Intelligent form distribution and data analysis system based on dialogue

Abstract

The invention relates to the technical field of data analysis, and discloses a dialogue-based intelligent form distribution and data analysis system, which comprises a dialogue receiving unit, a form generation unit, a dialogue state tracking unit, a data distribution unit, a data analysis unit, a feedback optimization unit, an authority management unit and a multi-mode interaction unit. The dialogue receiving unit extracts user intentions and key entities, the form generation unit dynamically generates form templates according to the user intentions and key entities, the data distribution unit directionally distributes data, the data analysis unit mines data values, the feedback optimization unit optimizes the system according to analysis results, the authority management unit guarantees data safety, and the multi-mode interaction unit expands interaction modes. According to the method, intelligent form processing and data analysis are realized, the user experience and the working efficiency are improved, the potential value of the data is mined, and the safety and reliability of the data are ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of data analysis, and specifically to an intelligent form distribution and data analysis system based on dialogue. Background Art

[0002] In today's digital age, the use of forms is ubiquitous in various business processes. Whether it is the daily office work and business applications of enterprises, the government affairs handling of government departments, or the business operations of financial institutions, all these scenarios rely on forms to collect and process information. However, there are many problems with traditional form processing methods, making it difficult to meet the growing intelligent needs.

[0003] In the form generation stage, existing forms are usually pre-designed fixed templates, lacking flexibility and intelligence. Users need to manually fill in a large amount of information according to a fixed format. Even if some information is not relevant to the current business requirements, they cannot skip it. This not only increases the user's filling burden but also easily leads to filling errors and reduces work efficiency. For example, in the employee leave application process of an enterprise, the leave application form may contain general fields for various leave types, but when an employee applies for leave each time, only one type is involved, yet they still need to fill in a large amount of unnecessary information. Moreover, when business requirements change, the cost of modifying the fixed form template is relatively high, requiring a large amount of manpower and time, and unable to adapt to the dynamic adjustment of the business in a timely manner.

[0004] In terms of form distribution, the traditional method mainly relies on manual operation or simple process configuration to determine the flow of forms. This means that in large organizations or complex business processes, form distribution is prone to errors and cannot quickly and accurately transfer form data to the target processing nodes. Taking cross-departmental project approval as an example, a project application form may need to be circulated and approved among multiple departments. However, due to the uncertainty of manual operations and the limitations of process configuration, the form may be misdelivered to the wrong department or there may be delays during the circulation process, seriously affecting the project's progress efficiency.

[0005] At the data analysis level, existing form data processing methods can often only perform simple statistical analysis and cannot deeply explore the potential value behind the data. After a large amount of form data is collected, only basic operations such as summation and counting are carried out, ignoring the correlation between data and deep-level rules. For example, in the order form data of an e-commerce platform, only surface data such as the number of orders and sales amount are concerned, while the internal connections between different product categories and customer purchase behaviors, geographical factors are not explored, and it is unable to provide strong support for the enterprise's precision marketing and product optimization.

[0006] In addition, with the increasing awareness of data security and privacy protection, the permission management of form data is of crucial importance. However, traditional systems are relatively crude in permission management and difficult to achieve fine-grained permission control. The access permission settings for form data by different user roles are not flexible enough, which easily leads to the risk of sensitive data leakage. For example, in a medical system, if the permission management is improper, ordinary medical staff may obtain highly sensitive medical information of patients, which poses a serious threat to patient privacy.

[0007] With the development of artificial intelligence technology, technologies such as natural language processing and machine learning have provided new ideas and methods for solving the above problems. However, the current research and practice of deeply integrating these technologies into form distribution and data analysis systems are still in the development stage, and no mature and perfect solution has been formed. Therefore, it is of great practical significance and market demand to develop an intelligent form distribution and data analysis system based on dialogue to solve the many problems existing in traditional form processing methods. Summary of the Invention

[0008] The purpose of the present invention is to provide an intelligent form distribution and data analysis system based on dialogue to solve the problems raised in the above background technology.

[0009] To achieve the above purpose, the present invention provides the following technical solution: An intelligent form distribution and data analysis system based on dialogue, the system includes:

[0010] A dialogue receiving unit, configured to receive the dialogue content input by the user and extract the user intention and key entities in the dialogue;

[0011] A form generating unit, configured to dynamically generate an adapted form template according to the user intention and key entities, and the form template includes field types, data verification rules, and distribution path configurations;

[0012] A dialogue state tracking unit, configured to record the multi-round dialogue history, update the dialogue context state, and adjust the field priorities of the form template according to the context state;

[0013] A data distribution unit, configured to direct and send the form data filled in by the user to the target processing node according to the distribution path configuration of the form template;

[0014] A data analysis unit, configured to perform real-time clustering analysis and association rule mining on the form data to generate a structured analysis result;

[0015] A feedback optimization unit, configured to optimize the generation logic of form fields and the distribution path configuration according to the analysis result;

[0016] Among them, the execution steps of the form generating unit include:

[0017] Match the field combinations in the preset form rule library based on the user's intention, fill in the default values of the fields according to the key entities, and dynamically add additional fields related to the context.

[0018] Preferably, the execution steps of the dialogue receiving unit include:

[0019] Perform word segmentation and semantic parsing on the dialogue content input by the user through a natural language processing model, and extract the intention classification label and entity slots;

[0020] According to the preset intention-form mapping table, map the intention classification label to the corresponding form type identifier;

[0021] Bind the key entities in the entity slots to the form fields to generate an initial form filling suggestion.

[0022] Preferably, the execution steps of the form generation unit further include:

[0023] Calculate the field filling urgency score according to the list of incomplete fields in the dialogue context state;

[0024] Dynamically sort the form fields based on the urgency score and generate a form interface with priority markings;

[0025] If the key entity is missing, trigger an additional follow-up field and initiate a follow-up request through the dialogue receiving unit.

[0026] Preferably, the execution steps of the dialogue state tracking unit include:

[0027] Construct a dialogue state graph to record the intention changes, entity additions, and field completion status of each dialogue turn;

[0028] Real-time update the dialogue state graph through a graph neural network model to predict the potential field requirements for the next round of dialogue;

[0029] Adjust the field combination strategy in the form generation unit according to the prediction result.

[0030] Preferably, the execution steps of the data distribution unit include:

[0031] Parse the distribution path configuration in the form template, and the configuration includes the target node identifier, data format conversion rules, and transmission protocol;

[0032] Query the routing table according to the target node identifier to determine the data transmission channel;

[0033] Package the form data according to the specified protocol through an asynchronous message queue and send it to the target node.

[0034] Preferably, the execution steps of the data analysis unit include:

[0035] Perform outlier detection on the numerical fields in the form data, and use a density-based clustering algorithm to divide the data distribution interval;

[0036] Perform topic modeling on the text fields, and extract high-frequency keywords and co-occurrence relationships;

[0037] Fuse the clustering results with the topic modeling results to generate a multi-dimensional association rule graph.

[0038] Preferably, the execution steps of the feedback optimization unit include:

[0039] According to the frequently occurring field combinations in the association rule graph, update the field priority weights in the form rule library;

[0040] Adjust the threshold range in the data verification rule based on the outlier detection result;

[0041] If the error rate of the form data on the same distribution path exceeds the preset threshold, trigger the re-optimization of the distribution path configuration.

[0042] Preferably, it further includes:

[0043] A permission management unit for dynamically controlling the accessibility of form fields and the data distribution range according to the user role identifier;

[0044] The execution steps of the permission management unit include:

[0045] Encrypt the form fields through an attribute-based encryption algorithm and bind the decryption key of the user role;

[0046] During data distribution, perform desensitization or filtering on sensitive fields according to the permission level of the target node.

[0047] Preferably, the execution steps of the permission management unit further include:

[0048] Embed a permission verification module in the dialogue receiving unit to real-time verify the operation permission of the user role for the current form field;

[0049] If an unauthorized access request is detected, trigger the dialogue state tracking unit to insert a permission application follow-up field and pause the form generation process.

[0050] Preferably, the system further includes:

[0051] A multi-modal interaction unit for supporting the form filling with mixed input of voice, image and text;

[0052] The execution steps of the multi-modal interaction unit include:

[0053] Convert the voice input into text through a voice recognition model, and extract the structured data in the image through an image recognition model;

[0054] Match the data of multi-modal input with the form fields. If the match fails, trigger the manual review process.

[0055] Compared with the prior art, the beneficial effects of the present invention are:

[0056] The present invention extracts the user intention and key entities through the dialogue receiving unit, and the form generation unit dynamically generates an adapted form template accordingly. This process can accurately match the user's needs and avoid the user from filling in unnecessary information. For example, when handling business, the system automatically generates a form containing key fields such as loan amount and repayment term according to the intention in the user's dialogue, such as "applying for a personal loan", and fills in the default values of the fields using the key entities. For example, the estimated loan amount field is initially filled according to the income information provided by the user. At the same time, context-related additional fields are dynamically added according to the dialogue context. For example, when the user mentions having a property, the property information field is automatically added, greatly simplifying the form filling process and improving the user experience and work efficiency.

[0057] According to the distribution path configuration of the form template, the data distribution unit can accurately send the form data filled by the user to the target processing node. In complex business processes, such as the procurement approval process of an enterprise, when an application is initiated from the procurement department, the form data will automatically and accurately flow to the finance department, management, etc. for approval according to the preset distribution path, reducing errors and delays caused by manual intervention and ensuring the efficient and smooth operation of the business process.

[0058] The data analysis unit performs real-time clustering analysis and association rule mining on the form data. Taking the e-commerce order form data as an example, by detecting outliers for numerical fields (such as order amount) and dividing the data distribution interval using density-based clustering algorithms, and performing topic modeling on text fields (such as product reviews) to extract high-frequency keywords and co-occurrence relationships, multi-dimensional association rules such as different products and customer purchase preferences, regional consumption differences, etc. can be mined. These analysis results provide strong data support for the decisions of the enterprise's precision marketing, product optimization, inventory management, etc., and help the enterprise enhance its competitiveness.

[0059] The feedback optimization unit optimizes the generation logic of form fields and the distribution path configuration according to the data analysis results. When it is found that a certain field in a certain type of form data often has filling errors or has weak correlation with other fields, the system will adjust the generation logic of this field in the form, such as modifying data verification rules, adjusting field priorities, or deleting unnecessary fields. If the error rate of form data on the same distribution path exceeds the preset threshold, the distribution path configuration will be re-optimized, enabling the system to continuously adapt to the dynamic changes of the business and continuously improve the performance and practicality of the system.

[0060] The permission management unit dynamically controls the accessibility of form fields and the data distribution scope according to the user role identifier. Encrypts the form fields through the attribute-based encryption algorithm and binds the decryption key of the user role to desensitize or filter sensitive fields during data distribution. In a medical system, medical staff at different levels can only access patient form data within their permission scope. For example, ordinary nurses can only view the basic nursing information of patients, while doctors can view more detailed diagnosis and treatment information, effectively protecting patient privacy and data security.

[0061] The multimodal interaction unit supports form filling with mixed input of voice, image, and text. In practical applications, users can quickly fill in forms through voice input. For example, in a mobile office scenario, users do not need to manually enter text and can directly complete the form filling by speaking the form content through voice, improving the operation convenience. For images containing structured data, such as invoice images, the system can extract the information therein through an image recognition model and automatically fill it into the corresponding form fields, further improving the efficiency and accuracy of form filling and meeting the diverse interaction needs of different users in different scenarios. Description of the Drawings

[0062] Figure 1 It is the working principle diagram of the intelligent form distribution and data analysis system described in the present invention;

[0063] Figure 2 It is the working flowchart of the dialogue receiving unit;

[0064] Figure 3 It is the working flowchart of the extended functions of the form generation unit;

[0065] Figure 4 It is the flowchart of the data analysis unit's work. Detailed Embodiments

[0066] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0067] Please refer to Figures 1-4 , the present invention provides a technical solution: an intelligent form distribution and data analysis system based on dialogue, and the system includes:

[0068] Dialogue receiving unit: responsible for receiving the dialogue content input by the user and extracting the user intention and key entities therefrom. Through natural language processing technology, the user input is parsed, and the unstructured dialogue is converted into intention and entity information that the system can understand, providing basic data for subsequent form generation and other operations.

[0069] Form generation unit: Dynamically generate an adapted form template based on the user intention and key entities extracted by the dialogue receiving unit. The form template contains key information such as field types, data verification rules, and distribution path configurations. During the generation process, the field combinations in the preset form rule library are matched based on the user intention, the default values of the fields are filled with the key entities, and context-related additional fields are dynamically added according to the context to ensure that the generated form can accurately meet the user's needs.

[0070] Dialogue state tracking unit: Record the multi-round dialogue history and update the dialogue context state. By continuously tracking the dialogue state, the field priorities of the form template can be adjusted according to the context state, making the form filling process more in line with the user's dialogue logic and actual needs, and improving the user experience.

[0071] Data distribution unit: According to the distribution path configuration of the form template, direct the form data filled by the user to the target processing node. Ensure that the data can be accurately transmitted to the corresponding processing links, realizing the effective flow of data and the smooth progress of the business process.

[0072] Data analysis unit: Perform real-time clustering analysis and association rule mining on the form data to generate structured analysis results. Through in-depth analysis of a large amount of form data, potential laws and association relationships behind the data are mined, providing strong data support for business decisions.

[0073] Feedback optimization unit: Optimize the generation logic of the form fields and the distribution path configuration according to the analysis results generated by the data analysis unit. Continuously improve the performance and adaptability of the system, and improve the accuracy and efficiency of form distribution and data analysis.

[0074] The present invention will be further described below in conjunction with Embodiments 1 to 5:

[0075] Embodiment 1:

[0076] The dialogue receiving unit performs word segmentation and semantic parsing on the dialogue content input by the user through a natural language processing model. The natural language processing model adopts advanced deep learning algorithms, such as models based on the Transformer architecture, which can perform accurate word segmentation on the input text, splitting the continuous text into individual words or tokens. At the same time, through semantic parsing technology, intent classification labels and entity slots are extracted. For example, when the user inputs "I want to apply for a business license, and my company name is XX Technology Co., Ltd.", after processing by the model, the extracted intent classification label is "Apply for a business license", and the entity slot contains "Company name: XX Technology Co., Ltd.".

[0077] According to the preset intent-form mapping table, the intent classification label is mapped to the corresponding form type identifier. The intent-form mapping table is pre-stored in the system database and is a table containing the corresponding relationships between various intent classification labels and form type identifiers. When the intent classification label of "Apply for a business license" is obtained, the system can quickly find the corresponding form type identifier from the mapping table, such as "Business license application form".

[0078] Bind the key entities in the entity slot to the form fields to generate initial form filling suggestions. Combining the above example, the system binds the "Company name" in "Company name: XX Technology Co., Ltd." to the "Company name" field in the business license application form, and takes "XX Technology Co., Ltd." as the initial filling suggestion for this field. In this way, when generating the form later, this field will automatically fill in this suggested value, reducing the user's manual input operation and improving the form filling efficiency.

[0079] Through the above steps, the dialogue receiving unit can accurately extract key information from the user dialogue and provide basic data for form generation, ensuring the accuracy and efficiency of subsequent form generation, which is an important basic link for the entire system to achieve intelligent form distribution and data analysis.

[0080] Embodiment 2:

[0081] Based on the form template generated according to the user's intention and key entities, the form generation unit calculates the urgency score for field filling according to the list of incomplete fields in the dialogue context state. The dialogue context state records the form-related information of the user in multiple rounds of dialogue, including the filled fields and the unfilled fields. The system calculates the urgency score for filling each incomplete field according to factors such as the importance of the unfilled field and its relevance to the current business process. For example, in a loan application form, if the user has filled in the basic personal information but has not filled in the income information, and the income information is crucial for evaluating the loan amount, then the urgency score for filling the "income" field will be relatively high.

[0082] Dynamically sort the form fields based on the urgency score and generate a form interface with priority markings. The system sorts the form fields according to the calculated urgency score, and displays the fields with higher scores prominently in the form interface. In this way, when the user fills in the form, the fields that are more critical to the current business process are seen first, which helps the user fill in the form in a reasonable order, improving the filling efficiency and accuracy. The generated form interface will mark the priority of each field, such as using numbers "1", "2", "3", etc., to facilitate the user to intuitively understand the filling order.

[0083] If the key entity is missing, an additional follow-up field is triggered, and a follow-up request is initiated through the dialogue receiving unit. For example, in a product purchase form, the product model is a key entity, but it is not mentioned in the user's input dialogue. At this time, the form generation unit will trigger an additional follow-up field, such as "Please provide the product model you need to purchase", and send this follow-up request to the user through the dialogue receiving unit. When the user replies, the dialogue receiving unit extracts the key entity again, and the form generation unit further improves the form generation process based on the newly obtained key entity.

[0084] Through these extended functions, the form generation unit can generate forms more intelligently according to the user's dialogue situation and business requirements, optimize the form filling process, and improve the user experience and the quality of form data.

[0085] Example 3:

[0086] This example details the specific implementation method of the dialogue state tracking unit and its effects in the system. This unit plays a key role in the intelligent interaction of the system and the optimization of form generation.

[0087] The dialogue state tracking unit constructs a dialogue state graph to record the intention changes, entity additions, and field completion status for each dialogue turn. The dialogue state graph is a structure that graphically represents the dialogue state. The nodes in the graph represent different dialogue states, and the edges represent the transition relationships between the states. For example, in a hotel reservation dialogue, the user's initial intention is to query hotels, and at this time, the dialogue state graph records the intention as "query hotels". As the dialogue progresses, the user indicates that they want to reserve a hotel for a specific date and room type, and the intention changes. The dialogue state graph will record this intention change and update the entity additions (such as entities like reservation date, room type, etc.) and the field completion status (such as whether fields like the name of the occupant have been filled in).

[0088] The dialogue state graph is updated in real time through a graph neural network model to predict the potential field requirements for the next dialogue turn. The graph neural network model can effectively process the graph-structured data of the dialogue state graph. By learning and updating the information of the nodes and edges, it can reflect the changes in the dialogue state in real time. At the same time, based on the analysis of historical dialogue data and the current dialogue state, the model can predict the fields that the user may need to fill in during the next dialogue turn. For example, in a hotel reservation dialogue, when the user has selected the room type and check-in date, the model predicts that the user may need to fill in fields such as the name and contact information of the occupant.

[0089] Adjust the field combination strategy in the form generation unit according to the prediction results. The dialogue state tracking unit feeds back the predicted potential field requirements to the form generation unit, and the form generation unit adjusts the field combination strategy based on this information. For example, add the predicted fields to the form template in advance, or adjust the priorities of these fields to make them more prominently displayed in the form interface for the convenience of the user to fill in. In this way, through the real-time tracking and prediction of the dialogue state tracking unit, the form generation unit can generate forms that better meet the user's needs more intelligently, improving the fluency of the dialogue interaction and the convenience of form filling.

[0090] Embodiment 4:

[0091] The data distribution unit first parses the distribution path configuration in the form template, which includes the target node identifier, data format conversion rules, and transmission protocol. For example, in an internal enterprise approval process, the form data needs to be sent to the finance department for review, and the target node identifier is the system identifier of the finance department. The data format conversion rule may be to convert the data in the form from JSON format to XML format to meet the data reception requirements of the finance system. The transmission protocol may use the HTTP protocol for data transmission.

[0092] Query the routing table based on the target node identifier to determine the data transmission channel. The routing table is stored in the system's network configuration database, which records the correspondence between each target node identifier and the corresponding transmission channel. When the data distribution unit obtains the target node identifier, it queries the routing table to find the corresponding transmission channel, such as a specific server IP address and port number, to ensure that the data can be transmitted to the target node accurately and without error.

[0093] Encapsulate and send the form data to the target node according to the specified protocol through an asynchronous message queue. The asynchronous message queue adopts advanced message queue technologies such as Kafka or RabbitMQ. First, the form data is encapsulated according to the specified transmission protocol. For example, under the HTTP protocol, the data is encapsulated into a data packet that conforms to the HTTP request format. Then, the encapsulated data is sent to the asynchronous message queue, and the message queue will send the data to the target node according to the first-in, first-out principle. This method can improve the reliability and efficiency of data transmission, and avoid data loss or transmission delay caused by network fluctuations or the target node being busy.

[0094] The data analysis unit performs outlier detection on the numerical fields in the form data and uses a density-based clustering algorithm to divide the data distribution range. The density-based clustering algorithm (DBSCAN) is a commonly used data mining algorithm, and its principle is to identify clusters based on the density of data points. For the numerical fields in the form data, such as order amount, age, etc., the DBSCAN algorithm calculates the density around the data points. If the density of a certain data point is lower than the set threshold, it is determined as an outlier. At the same time, the algorithm will divide the data into different ranges according to the density distribution of the data points. For example, the order amount is divided into three ranges: low, medium, and high, for subsequent analysis.

[0095] Perform topic modeling on the text fields to extract high-frequency keywords and co-occurrence relationships. The topic modeling uses the Latent Dirichlet Allocation (LDA) algorithm, which can discover latent topics in text data. For the text fields in the form, such as customer feedback comments, product descriptions, etc., the LDA algorithm analyzes the words in the text, extracts high-frequency keywords, and discovers the co-occurrence relationships between these keywords. For example, in customer feedback comments, it is found that keywords such as "product quality" and "after-sales service" often appear together, indicating that there may be a relationship between these two aspects.

[0096] Fuse the clustering results with the topic modeling results to generate a multi-dimensional association rule graph. The data analysis unit integrates the clustering results of numerical fields and the topic modeling results of text fields to construct a multi-dimensional association rule graph. The graph graphically shows the association relationships between different data elements, such as the association between order amount, product type, and customer feedback, providing comprehensive data insights for the enterprise and helping the enterprise make more accurate decisions.

[0097] Through the collaborative work of the data distribution unit and the data analysis unit, the effective transmission and in-depth analysis of form data are realized, providing strong support for the business operation and decision-making of the enterprise.

[0098] Example 5:

[0099] The permission management unit encrypts form fields through the attribute-based encryption algorithm and binds the decryption keys of user roles. The attribute-based encryption algorithm (ABE) is a new type of encryption technology that performs encryption and decryption operations based on user attributes. In this system, according to different attributes of user roles, such as employee level, department, etc., corresponding decryption keys are generated for each user role. For example, for senior employees in the finance department, their role attributes include "finance department" and "senior", and the system will generate a specific decryption key based on these attributes. When encrypting form fields, the form fields are associated with the attributes of user roles, and only user roles with corresponding attributes can use their decryption keys to decrypt the form fields, thus ensuring the security of form data.

[0100] During data distribution, desensitize or filter sensitive fields according to the permission level of the target node. The system pre-defines the permission levels of different target nodes and the range of accessible fields corresponding to each permission level. For example, for some sensitive fields involving user privacy, such as ID numbers, bank card numbers, etc., when the data is distributed to target nodes with lower permissions, the system will desensitize these sensitive fields, such as replacing the middle digits of the ID number with "*", or directly filtering out these fields without transmission, further protecting the security of user data.

[0101] The permission management unit embeds a permission verification module in the dialogue receiving unit to real-time verify the operation permissions of user roles for the current form fields. When the user inputs dialogue content, the permission verification module will, according to the user's role information, real-time verify whether the user has the right to operate the current form fields involved. For example, when an ordinary employee applies for reimbursement, the permission verification module will check whether the employee has the right to fill in fields such as reimbursement amount.

[0102] If an unauthorized access request is detected, the dialogue state tracking unit is triggered to insert a permission application follow-up field and pause the form generation process. When the permission verification module detects that a user has unauthorized access, it notifies the dialogue state tracking unit to insert a permission application follow-up field, such as "You do not have permission to operate this field. Do you want to apply for permission?", and at the same time pauses the form generation process and waits for the user's response. If the user chooses to apply for permission, the system will process it according to the preset permission application process; if the user gives up the application, the system will continue to prompt the user that they can only operate the fields for which they have permission.

[0103] The multimodal interaction unit converts the voice input into text through a speech recognition model and extracts the structured data in the image through an image recognition model. The speech recognition model adopts advanced speech recognition technologies, such as speech recognition algorithms based on deep learning, and can accurately convert the user's voice input into text information. The image recognition model uses technologies such as convolutional neural networks (CNNs) to analyze the images uploaded by the user and extract the structured data therein. For example, in an invoice image, it can recognize information such as the invoice number, amount, and date.

[0104] Match the multimodal input data with the form fields. If the match fails, trigger the manual review process. The multimodal interaction unit automatically matches the text converted from the voice and the data extracted from the image with the form fields. For example, it matches the amount data extracted from the invoice image with the "reimbursement amount" field in the reimbursement form. If the match fails, the system will automatically trigger the manual review process, and the data will be reviewed and processed manually to ensure the accuracy and integrity of the data.

[0105] Through the implementation of the functions of the permission management unit and the multimodal interaction unit, while ensuring data security, the system provides users with more convenient and diverse interaction methods, improving the overall performance and user experience of the system.

[0106] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "including", "comprising" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device.

[0107] Although embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An intelligent form distribution and data analysis system based on dialogue, characterized in that It includes: A dialogue receiving unit, which is used to receive the dialogue content input by the user and extract the user intention and key entities in the dialogue; A form generation unit, which is used to dynamically generate an adapted form template according to the user intention and key entities, and the form template includes field types, data verification rules and distribution path configurations; A dialogue state tracking unit, which is used to record the multi-round dialogue history, update the dialogue context state, and adjust the field priorities of the form template according to the context state; A data distribution unit, which is used to direct the form data filled in by the user to the target processing node according to the distribution path configuration of the form template; A data analysis unit, which is used to perform real-time clustering analysis and association rule mining on the form data to generate a structured analysis result; A feedback optimization unit, which is used to optimize the generation logic of form fields and the distribution path configuration according to the analysis result; Among them, the execution steps of the form generation unit include: Match the field combination in the preset form rule library based on the user intention, fill in the field default values according to the key entities, and dynamically add additional fields related to the context.

2. The system according to claim 1, wherein The execution steps of the dialogue receiving unit include: Perform word segmentation and semantic parsing on the dialogue content input by the user through a natural language processing model, and extract the intention classification label and entity slot; According to the preset intention-form mapping table, map the intention classification label to the corresponding form type identifier; Bind the key entities in the entity slot to the form fields to generate an initial form filling suggestion.

3. The system according to claim 2, wherein, The execution steps of the form generation unit further include: Calculate the field filling urgency score according to the unfinished field list in the dialogue context state; Dynamically sort the form fields based on the urgency score and generate a form interface with priority marks; If the key entity is missing, trigger an additional follow-up field and initiate a follow-up request through the dialogue receiving unit.

4. The system according to claim 3, wherein The execution steps of the dialogue state tracking unit include: Construct a dialogue state graph to record the intention change, entity addition and field completion status of each dialogue round; Real-time update the dialogue state graph through a graph neural network model to predict the potential field requirements of the next round of dialogue; Adjust the field combination strategy in the form generation unit according to the prediction result.

5. The system according to claim 1, wherein The execution steps of the data distribution unit include: Parse the distribution path configuration in the form template, and the configuration includes the target node identifier, data format conversion rules and transmission protocols; Query the routing table according to the target node identifier to determine the data transmission channel; Encapsulate and send the form data to the target node through an asynchronous message queue according to the specified protocol.

6. The system according to claim 5, wherein The execution steps of the data analysis unit include: Perform outlier detection on the numerical fields in the form data, and use a density-based clustering algorithm to divide the data distribution interval; Perform topic modeling on the text fields to extract high-frequency keywords and co-occurrence relationships; Fuse the clustering result and the topic modeling result to generate a multi-dimensional association rule graph.

7. The system according to claim 6, wherein The execution steps of the feedback optimization unit include: Update the field priority weights in the form rule library according to the frequently occurring field combinations in the association rule graph; Adjust the threshold range in the data verification rule based on the outlier detection result; If the error rate of form data on the same distribution path exceeds the preset threshold, trigger the re-optimization of the distribution path configuration.

8. The system according to claim 1, wherein It also includes: A permission management unit for dynamically controlling the accessibility of form fields and the data distribution range according to the user role identifier; The execution steps of the permission management unit include: Encrypt the form fields through an attribute-based encryption algorithm and bind the decryption key of the user role; During data distribution, desensitize or filter sensitive fields according to the permission level of the target node.

9. The system according to claim 8, wherein The execution steps of the permission management unit also include: Embed a permission verification module in the dialogue receiving unit to real-time verify the operation permission of the user role for the current form field; If an unauthorized access request is detected, trigger the dialogue state tracking unit to insert a permission application follow-up field and pause the form generation process.

10. The system according to claim 1, wherein It also includes: A multimodal interaction unit for supporting the form filling with mixed input of voice, image and text; The execution steps of the multimodal interaction unit include: Convert the voice input into text through a voice recognition model, and extract the structured data in the image through an image recognition model; Match the multimodal input data with the form fields, and trigger the manual review process if the match fails.

Images