the model generates a series of auxiliary question suggestions related to the user's needs. These suggestions not only cover common question expressions but also intelligently adjust and optimize based on the user's personalized characteristics and specific scenarios. For example, in a research scenario, if the user inputs a research direction in a certain field, the model will combine the latest research trends and hot topics in that field to provide the user with forward-looking and targeted question suggestions.

[0013] Meanwhile, the large language model processing module also has knowledge reasoning and association capabilities, which can mine potential related knowledge based on the information input by the user and integrate it into the auxiliary question suggestions to further broaden the user's questioning ideas.

[0014] Learning and Optimization Module: To continuously improve system performance and user experience, this system includes a learning and optimization module. This module records user behavior and feedback, including user selection of auxiliary question suggestions, modifications made, and problem-solving efficiency. Through analysis and mining of this data, the learning and optimization module understands user needs, preferences, and usage habits, and then optimizes and adjusts the parameters and algorithms of the large language model processing module, making the auxiliary question suggestions generated by the system more closely match users' actual needs. Simultaneously, the learning and optimization module regularly updates the system's knowledge base and model parameters to adapt to the ever-changing knowledge environment and user needs, ensuring the system always remains advanced and practical.

[0015] User privacy module: This module primarily ensures user privacy protection, prevents the leakage of private data, and allows users to use the system with peace of mind. It employs encryption technology to protect user data when recording user behavior and feedback information, ensuring user privacy and security.

[0016] Beneficial effects:

[0017] Improving Question Efficiency and Quality: Through intelligent analysis and assistance from a large language model, users can quickly obtain accurate and targeted question suggestions, avoiding the process of blindly asking questions and repeatedly revising them, thus greatly improving question efficiency. At the same time, the question suggestions generated by the system are based on a deep understanding of user needs and contextual information, helping users to ask higher-quality and more in-depth questions, thereby obtaining more valuable information.

[0018] Personalized Service: This system fully considers the individual characteristics and specific scenarios of users, providing each user with tailored assistance in asking questions. Whether you are a professional in a different field or an ordinary information seeker, you can find suitable ways to ask questions and suggestions in the system to meet your diverse needs.

[0019] Continuous learning and optimization: The learning and optimization module enables the system to learn and evolve on its own, continuously optimizing performance based on user feedback and changes in the knowledge environment, always providing users with high-quality service, and ensuring the long-term usability and competitiveness of the system.

[0020] User privacy protection: In an era that places increasing emphasis on privacy protection, this system fully respects user privacy and provides a high level of protection for users' sensitive information and data. Attached Figure Description

[0021] Figure 1 This is a schematic diagram of the overall structure of the question-and-answer display system of the present invention.

[0022] Figure 2 This is an example of an interface for a family doctor scenario, showing only the effect of displaying auxiliary question suggestions in table format.

[0023] Figure 3 This is an example of an interface for a mental health counseling scenario, showing only the effect of displaying auxiliary question suggestions in a table format.

[0024] Figure 4 To illustrate the layout of the system interface and the professional large language model, only the positional relationship between the system interface and the professional large language model is shown. This does not represent the actual position, size, quantity, or other specific data of the displayed system. Detailed Implementation

[0025] The principles and features of this application will be further described in detail below with reference to the accompanying drawings. It should be understood that the specific methods and embodiments described herein are for illustrative purposes only and are not intended to limit the invention. Furthermore, it should be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings.

[0026] Hardware environment:

[0027] The auxiliary question-and-answer display system for this large language model can be deployed on various hardware devices, including but not limited to servers, personal computers, mobile terminals, VR, AR, MR, and XR. Servers should have high-performance processors, large-capacity memory, and storage devices to meet the operational and data processing requirements of the large language model. Personal computers and mobile terminals need to be equipped with appropriate operating systems and input / output devices to ensure convenient operation and interaction for users.

[0028] Software environment:

[0029] The system development utilizes mainstream programming languages ​​and frameworks, such as Python, TensorFlow, or PyTorch. The operating system can be Windows, Linux, or macOS, depending on the actual deployment requirements. Furthermore, the system needs to integrate advanced algorithms and toolkits related to speech recognition, image recognition, and natural language processing to achieve its various functions.

[0030] Implementation steps:

[0031] The system setup and deployment are based on hardware and software requirements, establishing the system's operating environment. The user input module, large language model processing module, user privacy module, and learning and optimization module are integrated and deployed to ensure normal communication and data exchange between these modules.

[0032] Model training and optimization utilizes a large-scale text dataset to pre-train a large language model, enabling it to possess basic language understanding and generation capabilities. Then, for the question-assistance task, relevant domain data and user question data are collected to fine-tune the model, optimizing its performance in specific scenarios. Simultaneously, user feedback data recorded by the learning and optimization module is used to regularly update and optimize the model, continuously improving the system's question-assistance effectiveness.

[0033] Functional testing and debugging: After system deployment, comprehensive functional testing is conducted. This includes simulating different user scenarios and input conditions to check the compatibility and accuracy of the user input module, verifying the rationality and effectiveness of the auxiliary question suggestions generated by the large language model processing module, evaluating the display effects and interactive experience of the question display and interaction module, and ensuring that the learning and optimization module can correctly record and analyze user data. Problems discovered during testing are promptly debugged and fixed to ensure stable and reliable system operation.

[0034] Before the system's official launch, provide users with detailed training materials and operation guides to help them familiarize themselves with the system's functions and usage methods. Promote the system through a combination of online and offline methods to attract more users and further optimize system performance and service quality based on actual user feedback.

[0035] Example 1: Academic Research Scenario

[0036] In the field of academic research, a researcher wanted to learn about the application of quantum computing in cryptography. He entered the initial information "quantum computing cryptography" into the user input receiving module via text input.

[0037] The interaction history module records the input information. The prompt word generation module begins its work, and the semantic analysis submodule performs semantic analysis on the input information, extracting the two key pieces of information, "quantum computing" and "cryptography," and understanding that the researcher's intention is to understand the application of quantum computing in the field of cryptography. The knowledge association submodule searches and matches in the knowledge base of the large language model, associating it with the basic principles of quantum computing (such as qubits and quantum gate operations), common cryptographic algorithms (such as RSA and elliptic curve cryptography), and the potential impact of quantum computing on cryptography (such as the ability of quantum computing to break existing cryptographic algorithms and the development of quantum cryptography), and other related knowledge.

[0038] The prompt content generation submodule generates prompt content based on the analysis results, including suggested questions such as "How does quantum computing pose a threat to the RSA algorithm?" and "What are the advantages of quantum cryptography compared to traditional cryptography?"; keyword expansions such as "Current status of quantum computing technology development" and "Security assessment of cryptographic algorithms"; and question frameworks such as "In [specific application scenarios], what are the impacts of quantum computing on [cryptographic algorithms]?"

[0039] The system displays these prompts to researchers, who can then select topics of interest to further develop their questions. For example, one researcher selected the prompt "How does quantum computing pose a threat to the RSA algorithm?" and added the condition "Under current technological conditions," thus creating the complete question, "Under current technological conditions, how does quantum computing pose a threat to the RSA algorithm?"

[0040] After submitting the question to the large language model, the model returns a detailed answer, including the fundamental method by which Shor's algorithm breaks the RSA algorithm in quantum computing, and the impact of the current level of quantum computing technology development on the actual security of the RSA algorithm. It also records detailed information about the interaction and optimizes the rules for generating prompts based on this information. For example, if it finds that researchers are more interested in questions related to the details of quantum computing algorithms, it can add content related to this to subsequent prompt generation.

[0041] Example 2: Business Consulting Scenario

[0042] A market researcher from a company wanted to understand "the future development trend of the new energy vehicle market." When he input relevant voice information, the voice recognition submodule converted the speech into text information: "The future development trend of the new energy vehicle market."

[0043] The key information extracted, "new energy vehicle market" and "future development trends," reveals that the researchers' intention was to obtain information on the future dynamics and trends of the new energy vehicle market. The knowledge association submodule connects to relevant knowledge such as the technological development of new energy vehicles (e.g., improved battery range, charging infrastructure construction), policy impacts (e.g., subsidy policies, emission standards), and market competition (e.g., market share of major automakers, new entrants). It generates prompts, including suggested questions such as "What is the sales growth forecast for the new energy vehicle market in the next five years?" and "What policy factors will affect the development of the new energy vehicle market?"; keyword expansions such as "Development trends of new energy vehicle sub-markets (e.g., sedans, SUVs)" and "Expansion of new energy vehicles in overseas markets"; and question frameworks such as "Under the [policy environment] and [technological development level], how will the market size and competitive landscape of the [new energy vehicle sub-markets] change?" and "What is the sales growth forecast for the Chinese domestic new energy vehicle market in the next five years?".

[0044] Example 3: Medical scenarios, such as physiological health consultations or specific scenarios involving family doctors.

[0045] A user seeking medical consultation, wanting to understand their health issues, will have their questions displayed by the system after confirming a physiological health consultation. The system will present a series of questions based on the actual medical process. These questions include:

[0046] "I am 30 years old, have no bad habits, and enjoy hiking. Recently, I have been experiencing occasional tearing pain about 5 centimeters below my knee on the inside. The pain does not follow a clear pattern in terms of frequency, but I have not found any external injuries or similar family history. All my physical indicators are normal. What could this be? What could be causing it?"

[0047] Follow-up question: Based on your previous answers, I've basically determined it to be exercise-induced pes anserine bursitis. Could you please provide a detailed treatment plan, and what else should I be aware of? Do you have any other suggestions?

[0048] Follow-up question: How do I perform physical therapy to stretch the pes anserine muscle group? Could you recommend some relevant videos?

[0049] Follow-up question: Does pes anserine bursitis require surgery? In what situations is hospitalization necessary?

[0050] Follow-up question: What is pes anserine bursitis, and what is its pathological mechanism?

[0051] Follow-up question: I am a 30-year-old patient with pes anserine bursitis. How can I recover as quickly as possible and ensure that I can participate in the competition smoothly in one month? Please tailor a scientific and phased rehabilitation plan for me.

[0052] Implementation Case 4: Medical Scenarios, such as specific scenarios of mental health counseling

[0053] A user seeking medical consultation who wants to understand their psychological issues will, upon confirming their interest in mental health counseling, have the question-and-answer system display a set of questions based on the detailed process of an actual psychological consultation. These questions include:

[0054] "I am 40 years old, male, INTJ, with a master's degree, a family of four, smart children, a beautiful wife, a harmonious family, a successful career, and very wealthy materially. Logically speaking, everything is great, but why do I still not feel happy, often suffer from insomnia, and feel that life is meaningless and worthless? (Note: See follow-up questions for further information)."

[0055] Follow-up question A: Could you please explain in detail what Mindfulness-Based Cognitive Therapy (MBCT) mentioned in the answer is?

[0056] Follow-up question B: Please provide me with a complete Existential Anxiety Scale (EAS) and Meaning of Life Scale (MLQ), presented in tabular form, for my psychological assessment.

[0057] Follow-up question C: You previously diagnosed my anxiety as stemming from a midlife meaning crisis. What are the causes of this midlife meaning crisis? Is there any scientific evidence to support this?

[0058] Follow-up question D: Why is it said to be anxiety caused by a midlife crisis of meaning? What are the criteria and standards for this diagnosis, and what scientific principles and theories support this conclusion?

[0059] Follow-up question E: Based on the previous answers, please combine multiple disciplines such as neuroscience, developmental psychology, existential philosophy, and evolutionary psychology to tell me how to resolve the midlife meaning crisis, and provide me with a detailed treatment plan, presented in a table format.

[0060] "Follow-up question: As someone suffering from anxiety due to a midlife meaning crisis, how do you keep a 'meaning log'? Give me a simple and workable implementation plan and remind me to record it twice a week."

[0061] Follow-up question for G: As someone suffering from anxiety due to a midlife crisis of meaning, please recommend some books, movies, and video materials that are helpful for recovery, and explain your reasons for choosing them.

[0062] Follow-up question H: In your previous answer, you mentioned Erikson's psychosocial development theory. Could you please explain this theory? How does this theory help in the psychotherapy of midlife meaning crises?

[0063] "Follow-up Question I: How to analyze the midlife crisis of meaning in neuroexistentialism? Please explain the psychological mechanisms involved."

[0064] "Follow-up question J: As someone suffering from anxiety caused by a midlife meaning crisis, how can one combat death anxiety and resolve a midlife meaning crisis by creating 'idea offspring' (such as books, patents, open-source projects)? Please provide some well-known real-world examples."

[0065] "Follow-up question K: I am an INTJ type person suffering from midlife meaning crisis anxiety. I have already seen significant improvement through the method of creating 'thought offspring' that you mentioned earlier, but I would like to continue to try other treatment methods. Please provide me with a detailed plan."

[0066] "Follow-up question L: I am an INTJ type sufferer with midlife meaning crisis anxiety, and I have basically recovered. What self-adjustment methods and resources would you recommend for me?"

[0067] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

[0068] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

Claims

1. An auxiliary question display system for large language models, characterized in that, include: Content generation system and content display system; The content generation system is used to generate guiding prompts when a user uses a large language model. The content display system is used to display prompts separately, in addition to the display system for the large language model.

2. The content generation system according to claim 1, characterized in that, include: User input module, large language model processing module, learning and optimization module, user privacy module; The system includes a user input module for receiving initial information input by the user, which may include at least one of text, speech, and image. When the received initial information is speech, the user input module converts the speech into text using speech recognition technology. When the received initial information is an image, the user input module extracts key information from the image using an image recognition algorithm and converts it into a text description. A large language model processing module, connected to the user input module, performs in-depth analysis of the received text-based initial information to understand the user's intent, needs, and context, and generates a series of auxiliary question suggestions related to the user's needs based on the analysis results. The large language model processing module also possesses knowledge reasoning and association capabilities, enabling it to mine potential relevant knowledge based on the user's input information and integrate it into the auxiliary question suggestions. The learning and optimization module is connected to the user input module, the large language model processing module, and the user privacy module, respectively. It records user behavior and feedback information, including the user's selection of auxiliary question suggestions, modification content, and problem-solving efficiency. By analyzing and mining the user behavior and feedback information, the module understands the user's needs, preferences, and usage habits, and then optimizes and adjusts the parameters and algorithms of the large language model processing module. The learning and optimization module also regularly updates the system's knowledge base and model parameters.

3. The system according to claim 3, characterized in that, The large language model processing module employs a large language model based on the Transformer architecture. This model is pre-trained on a large-scale text dataset and fine-tuned using domain data and user question data for the auxiliary questioning task. The user privacy module uses encryption technology to protect user data when recording user behavior and feedback information, ensuring user privacy and security. The learning and optimization module uses machine learning algorithms to analyze the usage behavior and feedback information, uncovering user behavior patterns and demand trends. When the learning and optimization module periodically updates the system's knowledge base and model parameters, it uses incremental learning to update only the parts related to new knowledge, reducing the consumption of computing resources and time.

4. The system according to claim 1, characterized in that, The content display system includes: a display platform and media; the shape of the display frame; the number of display frames; the display presentation method; the scene and timing of the content display; and the specific content to be displayed.

5. The content display system according to claim 4, characterized in that, in, The display platform and medium are one of the following: mobile phone, computer, wearable device, naked-eye 3D device, and laser display device; the display frame has two shapes: a shaped frame and a hidden frame. The shaped frame can be a two-dimensional graphic or a three-dimensional graphic. The two-dimensional graphic includes one or more of the following: triangle, quadrilateral, polygon, circle, ellipse, sector, and irregular shape. The three-dimensional graphic includes one or more of the following: cuboid, cube, cylinder, cone, and sphere; the number of display frames is one or more with different functions; the display method is one or more of the following: text, table, image, voice, video, and mind map.

6. The content display system according to claim 4, characterized in that, in, The content is presented in two ways: flow-based or non-flow-based. Flow-based content is further divided into fixed flow and variable flow. Fixed flow displays a constant problem according to the actual process, while variable flow displays content that changes in real time according to the process.

7. The content display system according to claim 4, characterized in that, in, The scenarios and timing of the content presentation are divided into one of the following: medical scenarios, educational scenarios, legal scenarios, scientific research scenarios, and business consulting scenarios. The medical scenario is further divided into three types: mental health counseling, physical health counseling, and family doctor.

8. The scenario and timing for displaying content according to claim 7, characterized in that, In the medical setting, based on the basic workflow, the specific question types at each stage can be categorized into one or more of the following: structured questions, questions focusing on specific situations, questions requesting tool-based support, questions establishing continuous dialogue, metaphorical expression questions, and open-ended exploratory questions.

9. The scenario and timing for displaying content according to claim 7, characterized in that, Among them, medical The process of psychological health counseling in a specific setting can be divided into one or more of the following stages: preparation stage, psychological assessment stage, planning stage, psychological intervention stage, effect evaluation and adjustment stage, and follow-up and termination stage.

10. The scenario and timing for displaying content according to claim 7, characterized in that, Physiological health consultations and family doctors in medical settings follow a medical process that includes one or more of the following stages: initiation stage, consultation stage, examination stage, diagnosis and treatment stage, treatment and observation stage, follow-up visit and adjustment stage, termination stage, and follow-up work.

Images