Detailed Description
The application solves the technical problems of long implementation period, dependence on professionals and high cost of digitization, maintenance and modification faced by the prior art by providing a method for generating the service form template in a rapid digitization way.
In order to solve the above problems, the technical embodiment adopts the following overall concept:
Firstly, carrying out data analysis on input data, and generating a matching constraint based on a data analysis result, wherein the data analysis comprises the steps of carrying out data judgment on the input data; if the judgment result is the image data of the paper form, inputting the image data into an intelligent segmentation network; performing layout contour recognition of the image data through a layout contour recognition sub-network to generate a layout contour recognition result; synchronizing the layout contour recognition result to a character symbol recognition sub-network, performing granularity segmentation on the image data through the layout contour recognition result, and performing character symbol content recognition in a unit based on granularity; synchronizing a text symbol content recognition result and the layout contour recognition result to a positioning optimization sub-network, performing positioning optimization of relative coordinates, and constructing a business form template, wherein the layout contour recognition sub-network, the text symbol recognition sub-network and the positioning optimization sub-network are processing units of an intelligent segmentation network; visually displaying the business form template to the account main body, and recording feedback data of the account main body; and selecting a function form according to the feedback data, combining the function forms to generate a service form template, and sending the service form template to the account main body, wherein the service form template is in an editable state. And further, the technical effects of high usability, simple operation and reduced complicated workload are achieved without professional IT technicians to perform business analysis, knowledge base design, code implementation and the like.
In order to better understand the above technical solutions, the following detailed description will be given with reference to the accompanying drawings and specific embodiments, and it should be noted that the described embodiments are only some embodiments of the present application, and not all embodiments of the present application, and it should be understood that the present application is not limited by the exemplary embodiments described herein. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to fall within the scope of the application. It should be further noted that, for convenience of description, only some, but not all of the drawings related to the present application are shown.
Example 1
As shown in fig. 1, the present application provides a method for generating a service form template in a rapid digitization manner, where the method includes:
Carrying out data analysis on input data, and generating a matching constraint based on a data analysis result, wherein the data analysis comprises the steps of carrying out data judgment on the input data;
Alternatively, the input data is non-digitized raw form data, and there are a variety of data formats, including, by way of example, entity data (paper), image data, text data, keywords, and the like. Specifically, the input data comprises a combination of one or more of the above-described data formats.
Optionally, the matching constraint is a condition or rule for filtering or manipulating the data. Including filtering out data that does not meet certain conditions, selecting only data that meets certain conditions, and the like. Depending on the nature of the matching constraint, an appropriate tool or library is selected to achieve functional matching.
Optionally, first, a data judgment of the input data is performed, where the data judgment is used to determine a data format type included in the input data. The data judging method comprises machine vision-based judgment, data file format-based judgment and the like.
If the judgment result is the image data of the paper form, inputting the image data into an intelligent segmentation network;
Optionally, if the judgment result is the image data of the paper form, inputting the image data into the intelligent segmentation network. The intelligent segmentation network is a segmentation network based on a deep learning technology, and the intelligent segmentation network is used for analyzing, dividing and identifying the image so as to better understand the content in the image, and carrying out main body identification of the convolution network on the image data to form a form segmentation result matched with the image data. The intelligent segmentation network comprises a layout contour recognition sub-network, a character symbol recognition sub-network and a positioning optimization sub-network.
Performing layout contour recognition of the image data through a layout contour recognition sub-network to generate a layout contour recognition result;
Optionally, the layout profile recognition sub-network is used for performing data recognition on the image data to construct a layout profile of the target paper form, and specifically, the layout profile includes a table profile, an image profile, and the like, a dividing line, a merging cell, and the like.
Illustratively, first, preprocessing is performed on input image data, including denoising, graying, binarization, edge detection, and the like, to improve image quality. Then, key targets in the image, including table contours, image contours, segmentation lines, etc., are identified using a target detection algorithm, such as Convolutional Neural Network (CNN) or other deep learning model. The key objective is to build the basic elements of the layout outline. Further, the detected objects are subjected to contour analysis, the hierarchical relationship and the relative position between the objects are determined, and a hierarchical structure of the layout is constructed, for example, cells, headers, and the like in the table are determined. And finally, generating a complete layout contour based on the results of target detection and contour analysis, and obtaining a layout contour recognition result. The layout outline recognition result is a hierarchical structure containing tables, images and other elements, describing the layout relationship and position of each part in the images.
Synchronizing the layout contour recognition result to a character symbol recognition sub-network, performing granularity segmentation on the image data through the layout contour recognition result, and performing character symbol content recognition in a unit based on granularity;
Optionally, after the identification of the layout contour identification sub-network is completed, synchronizing the identification result to the character symbol identification sub-network, and then, completing granularity segmentation by the character symbol identification sub-network according to the synchronization result, taking each granularity segmentation result as a granularity unit, executing character symbol content identification in the granularity unit, and generating a character symbol content identification result; illustratively, the granularity segmentation results in a plurality of form regions composed of layout contours, including closed regions and non-closed regions.
And carrying out character symbol content recognition in a plurality of granularity units based on the granularity segmentation result, ensuring the one-to-one correspondence between the character symbol content recognition result and the form layout, and avoiding character symbol content recognition errors generated by recognition cross-region.
Synchronizing the text symbol content recognition result and the layout contour recognition result to a positioning optimization sub-network, performing positioning optimization of relative coordinates, and constructing a business form template, wherein the layout contour recognition sub-network, the text symbol recognition sub-network and the positioning optimization sub-network are processing units of an intelligent segmentation network.
Optionally, the positioning optimization sub-network is used for identifying the relative coordinates of the layout outline and the text symbol, and integrating the layout outline and the text symbol content identification result based on the relative coordinates to generate a form construction result. Specifically, the form construction result is an editable active form state.
Optionally, each functional unit is sequenced according to the similarity value, and a service form template is constructed for screening by the user. Illustratively, the order matching results are generated by performing a combination arrangement based on the occurrence order and the correlation relationship of the respective functional units in the matching constraint. Wherein, each functional unit is mutually combined and nested to form a complete digital form function realization path. The sequence matching result is used to determine the execution sequence of the respective functional units.
Optionally, the service form template includes a flotation keyword generated based on the recognition result, where the flotation keyword is generated based on the principles of methods such as semantic expansion, fuzzy search, normal association, and the like, and is associated with the recognition result. By providing flotation keywords for users to select, abnormal generation of subsequent forms caused by poor keyword presentation due to forgetting or inaccurate description of the users is avoided.
Optionally, the layout outline recognition sub-network, the character symbol recognition sub-network and the positioning optimization sub-network are processing units of the intelligent segmentation network, and jointly form the intelligent segmentation network. The layout contour recognition sub-network, the character symbol recognition sub-network and the positioning optimization sub-network are connected in an initial position, namely, the output end of the layout contour recognition sub-network is connected with the input end of the character symbol recognition sub-network, and the output end of the character symbol recognition sub-network is connected with the input end of the positioning optimization sub-network. In addition, the intelligent segmentation network also comprises a possible storage unit and a transmission unit.
Visually displaying the business form template to the account main body, and recording feedback data of the account main body;
optionally, the generated order matching results are visually presented to the account body. This may take the form of a graphical interface, chart or other form so that the account body can more easily understand the flow of processing data by the system.
Optionally, the generated business form template is visually displayed to the account body. The visualization displays flotation options through a user interface and provides interactive means for user selection. Illustratively, the user interface includes web pages, interactive devices, applications, applets, and the like.
Optionally, the service form template is inspected and fed back through the visualized delivery account body, and the selection result of the account body is recorded. The feedback data of the account body reflects understanding and attention points of the user to the business form template. The feedback data of the account body is a selected business form template and possible template reconstruction information.
And selecting a function form according to the feedback data, combining the function forms to generate a service form template, and sending the service form template to the account main body, wherein the service form template is in an editable state.
Optionally, the feedback data includes a function form selected by the user body and a form function sequence of the function form. The feedback data reflects the user's body's preferences, habits and special needs for the system. According to the feedback data, the technical effect of selecting the function form matched with the user requirement is achieved.
Optionally, the feedback data further includes layout optimization, text color adjustment, default setting, validation rule formulation, etc., performed by the user body. And combining the function forms by selecting the function forms according to the feedback data to form a complete service form template. The business form template comprises various input fields, check boxes, drop-down menus and the like so as to meet different requirements of users.
Optionally, based on the feedback data, each of the function forms is arranged in a combined manner to form a digitized service form template, and the service form template is sent to the user body in an editable state. The user main body can be modified and customized according to the needs and the preferences of the user main body, so that the personalized needs of the user main body are met, the cost of digitizing the user main body form is reduced, and the capability of the user main body for realizing digital transformation through the system is improved.
Optionally, after receiving the service form template, the user may issue a form to implement data collection, perform related maintenance work of the data, and implement digital use of the form. The business form template sent to the account main body is selected to be a practical form which can be directly put into use, and has all form functions and effects expected by the user main body. In addition, the system automatically realizes the construction of the data table based on the published form, thereby facilitating the inquiry and management of the subsequent data.
Further, as shown in fig. 2, the method further includes:
establishing a form knowledge base, wherein the form knowledge base is constructed based on big data, and each form data in the form knowledge base is in an editable state;
Performing multi-granularity splitting on form data in the form knowledge base, performing functional identification on a multi-granularity splitting result, and performing form clustering according to the functional identification result to generate a multi-granularity clustering result;
And acquiring input account main body data, taking the input account main body data as auxiliary constraint, taking the matching constraint as main body constraint, executing function matching of the multi-granularity clustering result, and generating a business form template.
Optionally, various form samples in the target business field are collected through big data, including forms of different types and structures, a form knowledge base is built, and the form knowledge base covers various situations which may occur in the target business process. Knowledge base types include relational (e.g., mySQL, postgreSQL) or non-relational (e.g., mongoDB, cassandra).
Optionally, setting the knowledge base data model based on the flow in the target service field and the related form features, where the setting of the knowledge base data model includes a mode of storing form data and a relation between forms.
Optionally, by adding corresponding fields in the knowledge base, the editing state of the form data is identified, and each form data in the knowledge base is ensured to be in an editable state. By constructing the form knowledge base, a reliable basis is provided for storing and managing a large amount of form data, and meanwhile, the data is ensured to be in an editable state so as to meet business requirements.
Optionally, appropriate security measures, such as access control, encryption, etc., are implemented in order to protect the security of the form data. The form data has corresponding version and time identification, and is used for performing version control on the form data so as to track and manage the change of the data.
Multi-granularity splitting refers to splitting or splitting an entire or a data set according to different granularities. In the form data analysis, data are split according to different outline grades, such as primary title, secondary title, text, attachment and the like; through multi-granularity splitting, one large form data is split into a plurality of small data clusters, so that function allocation and identification are facilitated. The data or tasks can be managed and analyzed more flexibly, and different requirements and scenes can be better adapted. The complex problems are better understood and handled, and the efficiency and accuracy are improved.
Optionally, different granularity division results correspond to different sub-module structures, each sub-module structure has a corresponding module function, and the multi-granularity division results are subjected to function identification, so that the form functions are convenient to refine, analyze and understand, and form clustering based on the functions is facilitated.
Alternatively, the result of the function identification is used for form clustering, i.e. form data with similar functions or attributes are classified into the same category. Form clustering is realized through a clustering algorithm, such as K-means clustering, hierarchical clustering and the like. Specifically, each form in the form repository has a form type identifier, and forms belonging to the same form type have the same or similar module function combinations. And clustering the forms by using clustering constraint based on module function combination aiming at the similar forms. The method has the advantages of reducing the calculated amount in the subsequent matching process and improving the calculation speed.
Alternatively, the multi-granularity clustering result is a hierarchical structure, which contains clustering information on different granularities. Form classification on multiple granularity levels is realized, and then follow-up form matching and screening of different granularity levels are facilitated, various and changeable form digitization requirements can be better met, and matching degree between a form generation result and user requirements is improved.
Optionally, based on the update of form data in the knowledge base, the multi-granularity clustering result is updated in real time to reflect the latest data state. Illustratively, this is accomplished by periodic batch processing or real-time streaming processing.
Further, the step of performing data analysis on the input data and generating a matching constraint based on the data analysis result further includes:
If the judging result is the combined data of the equipment image data and the keyword data, inputting the equipment image data into a convolution network, carrying out equipment main body identification, and generating an image keyword based on the main body identification result;
reconstructing a matching keyword by using the image keyword and the keyword data, and generating the matching constraint by using the matching keyword.
If the judging result shows that the shot device image and the keywords are taken as input data, carrying out main body recognition based on a convolution network on the device image data, and regenerating the image keywords according to the main body recognition result and the keywords. Wherein the image keywords describe the content of the device by mapping the device body to the corresponding keywords or tags. The convolutional network is used for image recognition tasks to identify objects in the device image by learning features in the image.
Optionally, the introduced form is subject-recognized by introducing OCR technology, NLP technology, and job form large model technology, to form a layout unit, a text unit, and an input unit. Wherein the layout element includes identifying layout-related portions of the form, such as dividing lines, groupings, merging cells, etc., based on the techniques or models described above. The text unit includes text portions, such as titles, subtitles, descriptive text, etc., in the form identified based on the techniques or models described above. The input unit includes a part of the form identified based on the above technique or model, such as a check box, a radio box, text filling, etc., that needs to be filled in or checked by the user.
Optionally, the generated image keywords are reconstructed and matched with the keyword data in the input. The method comprises the steps of keyword de-duplication, combination, expansion or other modes so as to comprehensively consider the image keywords and the keyword data in the input and obtain the matching keywords. Further, matching constraints are provisions for data relationship or consistency, facilitating subsequent processing and analysis.
Further, the step of performing data analysis on the input data and generating a matching constraint based on the data analysis result further includes:
establishing a keyword association database, and performing association matching of the keyword association database according to the image keywords after the image keywords are generated;
Generating a flotation phrase based on the association matching result, visually displaying the flotation phrase to the account main body, and recording the selection result of the account main body;
And adding the selection result to the matching keyword to regenerate the matching constraint.
Optionally, the association matching of the keyword association database is used for determining the associated keywords of the image keywords, so that the semantic scope of the image keywords is enlarged. The flotation phrase generated by the association matching result contains more possible image related keywords, which is helpful for users and systems to more comprehensively understand the content represented by the images.
Optionally, the generated flotation phrase is visually presented to the account body (user). The visualization displays the flotation phrase through a user interface and provides an interactive way for the user to select. Illustratively, the user interface includes web pages, interactive devices, applications, applets, and the like.
Optionally, selecting a flotation phrase by visually delivering the flotation phrase to the account main body, and recording the selection result of the account main body. The selection result of the account body reflects understanding and attention points of the user to the image keywords. The understanding of the image keywords is continuously optimized through participation and feedback of the user, the intelligent level of the system is improved, and the comprehensiveness and accuracy of the acquisition of the image keywords are ensured.
Further, obtaining input account main body data, taking the input account main body data as auxiliary constraint, taking the matching constraint as main body constraint, and executing function matching of multi-granularity clustering results, wherein the steps comprise:
Establishing an account database, wherein the account database is constructed according to the registration data and the historical use data of an account;
Calling the input account main body data from the account database, and taking the input account main body data as auxiliary constraint;
Before performing functional matching of the multi-granularity clustering result, performing pre-matching of the multi-granularity clustering result through the auxiliary constraint;
and performing functional matching of the main body constraint and the multi-granularity clustering result based on the pre-matching result.
Optionally, the account database includes basic information of the user, data provided at the time of registration, and historical operation and usage records thereof in the system by acquiring registration data and historical usage data of the account.
Optionally, when the account main body data is required to be used, the account database is activated, the data of the corresponding account is called from the account database through the data query statement, and the acquired account main body data is used as the auxiliary constraint. Auxiliary constraints are used to help the system better understand and meet the needs of the user.
Optionally, before functional matching, pre-matching of multi-granularity clustering results based on auxiliary constraint is performed. And carrying out preliminary screening and sorting on possible function matching results in advance through account main body data. The range of function matching is reduced, the matching flow is optimized, and the efficiency and the speed of function matching are improved.
Further, performing functional matching between the subject constraint and the multi-granularity clustering result based on the pre-matching result, and the steps further include:
Carrying out importance division on the keywords through the matching constraint to generate division weights;
And setting the granularity of the unit cells by the dividing weight, and completing functional matching of the main constraint and the multi-granularity clustering result according to the granularity setting result and the pre-matching result.
Optionally, the AI digitizing form engine is activated to match and translate the units identified by the multi-granularity clustering result into a digitizing component. Exemplary, include an input box component, a radio box component, a check box component, a signature component, a layout component, a text component, and the like.
Optionally, taking the data type in the pre-matching as the newly added feature constraint, carrying out data matching on the sub-module structure of the lowest granularity division result according to the importance degree of the newly added feature constraint on the corresponding format unit, and filling the sub-module structure into the pre-matching result according to the matching result. And finally generating an intelligent form. In addition, according to the requirement of the user, the coverage matching filling of the lowest granularity division result can be performed by taking the format unit as a unit.
Further, in a possible embodiment, the method further comprises:
establishing an account feature set of an account, wherein the account feature set is a habit feature set and a general feature set in an account construction completion form;
Taking the matching constraint as a matching feature to perform account feature set matching, and generating a calling constraint according to a matching result;
and carrying out combination optimization on the business form templates based on the calling constraint.
Optionally, the account feature set is established according to the account form construction record, and the habit feature set and the general feature set are obtained by carrying out feature extraction on the habit feature and the general feature in the account construction completion form. The habit feature set reflects personalized preference and operation habit of the user, and the general feature set contains general user information. Illustratively, the custom feature set includes a user's commonly used form construction features, form level setting features, hierarchical phrase features, unit features, font features, and the like. The general feature set includes service domain features, service aging features, etc. of the user.
And matching the account feature set with the matching features of the business form by using the matching constraint. The system is helped to better know the demands, preferences and operation habits of the user, generate a business form which better accords with the expectations of the user, reduce the frequency and the data volume of the feedback data of the account main body, and improve the user experience. Wherein the invocation constraint provides detailed information about the user's needs. And carrying out combination optimization of the business form templates according to the calling constraint, wherein the combination optimization comprises the steps of adjusting the arrangement sequence of form elements, changing the types of fields or adding specific functions. The combination optimization of the business form templates improves the adaptation degree of form generation and user requirements.
In summary, the method for generating the service form template in a rapid digitization manner has the following technical effects:
Generating a matching constraint based on a data analysis result by carrying out data analysis on the input data, wherein the data analysis result comprises data judgment on the input data; if the judgment result is the image data of the paper form, inputting the image data into an intelligent segmentation network; performing layout contour recognition of the image data through a layout contour recognition sub-network to generate a layout contour recognition result; synchronizing the layout contour recognition result to a character symbol recognition sub-network, performing granularity segmentation on the image data through the layout contour recognition result, and performing character symbol content recognition in the unit based on granularity; synchronizing the text symbol content recognition result and the layout contour recognition result to a positioning optimization sub-network, and executing positioning optimization of relative coordinates to construct a business form template, wherein the layout contour recognition sub-network, the text symbol recognition sub-network and the positioning optimization sub-network are processing units of an intelligent segmentation network; visually displaying the business form template to the account main body, and recording feedback data of the account main body; selecting a function form according to the feedback data, combining the function forms to generate a service form template, and sending the service form template to the account main body, wherein the service form template is in an editable state. And further, the technical effects of high usability, simple operation and reduced complicated workload are achieved without professional IT technicians to perform business analysis, knowledge base design, code implementation and the like.
Example two
Based on the same conception as the method for generating the rapid digitization of the business form template in the embodiment, as shown in fig. 3, the application also provides a system for generating the rapid digitization of the business form template, which comprises:
the constraint generating module 11 is configured to perform data analysis on input data, generate a matching constraint based on a data analysis result, and perform data judgment on the input data;
The intelligent segmentation module 12 is configured to input the image data into an intelligent segmentation network if the judgment result is the image data of the paper form;
a layout contour recognition module 13, configured to perform layout contour recognition of the image data through a layout contour recognition sub-network, and generate a layout contour recognition result;
The granularity content recognition module 14 is used for synchronizing the layout contour recognition result to the character symbol recognition sub-network, performing granularity segmentation of the image data through the layout contour recognition result, and performing character symbol content recognition in the unit based on the granularity unit;
The positioning optimization module 15 is configured to synchronize a text symbol content recognition result and the layout contour recognition result to a positioning optimization sub-network, perform positioning optimization of relative coordinates, and construct a service form template, where the layout contour recognition sub-network, the text symbol recognition sub-network, and the positioning optimization sub-network are processing units of an intelligent segmentation network;
the sequence display module 16 is configured to visually display the service form template to the account body, and record feedback data of the account body;
The template generating module 17 is configured to select a functional form according to the feedback data, combine the functional forms to generate a service form template, and send the service form template to the account main body, where the service form template is in an editable state.
Further, the granular content identification module 14 is further configured to:
establishing a form knowledge base, wherein the form knowledge base is constructed based on big data, and each form data in the form knowledge base is in an editable state;
Performing multi-granularity splitting on form data in the form knowledge base, performing functional identification on a multi-granularity splitting result, and performing form clustering according to the functional identification result to generate a multi-granularity clustering result;
further, the constraint generating module 11 is further configured to:
If the judging result is the combined data of the equipment image data and the keyword data, inputting the equipment image data into a convolution network, carrying out equipment main body identification, and generating an image keyword based on the main body identification result;
reconstructing a matching keyword by using the image keyword and the keyword data, and generating the matching constraint by using the matching keyword.
Further, the sequential presentation module 16 is further configured to:
establishing a keyword association database, and performing association matching of the keyword association database according to the image keywords after the image keywords are generated;
Generating a flotation phrase based on the association matching result, visually displaying the flotation phrase to the account main body, and recording the selection result of the account main body;
And adding the selection result to the matching keyword to regenerate the matching constraint.
Further, the granular content identification module 14 is further configured to:
Establishing an account database, wherein the account database is constructed according to the registration data and the historical use data of an account;
Calling the input account main body data from the account database, and taking the input account main body data as auxiliary constraint;
Before performing functional matching of the multi-granularity clustering result, performing pre-matching of the multi-granularity clustering result through the auxiliary constraint;
and performing functional matching of the main body constraint and the multi-granularity clustering result based on the pre-matching result.
Further, the granular content identification module 14 is further configured to:
Carrying out importance division on the keywords through the matching constraint to generate division weights;
And setting the granularity of the unit cells by the dividing weight, and completing functional matching of the main constraint and the multi-granularity clustering result according to the granularity setting result and the pre-matching result.
Further, the system further comprises a user feature optimization unit for:
establishing an account feature set of an account, wherein the account feature set is a habit feature set and a general feature set in an account construction completion form;
Taking the matching constraint as a matching feature to perform account feature set matching, and generating a calling constraint according to a matching result;
and carrying out combination optimization on the business form templates based on the calling constraint.
It should be understood that the embodiments mentioned in this specification focus on the differences from other embodiments, and the specific embodiment in the foregoing first embodiment is equally applicable to the system for generating the rapid digitization of a service form template described in the second embodiment, which is not further developed herein for brevity of the specification.
It is to be understood that both the foregoing description and the embodiments of the present application enable one skilled in the art to utilize the present application. While the application is not limited to the embodiments described above, obvious modifications and variations of the embodiments described herein are possible and are within the principles of the application.