AI dialogue system and method based on bidirectional interactive form

Abstract

The invention relates to the technical field of computers, and discloses an AI dialogue system and method based on a bidirectional interactive form. According to the method, through dynamic generation of the structured form and an intelligent pre-filling mechanism, the user interaction burden and the prompt word engineering threshold are remarkably reduced, and the problems of passive interaction, low multi-round efficiency, cooperative fault among models and the like in the prior art are solved. The system can automatically identify the information gap and generate the optimal form, supports multi-modal input and cross-model cooperation, and ensures the integrity and accuracy of information transmission. Meanwhile, through a feedback-driven self-optimization mechanism, the adaptation degree of a form generation strategy and a template is continuously improved, and efficient, accurate and extensible intelligent interaction experience is achieved.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to an AI dialogue system and method based on a two-way interactive form. Background Art

[0002] Currently, AI interaction systems mainly rely on natural language dialogue mode, where users interact with the model through open-ended questions or instructions. The core of this interaction method is that users input prompts, and the model generates responses based on the context. Although this model performs well in simple scenarios, it has significant defects when dealing with complex tasks, as shown below: Interaction passivity and cue word dependence: The response quality of existing systems is highly dependent on the completeness and accuracy of user input. If the user does not provide sufficient contextual information (such as target scenarios, constraints, preference parameters, etc.), it is difficult for the model to generate accurate results. For example, in a medical consultation scenario, if the user does not actively explain the duration of symptoms or past medical history, the model may only provide generalized suggestions rather than personalized diagnosis.

[0003] Multiple rounds of interaction are inefficient: Existing systems usually complete information through multiple questioning mechanisms, which increases the number of interaction rounds and significantly delays tasks. For example, in AI agent collaboration scenarios, if the task planning model does not clearly specify the data format requirements, the execution model may return unparseable results that require repeated corrections, greatly reducing the efficiency of task execution.

[0004] Information gap in collaboration between models: In the multi-model collaborative work scenario, the lack of dynamic information supplementation protocols between agents leads to a break in the information transmission chain. Experiments show that when the task decomposition model does not pass resource priority parameters, the probability of the scheduling model incorrectly allocating computing power increases by 62%. This information gap seriously affects the execution effect of complex tasks.

[0005] Limitations of long context processing: Although some systems support long text input, it is difficult for users to predict all the information they need to provide, and redundant descriptions may interfere with the extraction of core intent. For example, in a business analysis scenario, users may provide a large amount of irrelevant background information, making it difficult for the model to focus on key requirements.

[0006] Existing technologies attempt to alleviate the above problems through templated interactions or predefined protocols, but there are still bottlenecks such as insufficient flexibility and poor cross-domain adaptability. For example, a Gartner report pointed out that 73% of companies have a lower-than-expected return on investment (ROI) for AI applications due to improper prompt word design. In addition, traditional methods do not solve the coupling problem between dynamic information completion and multi-model collaboration, making it difficult to achieve efficient interaction in complex scenarios.

[0007] In summary, the existing technology has significant defects such as passive interaction, low efficiency, and collaboration gap when processing complex tasks. There is an urgent need for a new interactive system that can lower the threshold of prompt word engineering, dynamically complete information, and support efficient cross-model collaboration. For this purpose, an AI dialogue system and method based on a two-way interactive form are proposed. Summary of the invention

[0008] In view of the deficiencies of the prior art, the present invention provides an AI dialogue system and method based on a two-way interactive form to solve the background technology problems.

[0009] In the first aspect, to achieve the above-mentioned purpose, the present invention provides the following technical solution: an AI dialogue system based on a two-way interactive form, comprising the following contents: Information integrity analysis unit, used to parse user input, identify intent and entities, compare the complete information model in the domain knowledge base, mark missing information items and calculate priorities; The form generation engine is used to select the self-generation mode, service call mode or mental model mapping mode according to the type of information gap, generate a structured interactive form, and automatically assign a unique variable identifier to each field. The naming rule is "scenario-field type-hash value", and the variable identifier is mapped with the semantic description and stored in the global variable registry; Multimodal interactive interface, used to support multiple input methods such as text, voice, and image to fill in the form, including drop-down selection and fill-in-the-blank input in text mode, voice-to-text filling in voice mode, and feature extraction and key information analysis in image mode; Cross-model collaboration adapter, which is used to encapsulate form data into a machine-readable standardized format in the interaction scenario between AI agents, and attach data validation rules to ensure the integrity of information transmission; Intelligent pre-fill module, which is used to automatically fill in form fields based on user device type, current activity and data access permissions, and supports multiple pre-fill rules including but not limited to GPS positioning, enterprise CRM system data call, and wearable device data acquisition; A feedback-driven self-optimization system is used to record users' modification behaviors on pre-filled fields, build field correction heat maps, dynamically adjust pre-filling strategies, and evaluate template adaptability based on large model output quality to achieve template evolution and semantic gap detection.

[0010] Preferably, the generation strategy of the form generation engine includes: Self-generation mode, which is used to extract necessary information dimensions based on semantic analysis when there is no predefined template in the problem domain, dynamically construct form fields, and automatically generate natural language sentences to embed variables in a logical order; Service call mode, used to connect to external systems and directly call industry-standard form templates, which contain variable placeholders; The thinking model mapping mode is used to match the problem type to the professional analysis framework, generate information collection items according to the framework requirements, and bind variable identifiers.

[0011] Preferably, the multimodal interaction interface includes: Text mode is used to provide structured controls such as drop-down selection and fill-in-the-blank input. The control value is dynamically bound to the variable identifier. Voice mode, which is used to convert voice descriptions into form field values ​​and automatically map them to corresponding variable identifiers; Image mode is used to automatically parse key information in uploaded images through feature extraction technology and bind the parsing results to variable identifiers.

[0012] Preferably, the prefilling rules of the intelligent prefilling module include: If the form contains a "geolocation" field and the user authorizes location, the GPS coordinates are automatically filled in and converted to a standard address format and bound to the variable identifier; If it is detected that the user is using a corporate account, the customer number and project name are automatically pre-filled from the internal CRM system and bound to the variable identifier; In medical emergency scenarios, wearable devices are used to obtain heart rate and blood oxygen data, complete health monitoring forms, and bind them to variable identifiers; In the search scenario, public information or complex information in the database is obtained by searching the big model, and possible answers are pre-filled into the form.

[0013] Preferably, the cross-model collaboration adapter includes: Define machine-readable metadata specifications, including field descriptions, validation rules, and urgency indicators; When AI agents interact with each other, the prefill assistant can initiate data requests to other agents, and automatically calculate the recommended values ​​after the responding agent returns the relevant data; Supports step-by-step filling and mid-course correction, and real-time verification of data validity.

[0014] Preferably, the feedback-driven self-optimization system comprises: Record users’ modification behaviors on pre-filled fields, build field modification heat maps, and identify high-frequency modification fields; Dynamically adjust prefill strategy based on correction patterns; Evaluate template adaptability based on large model output quality, and automatically downgrade low-scoring templates; When a template triggers user questions continuously, mark the variable fields that need to be supplemented.

[0015] In the second aspect, an AI dialogue method based on a two-way interactive form is implemented according to the AI ​​dialogue system based on a two-way interactive form described in the first aspect. The following method steps are implemented: Information gap detection phase: parse the input content, identify intent and entities, compare the complete information model in the domain knowledge base, mark missing items and calculate priorities; Form generation decision stage: First check whether there is an industry standard form that can be called. If there is no matching template, select the thinking model to generate fields according to the question type, or independently create a dynamic form containing required / optional fields; Interactive completion stage: Return the form to the user or question model, clearly mark the required items, provide input format suggestions, and automatically fill in the fields according to the pre-fill rules; Data integration and response phase: The form data is integrated with the original input to generate a complete context, triggering the AI ​​model to recalculate and output accurate responses, while recording missing patterns to optimize subsequent form generation strategies.

[0016] Preferably, the form generation decision stage further includes: Assign a unique variable identifier to each form field, using the naming convention of "scenario-field type-hash value"; Call the prompt word template in the preset template library, which contains variable placeholders; For new scenarios, natural language sentences are automatically generated based on field semantics, and variables are embedded in a logical order.

[0017] Preferably, the interactive completion stage further includes: Return the form to the user or question model, clearly mark the required fields, input format suggestions, and bind the input data to the corresponding variable identifiers; Intelligently adjust the form display method based on the user's device type and current activity; Verify user data access permissions and automatically populate fields within the authorized scope.

[0018] Preferably, the data integration and response phase further includes: Dynamically match form data with external database fields based on variable identifiers; Calling the intelligent prompt word synthesis engine to generate domain-limited structured query instructions; Trigger the large language model to recalculate and output accurate responses, while recording missing patterns to optimize subsequent form generation strategies.

[0019] Compared with the prior art, the present invention has the following beneficial effects: The present invention provides an AI dialogue system and method based on a two-way interactive form. By dynamically generating structured forms and intelligent pre-filling mechanisms, the user interaction burden and the threshold of prompt word engineering are significantly reduced, and the problems of passive interaction, low efficiency of multiple rounds, and faults in collaboration between models in the prior art are solved. The system can automatically identify information gaps and generate optimal forms, support multi-modal input and cross-model collaboration, and ensure the integrity and accuracy of information transmission. At the same time, through a feedback-driven self-optimization mechanism, the form generation strategy and template adaptability are continuously improved, achieving an efficient, accurate, and scalable intelligent interactive experience.

[0020] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures pointed out in the description, claims and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 This is a block diagram of the AI ​​dialogue system based on a two-way interactive form of the present invention; Figure 2 The present invention is a flow chart of the AI ​​dialogue method based on a two-way interactive form. DETAILED DESCRIPTION

[0022] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this technical field without creative work are within the scope of protection of the present invention.

[0023] Please refer to the figure, the AI ​​dialogue system and method based on a two-way interactive form in the present invention, and its specific embodiments are as follows: Dynamic form generation and interaction process in common scenarios 1. User input parsing and information gap detection The user enters a natural language request (for example: "I need a market analysis report").

[0024] The information integrity analysis unit extracts intent ("market analysis") and key entities (such as "industry field" and "time range") through semantic analysis, compares them with the predefined complete information model (market analysis must include fields such as "target market", "competitive product data" and "user portrait"), marks missing items and calculates priorities (such as "competitive product data" has a high priority).

[0025] 2. Form generation decision and multimodal interaction The form generation engine detects that there is no pre-existing industry standard template, selects "Thinking Model Mapping Mode", matches to "PESTEL Analysis Framework" (political, economic, social, technological, environmental, legal), and generates a form with the following fields: Required fields: target market (text input box), time range (date picker); Optional items: policy document upload (image mode), industry white paper link (URL input).

[0026] The multimodal interactive interface supports users to upload industry reports in PDF format (the image mode automatically parses key chart data) and provide additional explanations through voice input (such as "focus on the field of artificial intelligence").

[0027] 3. Intelligent pre-fill and cross-model collaboration The intelligent pre-fill module detects that the user account has enterprise-level permissions and automatically pre-fills "Target market: Chinese AI hardware market" from the internal database; The cross-model collaboration adapter encapsulates the form data in JSON format, adds validation rules (such as the time range must be in YYYY-MM-DD format), and sends it to the data analysis Agent cluster: The economic analysis agent calls the statistics bureau API to obtain macroeconomic data; Technology Trend Agent searches the patent database for the past three years.

[0028] 4. Data integration and response generation After each agent returns data, the system generates structured prompt words: “Analysis of China’s AI hardware market (202x-202x) based on the PESTEL framework. Policy documents are attached. Technology trends refer to patent number CN2025xxxxxxx.” The large model output includes an analysis report with data visualization, and marks the missing field "User Profile" for subsequent supplementation.

[0029] 5. Feedback optimization and template iteration The self-optimization system records the "User Profile" field manually added by the user, and automatically adds this field as a required field in subsequent similar requests; Based on the relevance score of the output report (user rating 4.8 / 5), the weight of the PESTEL template is increased, and the inefficient SWOT template is downgraded.

[0030] Specific implementation scenarios: Example 1: Intelligent pre-filling of taxi-hailing scenarios Implementation steps: 1. User input: "I want to take a taxi to Computer Plaza" 2. Scene recognition: Semantic analysis identifies the "taxi-hailing" intent and matches the travel service form template 3. Gap detection: Required fields are missing: departure location (current_location), car type preference (car_type) 4. Intelligent pre-fill: Call the mobile phone GPS to obtain the real-time location → Fill in "Departure point: Computer subway station exit 1" Read historical order data → Pre-select "Car model preference: Comfort type (80% selected in the past 12 months)" 5. Interactive confirmation: The user changes the destination to "Computer Plaza South Gate" and confirms 6. Service Execution: Submit structured data to a taxi API and get driver order information back within 3 seconds Example 2: Gift recommendation scenario Implementation steps: 1. User input: "I want to buy a gift for my friend, give me some suggestions" 2. Form generation: Matching gift recommendation thinking model, generating fields: markdown | Field | Type | |----------------|--------------- | | Relationship| Single choice (colleague / family / lover) | | Age| Numeric Input| | Gender| Single Choice| | Budget (RMB) | Slider (100-5000) | | Hobbies| Multiple-select tags (sports / beauty / digital...)| 3.User fill in: Select "Relationship: Colleague | Age: Computer | Gender: Female | Budget: Computer x | Interests: Coffee, Literature" 4. Data conversion: Generate structured prompt words: "Recommend a computer gift worth about x yuan, suitable for a female colleague who is 6 years old and likes coffee and literature" 5. Result output: The large model returns to the "Computer x Coffee Joint Hand Pour Set + Custom Bookmark Combination" with a purchase link Example 3: Industry Research Agent Collaboration Implementation steps: 1. User input: "Help me analyze whether new energy vehicle A is worth investing in" 2. Model matching: Load the SWOT analysis framework and generate a form: | Analysis Dimensions| Information Items to be Collected| |----------|------------------------------| | Advantages(S) | Number of core technology patents| | Disadvantages (W) | Supply Chain Risk Points | | Opportunities (O) | Policy support direction | | Threat(T) | Competitors' new product launch plans | 3.Agent collaboration: The investment analysis agent breaks down the form into four subtasks and schedules them separately: Patent search agent query "a car 2023 patent application" Public Opinion Monitoring Agent Scans "Power Battery Supplier Default Risk" Policy Analysis Agent extracts "New Policy on Subsidies for New Energy Vehicles" Competitive product monitoring agent obtains "B car launch plan" 4. Report Generation: The comprehensive data generates a visual SWOT matrix and gives an investment recommendation index: ★★★☆☆ Example 4: Medical emergency scenario Implementation steps: 1. User calls for help: "I feel sick in my heart and need emergency treatment!" 2. Emergency Response: Identify medical emergency scenarios and generate fields: markdown | Emergency Field| Data Source| |------------------|-------------------------| | Current location | GPS location | | Medical History| Electronic Health Records| | Real-time vital signs | Smart watch data | | History of drug allergy | Manual confirmation required | 3. Quick pre-fill: Automatically obtain: Location: Computer No. 1, Computer Road, Computer District, Computer City Medical history: Hypertension (diagnosed in 202x) Heart rate: 125 beats / minute (over-threshold alarm) 4. Key confirmation: Flashing warning "Drug allergy history is not filled in", the user quickly clicks "Confirm no allergy history" 5. Rescue linkage: The data is synchronized to the 120 dispatch center to generate an emergency plan: Distance of dispatched vehicle: 1.2 km (estimated arrival time: 3 minutes) Prepare medicines: nitroglycerin (predicted based on medical history) First aid tips: "Keep sitting and do not move the patient."

[0031] An embodiment of the invention discloses a large-scale language model dialogue system with bidirectional interactive form generation capability. It identifies information gaps in user input through an information integrity analysis unit, dynamically generates structured forms in combination with a form generation engine, supports multimodal input and intelligent pre-filling, and significantly reduces the user interaction burden. The cross-model collaboration adapter realizes efficient information transmission and data verification between AI Agents, and improves the execution efficiency of complex tasks. The feedback-driven self-optimization system continuously optimizes the form generation strategy and template adaptability by recording user correction behavior and large model output quality, ensuring the high accuracy and generalization ability of the system in multiple scenarios such as medical care, investment, and scientific research.

Claims

1. An AI dialogue system based on a two-way interactive form, characterized by: Includes the following: Information integrity analysis unit, used to parse user input, identify intent and entities, compare the complete information model in the domain knowledge base, mark missing information items and calculate priorities; The form generation engine is used to select the self-generation mode, service call mode or mental model mapping mode according to the type of information gap, generate a structured interactive form, and automatically assign a unique variable identifier to each field. The naming rule is "scenario-field type-hash value", and the variable identifier is mapped with the semantic description and stored in the global variable registry; Multimodal interactive interface, used to support multiple input methods such as text, voice, and image to fill in the form, including drop-down selection and fill-in-the-blank input in text mode, voice-to-text filling in voice mode, and feature extraction and key information analysis in image mode; Cross-model collaboration adapter, which is used to encapsulate form data into a machine-readable standardized format in the interaction scenario between AI agents, and attach data validation rules to ensure the integrity of information transmission; Intelligent pre-fill module, which is used to automatically fill in form fields based on user device type, current activity and data access permissions, and supports multiple pre-fill rules including but not limited to GPS positioning, enterprise CRM system data call, and wearable device data acquisition; A feedback-driven self-optimization system is used to record users' modification behaviors on pre-filled fields, build field correction heat maps, dynamically adjust pre-filling strategies, and evaluate template adaptability based on large model output quality to achieve template evolution and semantic gap detection.

2. The AI ​​dialogue system based on a two-way interactive form according to claim 1, characterized in that: The generation strategy of the form generation engine includes: Self-generation mode, which is used to extract necessary information dimensions based on semantic analysis when there is no predefined template in the problem domain, dynamically construct form fields, and automatically generate natural language sentences to embed variables in a logical order; Service call mode, used to connect to external systems and directly call industry-standard form templates, which contain variable placeholders; The thinking model mapping mode is used to match the problem type to the professional analysis framework, generate information collection items according to the framework requirements, and bind variable identifiers.

3. The AI ​​dialogue system based on a two-way interactive form according to claim 1 is characterized in that: The multimodal interaction interface includes: Text mode is used to provide structured controls such as drop-down selection and fill-in-the-blank input. The control value is dynamically bound to the variable identifier. Voice mode, which is used to convert voice descriptions into form field values ​​and automatically map them to corresponding variable identifiers; Image mode is used to automatically parse key information in uploaded images through feature extraction technology and bind the parsing results to variable identifiers.

4. The AI ​​dialogue system based on a two-way interactive form according to claim 1, characterized in that: The prefilling rules of the intelligent prefilling module include: If the form contains a "geolocation" field and the user authorizes location, the GPS coordinates are automatically filled in and converted to a standard address format and bound to the variable identifier; If it is detected that the user is using a corporate account, the customer number and project name are automatically pre-filled from the internal CRM system and bound to the variable identifier; In medical emergency scenarios, wearable devices are used to obtain heart rate and blood oxygen data, complete health monitoring forms, and bind them to variable identifiers; In the search scenario, public information or complex information in the database is obtained by searching the big model, and possible answers are pre-filled into the form.

5. The AI ​​dialogue system based on a two-way interactive form according to claim 1, characterized in that: The cross-model collaboration adapter includes: Define machine-readable metadata specifications, including field descriptions, validation rules, and urgency indicators; When AI agents interact with each other, the prefill assistant can initiate data requests to other agents, and automatically calculate the recommended values ​​after the responding agent returns the relevant data; Supports step-by-step filling and mid-course correction, and real-time verification of data validity.

6. The AI ​​dialogue system based on a two-way interactive form according to claim 1, characterized in that: The feedback driven self-optimizing system comprises: Record users’ modification behaviors on pre-filled fields, build field modification heat maps, and identify high-frequency modification fields; Dynamically adjust prefill strategy based on correction patterns; Evaluate template adaptability based on large model output quality, and automatically downgrade low-scoring templates; When a template triggers user questions continuously, mark the variable fields that need to be supplemented.

7. The AI ​​dialogue method based on a two-way interactive form is characterized by: The AI ​​dialogue system based on a two-way interactive form according to claims 1-6 implements the following method steps: Information gap detection phase: parse the input content, identify intent and entities, compare the complete information model in the domain knowledge base, mark missing items and calculate priorities; Form generation decision stage: First check whether there is an industry standard form that can be called. If there is no matching template, select the thinking model to generate fields according to the question type, or independently create a dynamic form containing required / optional fields; Interactive completion stage: Return the form to the user or question model, clearly mark the required items, provide input format suggestions, and automatically fill in the fields according to the pre-fill rules; Data integration and response phase: The form data is integrated with the original input to generate a complete context, triggering the AI ​​model to recalculate and output accurate responses, while recording missing patterns to optimize subsequent form generation strategies.

8. The AI ​​dialogue method based on a two-way interactive form according to claim 7, characterized in that: The form generation decision phase further includes: Assign a unique variable identifier to each form field, using the naming convention of "scenario-field type-hash value"; Call the prompt word template in the preset template library, which contains variable placeholders; For new scenarios, natural language sentences are automatically generated based on field semantics, and variables are embedded in a logical order.

9. The AI ​​dialogue method based on a two-way interactive form according to claim 7, characterized in that: The interactive completion stage further includes: Return the form to the user or question model, clearly mark the required fields, input format suggestions, and bind the input data to the corresponding variable identifiers; Intelligently adjust the form display method based on the user's device type and current activity; Verify user data access permissions and automatically populate fields within the authorized scope.

10. The AI ​​dialogue method based on a two-way interactive form according to claim 7, characterized in that: The data integration and response phase further includes: Dynamically match form data with external database fields based on variable identifiers; Calling the intelligent prompt word synthesis engine to generate domain-limited structured query instructions; Trigger the large language model to recalculate and output accurate responses, while recording missing patterns to optimize subsequent form generation strategies.

Images