A structure-driven method and device for dynamically generating prompt words for conversational form filling

Abstract

The present invention belongs to the field of intelligent interaction within the field of artificial intelligence, and provides a structure-driven method and device for dynamically generating prompt words for conversational form filling. Its technical solution is to obtain the business background and related document information of the document to be parsed; parse the document data structure, and generate initial prompt words based on the form filling rules of each business field and the document data structure; input the document information under different business backgrounds and the corresponding generated initial prompt words into a large model, perform secondary adjustment and optimization on the prompt words, and generate customized prompt words for each actual business requirement; respond to the user's document filling instructions, guide the document filling based on the customized prompt words based on the actual business requirements, combine the user's document filling feedback information, and call the large model again based on the feedback information to adjust and optimize the prompt words to obtain the adjusted form filling prompt words. This improves the efficiency of document filling and enhances the user's document filling experience.

Description

Technical Field

[0001] The present invention belongs to the field of intelligent interaction in the field of artificial intelligence, and in particular relates to a structure-driven method and device for dynamically generating prompt words for conversational form filling. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] Intelligent form-filling interactions are a common intelligent requirement within enterprises. Traditional form-filling methods based on rules or semantic models rely on fixed word slots, resulting in a mechanical interaction and poor user experience.

[0004] After the outbreak of big model technology, prompt words are mostly used to control big models and guide users to complete the filling of documents. The prompt words are strongly bound to the document process that needs to be filled out, that is, the required document filling prompt words are strongly correlated with the structure and logic of the document. Each document requires specific prompt words to guide. On the one hand, this method has to readjust the prompt words when the document changes due to the strong coupling between the prompt words and the configuration. There is a lack of a data structure-driven mechanism, and the prompt words cannot be automatically adapted according to the changes in the document data structure. It can only be passively modified manually, which reduces work efficiency. On the other hand, relying on static templates to generate prompt words is completely unable to adapt to secondary development needs. When the business scenario changes or the function needs to be expanded, the prompt words are difficult to flexibly adjust to meet the new requirements and cannot be applied to changing business scenarios. Summary of the Invention

[0005] In order to solve at least one technical problem existing in the above-mentioned background technology, the present invention provides a structure-driven conversational form filling prompt word dynamic generation method and system, which avoids the problem of mismatch between different business scenarios caused by the strong correlation between existing document filling prompt words and document structure and logic. At the same time, there is no need to repeatedly adjust the prompt words according to the business and type of the document, thereby improving the efficiency of document filling and enhancing the user's document filling experience.

[0006] According to one aspect of an embodiment of the present application, a structure-driven method for dynamically generating prompt words for a conversational form filling is provided, comprising the following steps:

[0007] Obtain the business background and related document information of the document to be parsed;

[0008] Parse the data structure of the document, combine the form filling rules of each business field and the data structure of the document, and generate the initial prompt word;

[0009] Input document information and the corresponding initial prompt words generated under different business backgrounds into the big model, conduct secondary adjustment and optimization on the prompt words, and generate customized prompt words based on actual business needs;

[0010] In response to the document filling instructions, customized prompt words based on actual business needs are used to guide document filling. Combined with the feedback information of document filling, the big model is called again according to the feedback information to adjust and optimize the prompt words to obtain the adjusted form filling prompt words.

[0011] Furthermore, the data structure of the document is parsed, including: analyzing the layout of the document, segmenting each area and identifying text blocks based on the layout analysis results, obtaining the title area, main body area, table body area and footer based on area segmentation, performing text extraction based on text block recognition to obtain text extraction results, and combining the area segmentation and text extraction results to obtain the final document data structure.

[0012] Furthermore, the data structure of the document obtained by parsing includes field types, field names, associations between fields, and possible options.

[0013] Furthermore, the form filling rules for each business field for generating the initial prompt words include:

[0014] Determine the classification of fields based on business attributes, and maintain and add them to the corresponding field descriptions based on the field filling requirements;

[0015] Get the optional enumeration value of the enumeration field and add it to the fixed prompt word database;

[0016] Determine the predicted fields and add instructions for the data extraction tools for the predicted fields;

[0017] Clarify the summary fields and field summary rules.

[0018] Furthermore, the secondary adjustment and optimization of the prompt words specifically includes: determining the process rules and interaction constraints of each actual business, and adding cases of fixed prompt words and final prompt words in the interaction constraints; the large model learns according to the process rules and interaction constraints of each actual business, and outputs the prompt word optimization results.

[0019] Furthermore, combined with the feedback information of document filling, when the big model is called again to adjust and optimize the prompt words based on the feedback information, the frequency of user-filled errors and the user's rating of the prompt words are used to further adjust the process rules of each actual business.

[0020] Furthermore, in response to the document filling instruction, the input content is analyzed and verified in real time, and when the input content does not conform to the prompt word requirements or the logic of the document, an error prompt and a guidance prompt are generated.

[0021] Furthermore, when generating error prompts and guidance prompts, the type of error is judged. If the error is of the set error type, the error is automatically corrected. If the dislocation is of a non-set dislocation type, the error position is marked, and the specific document rules are displayed based on the error position, and correction examples are given.

[0022] According to one aspect of an embodiment of the present application, a structure-driven dynamic generation device for prompt words for conversational form filling is provided, the device comprising:

[0023] The document information acquisition module is used to obtain the business background and related document information of the document to be parsed;

[0024] The initial prompt word generation module is used to parse the data structure of the document and generate the initial prompt word based on the form filling rules of each business field and the data structure of the document;

[0025] The prompt word dynamic optimization module is used to input document information and the corresponding initial prompt words generated under different business backgrounds into the large model, perform secondary adjustment and optimization on the prompt words, and generate customized prompt words based on actual business needs;

[0026] In response to the document filling instructions, customized prompt words based on actual business needs are used to guide document filling. Combined with the feedback information of document filling, the big model is called again according to the feedback information to adjust and optimize the prompt words to obtain the adjusted form filling prompt words.

[0027] According to one aspect of an embodiment of the present application, a computer device is provided, comprising a processor and a memory, wherein the memory stores a computer program, and the computer program is loaded and executed by the processor to implement the above-mentioned structure-driven method for dynamically generating conversational form filling prompt words.

[0028] According to one aspect of an embodiment of the present application, a computer-readable storage medium is provided, in which a computer program is stored. The computer program is loaded and executed by a processor to implement the above-mentioned structure-driven method for dynamically generating conversational form filling prompt words.

[0029] According to one aspect of an embodiment of the present application, a computer program product is provided, which includes a computer program, and the computer program is loaded and executed by a processor to implement the above-mentioned structure-driven method for dynamically generating conversational form filling prompt words.

[0030] Compared with the prior art, the present invention has the following beneficial effects:

[0031] The present invention generates a basic prompt word framework based on the data structure of the document at the rule layer, and conducts secondary adjustment and optimization of the prompt words in combination with the prompt word framework of the rule layer and the actual business scenarios at the large model layer, thereby avoiding the problem of mismatch between different business scenarios caused by the strong correlation between the existing document filling prompt words and the document structure and logic. At the same time, there is no need to repeatedly adjust the prompt words according to the different business and types of the document, which improves the efficiency of document filling and enhances the user experience of document filling.

[0032] Advantages of additional aspects of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] The accompanying drawings, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0034] Figure 1 This is a schematic diagram of an implementation environment for a solution provided by an embodiment of the present application;

[0035] Figure 2 This is a flow chart of a structure-driven method for dynamically generating prompt words for conversational form filling provided by an embodiment of the present application;

[0036] Figure 3 This is a block diagram of a structure-driven, conversational form filling prompt word dynamic generation device provided by one embodiment of the present application;

[0037] Figure 4 This is a structural block diagram of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0038] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0039] It should be noted that the following detailed descriptions are illustrative and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which the present invention belongs.

[0040] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, they indicate the presence of features, steps, operations, devices, components and / or combinations thereof.

[0041] A Large Language Model (LLM) is an AI algorithm based on deep learning technology. Its goal is to enable computers to understand and generate natural language. It analyzes large amounts of language data, such as text, speech, or images, to learn the structure and patterns of language and use this knowledge to perform various natural language processing tasks, such as machine translation, speech recognition, text classification, and question-answering systems. Large language models typically use the Transformer architecture from deep learning to model text sequences to understand context and semantics. Their training typically involves extensive data and computing resources, such as large-scale corpora and high-performance computing platforms. During training, large language models gradually learn the characteristics and patterns of language, developing the ability to understand and express it.

[0042] Intelligent document filling refers to the scenario of guiding the filling of business documents through voice or card-based methods, such as filling out documents such as conference room applications and reimbursement applications. Most current solutions use prompt words to guide the interaction between the large model and the user to complete the filling of documents. However, the prompt words currently required for document filling are strongly correlated with the structure and logic of the document. Each document requires specific prompt words for guidance. On the one hand, this method has a strong coupling between the prompt words and the configuration, which means that when the document changes, the prompt words must be readjusted. There is a lack of a data structure-driven mechanism, and the prompt words cannot be automatically adapted according to changes in the document data structure. They can only be passively modified manually, which reduces work efficiency. On the other hand, relying on static templates to generate prompt words is completely unable to adapt to secondary development needs. When the business scenario changes or the function needs to be expanded, the prompt words are difficult to flexibly adjust to meet the new requirements and cannot be applied to changing business scenarios.

[0043] Please refer to Figure 1 , which shows a schematic diagram of a solution implementation environment provided by an embodiment of the present application. The solution implementation environment can implement a structure-driven dynamic generation system for prompt words for conversational form filling. The solution implementation environment may include: a terminal device 10 and a server 20.

[0044] The number of terminal devices 10 may be one or more. The terminal device 10 may be, but is not limited to, an electronic device such as a smartphone, tablet computer, laptop computer, desktop computer, smart speaker, smart watch, etc. The terminal device 10 may have a client for a target application installed. Optionally, the application may be an application that needs to be downloaded and installed, or an application that can be clicked and used, which is not limited in this embodiment of the present application.

[0045] The server 20 is used to provide background services for the client of the target application installed and running on the terminal device 10. For example, the server 20 can be the background server of the aforementioned application. The server 20 can be a standalone physical server, a server cluster consisting of multiple servers, or a cloud computing service center. Optionally, the server 20 provides background services for applications running on multiple terminal devices 10 simultaneously. The terminal device 10 and the server 20 can communicate with each other via a network.

[0046] In the embodiment of the present application, the execution subject of each step may be a computer device, which refers to an electronic device with data calculation, processing and storage functions. For example, the computer device may be Figure 1 The terminal device 10 in the embodiment may also be a server 20.

[0047] Please refer to Figure 2 , which shows a flowchart of a structure-driven method for dynamically generating prompts for conversational form filling, provided in one embodiment of the present application. Each step of the method may be performed by a computer device. The method may include at least one of the following steps 210-250.

[0048] Step 210: Obtain the documents to be parsed and document information for each business scenario;

[0049] Different documents correspond to different business contexts. For example, financial reimbursement documents include invoices and travel documents; supply chain management documents include purchase orders and waybills; contract signing documents include agreements and POs; and customer service documents include application forms and complaint forms.

[0050] Taking the filling out of a travel application form as an example, users are required to enter multiple fields such as the departure city, departure date, return date, arrival city, trip type, project, and companions. There are also control fields such as whether it is required, whether it is multiple-choice, entry method, default value, and description. In a business scenario, filling out a simple document requires writing thousands of words of prompts.

[0051] Step 220, parsing the data structure of the document;

[0052] When parsing the document's data structure, you can first analyze the document's layout. Based on the layout analysis results, you can segment each area and identify text blocks. Based on the area segmentation, you can obtain the title area, main body area, table body area, and footer area. Based on the text block recognition, you can perform text extraction to obtain the text extraction results. Combine the area segmentation and text extraction results to obtain the final document data structure.

[0053] In this embodiment, when analyzing the format of the document, existing format segmentation technologies such as contour detection method and projection analysis method are used; unstructured or semi-structured document data is converted into computer-processable structured data such as JSON, XML, and database table. The specific parsing method can adopt the existing parsing method.

[0054] It should be noted that when segmenting each area, the existing segmentation algorithm can be used; when recognizing text blocks, the existing image processing, OCR and NLP technologies can be used, which is not the focus of this application.

[0055] In this embodiment, the document data structure obtained by parsing includes field types, field names, associations between fields, and possible options;

[0056] Specifically, field types include text (String), number (Number), date (Date), enumeration (Enum), Boolean (Boolean), and object (Object), etc.

[0057] For example, text (String) includes name, address, and description fields;

[0058] Number: includes calculated fields such as amount, quantity, and tax rate;

[0059] Date (Date) includes document date, validity period, etc.;

[0060] Boolean includes whether the class tag field;

[0061] Objects include nested structures such as product detail rows;

[0062] For a travel application, its data structure may include multiple fields such as departure city (enumeration), departure date (manual entry), return date (manual entry), arrival city (manual entry), trip type (enumeration), project (AI prediction), companions (manual entry), reason for trip (AI summary), etc.

[0063] Step 230: Generate an initial prompt word based on the form filling rules of each business field and the data structure of the parsed document;

[0064] The specific steps include:

[0065] Step 231: Determine the classification of the field based on the business attributes, and maintain and add it to the corresponding field description according to the field filling requirements;

[0066] In this embodiment, the classification of fields includes required fields and non-required fields; required fields and non-required fields, and the corresponding field filling requirements are maintained in the field description to ensure that the prompt words clearly indicate the required fields for filling in the form.

[0067] Step 232: Obtain the optional enumeration value of the enumeration field and add it to the fixed prompt word database;

[0068] For example, if the business trip type is an enumeration field, the options that can be filled in are maintained as fixed prompt words.

[0069] Step 233, determine the predicted field and add the description of the data acquisition tool of the predicted field;

[0070] Step 234: clarify the summary fields and the summary rules of the fields.

[0071] A summary field is the result of applying a specific calculation rule (such as an aggregation function) to one or more base data fields. A field's summary rule specifies the aggregation function used to calculate the field, generating the final summary directly without relying on conversation feedback.

[0072] Step 235 , generating an initial prompt word by combining the form filling rules of steps 231 to 234 .

[0073] Taking the business meaning and filling method of the travel application form as an example, the initial prompt words generated are as follows:

[0074] Required fields (all must be collected): departure city, departure date, return date, arrival city, business trip type, whether multiple projects, reason for business trip; non-mandatory fields (users can choose to provide, no need to actively ask): companions (multiple choices), transit city; predicted field: project.

[0075] Enumeration fields:

[0076] {"data": {"type":"AiOptions","optionsname":"Business trip type","options":["Official business trip","Training business trip","Closed R&D"]}};

[0077] Summary Fields:

[0078] All summary fields directly generate the target summary results, without the need to rely on conversation feedback to directly generate the final business trip reason.

[0079] Prediction fields:

[0080] Project fields, you need to call the project field acquisition tool.

[0081] In this way, a basic prompt word list can be automatically generated based on the rules according to the data structure of different documents.

[0082] To improve the accuracy of prompt word generation, the present invention can further refine and optimize the rules. For example, more targeted rules can be developed based on the characteristics of different industries and business scenarios. In financial documents, for fields involving amounts, stricter formatting requirements and risk warning rules can be added to generate professional prompt words.

[0083] Step 240: Input the document information under different business contexts and the corresponding fixed prompt words generated into the large model, perform secondary adjustment and optimization on the prompt words, and generate customized prompt words based on actual business needs;

[0084] In this embodiment, the secondary adjustment and optimization of prompt words specifically includes: determining the process rules and interaction constraints of each actual business, and adding cases of fixed prompt words and final prompt words to the interaction constraints; the large model learns based on the process rules and interaction constraints of each actual business, and outputs the prompt word optimization results;

[0085] It should be noted that the process rules and interaction constraints of each actual business can be set according to user needs; such as the filling rules of some optional fields, whether multiple fields need to be filled in during the interaction process, etc.

[0086] The process of secondary adjustment and optimization of prompt words is still carried out based on the travel application form:

[0087] 1. Field collection specifications:

[0088] Today is {{current_date}}

[0089] Required fields (all must be collected): departure city, departure date, return date, arrival city, business trip type, whether multiple projects, business trip reason. Optional fields (users can choose to provide, no need to actively ask): companions (multiple choices), transit city prediction field: project;

[0090] 2. Interaction requirements:

[0091] Use natural conversational methods to guide users to provide information. When encountering enumeration fields, the options must be presented strictly in the specified format.

[0092] The fields obtained through prediction must be queried last, and the prediction results must be displayed strictly in the specified format;

[0093] If a lot of information is provided at one time, there is no need to confirm each one and go straight to the next piece of information;

[0094] 3. Enumeration field processing specifications:

[0095] When the user is required to select the following enumeration fields, the following format must be returned and the format cannot be modified:

[0096] {"data": {"type":"AiOptions","optionsname":"Business trip type","options":["Official business trip","Training business trip","Closed R&D"]}};

[0097] 4. Summary of field processing specifications:

[0098] All summary fields directly help users generate summaries without requiring user feedback;

[0099] Based on the conversation, generate a business trip reason, for example: Apply for a business trip to xx place for xxx project;

[0100] 5. Prediction acquisition field processing specifications:

[0101] When users are required to select a prediction field, fixed-format content must be returned, and default and recommended values ​​must be obtained by calling the tool.

[0102] To improve the generation effect of the large model, domain adaptation and fine-tuning are performed on the large model. The large model is further trained using a large amount of conversation data and document filling cases related to the target business domain, enabling it to better understand the professional terminology and business processes in the field, thereby generating customized prompts that better meet business needs.

[0103] Step 250, in response to the document filling instruction, guide the document filling based on the customized prompt words under the actual business needs, combine the feedback information of the document filling, call the big model again according to the feedback information to adjust and optimize the prompt words, and obtain the adjusted form filling prompt words.

[0104] In this embodiment, feedback information is collected during the use of the interactive form filling system, such as the frequency of user errors and user evaluation of the prompt words. Based on this feedback information, the rules and parameters of the large model are further optimized to continuously improve the quality of the prompt words and system performance.

[0105] Furthermore, a semantic understanding and verification mechanism for input content has been added. When filling out documents, natural language processing technology is used to analyze and verify the input content in real time. If the input is found to be inconsistent with the prompt word requirements or the logic of the document, timely error prompts and guidance are given to help users fill in the document information more accurately.

[0106] Specifically, the NLP analysis engine can be used to perform real-time analysis and verification of user input, and pre-trained models such as BERT / RoBERTa can be fine-tuned to identify the purpose of user operations, such as "apply for reimbursement" and "check progress."

[0107] Based on the form filling rules of each business field and the adjusted form filling prompts and user operations, error prompts are issued when errors such as format errors (date format should be YYYY-MM-DD), logical conflicts such as the end time being earlier than the start time, business violations exceeding the single approval authority, missing information and failure to fill in the emergency contact are encountered;

[0108] When guiding the user, the type of error is determined. If the error is of the set error type, it will be corrected automatically. If the misalignment is of a non-set misalignment type, the error location will be marked, and the specific document rules based on the error location will be displayed, and correction examples will be given.

[0109] Please refer to Figure 3 , which shows a block diagram of a structure-driven dynamic generation device for dialog-style form filling prompts provided by an embodiment of the present application. The device has the function of implementing the above-mentioned content generation method, and the function can be implemented by hardware or by hardware executing corresponding software. The device can be the computer device described above, or it can be set in a computer device. Figure 3 As shown, the device 300 may include: a document information acquisition module 310 , an initial prompt word generation module 320 and a prompt word dynamic optimization module 330 .

[0110] The document information acquisition module 310 is used to obtain the business background and related document information of the document to be parsed;

[0111] The initial prompt word generation module 320 is used to parse the data structure of the document and generate the initial prompt word based on the form filling rules of each business field and the data structure of the document;

[0112] The prompt word dynamic optimization module 330 is used to input document information under different business backgrounds and the corresponding initial prompt words generated into the large model, perform secondary adjustment and optimization on the prompt words, and generate customized prompt words based on actual business needs;

[0113] In response to the user's document filling instructions, customized prompt words based on actual business needs are used to guide document filling. Combined with the feedback information of document filling, the big model is called again according to the feedback information to adjust and optimize the prompt words to obtain the adjusted form filling prompt words.

[0114] It should be noted that the apparatus provided in the above embodiments, when implementing its functions, is only illustrated by the division of the above functional modules. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the content structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the apparatus and method embodiments provided in the above embodiments are based on the same concept. The specific implementation process is detailed in the method embodiment and will not be repeated here.

[0115] Please refer to Figure 4 , which shows a block diagram of a computer device 400 provided in one embodiment of the present application. The computer device 400 can be any electronic device with data computing, processing, and storage functions. The computer device 400 can be used to implement the content generation method provided in the above embodiment.

[0116] Typically, the computer device 400 includes a processor 401 and a memory 402 .

[0117] Processor 401 may include one or more processing cores, such as a quad-core processor or an octa-core processor. Processor 401 may be implemented in hardware using at least one of a DSP (Digital Signal Processing), an FPGA (Field Programmable Gate Array), and a PLA (Programmable Logic Array). Processor 401 may also include a main processor and a coprocessor. The main processor is used to process data in the awake state, also known as a CPU (Central Processing Unit); the coprocessor is a low-power processor used to process data in the standby state. In some embodiments, processor 401 may be integrated with a GPU (Graphics Processing Unit), which is responsible for rendering and drawing content displayed on the display screen. In some embodiments, processor 401 may also include an AI processor for handling computational operations related to machine learning.

[0118] Memory 402 may include one or more computer-readable storage media, which may be non-transitory. Memory 402 may also include high-speed random access memory and non-volatile memory, such as one or more magnetic disk storage devices or flash memory storage devices. In some embodiments, the non-transitory computer-readable storage media in memory 402 is used to store a computer program, which is configured to be executed by one or more processors to implement the above-described content generation method.

[0119] Those skilled in the art will understand that Figure 4 The structure shown in the figure does not constitute a limitation on the computer device 400, and the computer device 400 may include more or fewer components than shown in the figure, or combine some components, or adopt a different component arrangement.

[0120] In an illustrative embodiment, a computer-readable storage medium is also provided, storing a computer program that, when executed by a processor of a computer device, implements the aforementioned content generation method. Optionally, the computer-readable storage medium may be a ROM (Read-Only Memory), a RAM (Random Access Memory), a CD-ROM (Compact Disc Read-Only Memory), a magnetic tape, a floppy disk, or an optical data storage device.

[0121] In an exemplary embodiment, a computer program product is further provided, the computer program product including a computer program stored in a computer-readable storage medium. A processor of a computer device reads the computer program from the computer-readable storage medium and executes the computer program, causing the computer device to perform the above-mentioned content generation method.

[0122] It should be noted that this application may display a prompt interface, pop-up window, or output a voice prompt message before and during the process of collecting the user's relevant data. The prompt interface, pop-up window, or voice prompt message is used to remind the user that their relevant data is currently being collected, so that this application only starts to execute the relevant steps of obtaining the user's relevant data after obtaining the user's confirmation operation on the prompt interface or pop-up window. Otherwise (that is, when the user's confirmation operation on the prompt interface or pop-up window is not obtained), the relevant steps of obtaining the user's relevant data are terminated, that is, the user's relevant data is not obtained. In other words, all user data collected by this application are processed strictly in accordance with the requirements of relevant national laws and regulations. The informed consent or separate consent of the personal information subject is obtained only when the user agrees and authorizes the collection, and the subsequent data use and processing are carried out within the scope of the laws and regulations and the authorization of the personal information subject. The collection, use, and processing of relevant user data must comply with the relevant laws, regulations, and standards of the relevant countries and regions.

[0123] It should be understood that the "multiple" mentioned in this article refers to two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent three situations: A exists alone, A and B exist at the same time, and B exists alone. The character " / " generally indicates that the previous and subsequent associated objects are in an "or" relationship. In addition, the step numbers described in this article only illustrate a possible execution sequence between the steps. In some other embodiments, the above steps may not be executed in the order of the numbers, such as two steps with different numbers are executed at the same time, or two steps with different numbers are executed in the opposite order to the diagram. The embodiments of the present application do not limit this.

[0124] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be within the scope of protection of the present invention.

Claims

1. A structure-driven method for dynamically generating prompt words for conversational form filling, characterized in that: The steps include: Obtain the business background and related document information of the document to be parsed; Parse the data structure of the document, combine the form filling rules of each business field and the data structure of the document, and generate the initial prompt word; Input document information and the corresponding initial prompt words generated under different business backgrounds into the big model, conduct secondary adjustment and optimization on the prompt words, and generate customized prompt words based on actual business needs; In response to document filling instructions, customized prompts based on actual business needs are used to guide document filling. Combined with the feedback information from document filling, the large model is called again based on the feedback information to adjust and optimize the prompts to obtain the adjusted form filling prompts; The secondary adjustment and optimization of the prompt words specifically includes: determining the process rules and interaction constraints of each actual business, and adding cases of fixed prompt words and final prompt words to the interaction constraints; the large model learns based on the process rules and interaction constraints of each actual business, and outputs the prompt word optimization results; Among them, combined with the feedback information of document filling, when the big model is called again to adjust and optimize the prompt words based on the feedback information, the process rules of each actual business are further adjusted according to the frequency of errors filled in by users and the user's rating of the prompt words.

2. A structure-driven method for dynamically generating prompt words for a conversational form filling as claimed in claim 1, characterized in that: Parse the document's data structure, including analyzing the document's layout, segmenting each area and identifying text blocks based on the layout analysis results, obtaining the title area, main body area, table body area, and footer based on area segmentation, performing text extraction based on text block recognition to obtain text extraction results, and combining the area segmentation and text extraction results to obtain the final document's data structure.

3. A structure-driven method for dynamically generating prompt words for a conversational form filling as claimed in claim 1, characterized in that: The data structure of the document obtained by parsing includes field types, field names and the relationship between fields.

4. A structure-driven method for dynamically generating prompt words for a conversational form filling as claimed in claim 1, characterized in that: The form filling rules for various business areas that generate initial prompt words include: Determine the classification of fields based on business attributes, and maintain and add them to the corresponding field descriptions based on the field filling requirements; Get the optional enumeration value of the enumeration field and add it to the fixed prompt word database; Determine the predicted fields and add instructions for the data extraction tools for the predicted fields; Clarify the summary fields and field summary rules.

5. A structure-driven method for dynamically generating prompt words for a conversational form filling as claimed in claim 1, characterized in that: When responding to document filling instructions, the input content is analyzed and verified in real time. When the input content does not conform to the prompt word requirements or the logic of the document, error prompts and guidance prompts are generated.

6. A structure-driven method for dynamically generating prompt words for a conversational form filling as claimed in claim 5, characterized in that: When generating error prompts and guidance prompts, the type of error is determined. If the error is of the set error type, it is automatically corrected. If the dislocation is of a non-set dislocation type, the error location is marked, and the specific document rules are displayed based on the error location, and correction examples are given.

7. A structure-driven dynamic generation device for prompt words for conversational form filling, characterized in that: A structure-driven method for dynamically generating prompt words for conversational form filling according to any one of claims 1 to 6, comprising: The document information acquisition module is used to obtain the business background and related document information of the document to be parsed; The initial prompt word generation module is used to parse the data structure of the document and generate the initial prompt word based on the form filling rules of each business field and the data structure of the document; The prompt word dynamic optimization module is used to input document information and the corresponding initial prompt words generated under different business backgrounds into the large model, perform secondary adjustment and optimization on the prompt words, and generate customized prompt words based on actual business needs; In response to the document filling instructions, customized prompt words based on actual business needs are used to guide document filling. Combined with the feedback information of document filling, the big model is called again according to the feedback information to adjust and optimize the prompt words to obtain the adjusted form filling prompt words.

8. A program product, wherein the program product is a computer program product, comprising a computer program, characterized in that: When the computer program is executed by a processor, the steps of the structure-driven method for dynamically generating prompt words for conversational form filling are implemented as described in any one of claims 1-6.

Images