An intelligent filling method based on natural language processing

Abstract

The application belongs to the technical field of software development, and particularly relates to an intelligent filling method based on natural language processing. The application is used to solve the problem of low filling efficiency of the existing intelligent filling. The application provides an intelligent filling method based on natural language processing, solves the problems of low manual input efficiency and low file analysis automation degree in the prior art, automatically extracts key data in multiple types of files by relying on OCR graphic-text recognition and NLP semantic analysis technology, substitutes traditional manual input by combining voice interaction technology, greatly compresses the data collection and filling period, and significantly improves the intelligent filling efficiency; the voice information effectiveness check of the field logic guidance of the sequential filling avoids the missing filling and data confusion in the filling process, reduces the human filling error by providing the standardized format examples of the form viewing module, and effectively guarantees the compliance and logical consistency of the filling data.

Description

Technical Field

[0001] This invention belongs to the field of software development technology, specifically relating to an intelligent data entry method based on natural language processing. Background Technology

[0002] As global digital transformation deepens, data, as a core production factor, is driving a surge in demand for efficient and accurate data reporting across various industries: enterprises need to achieve multi-business system form interaction, government agencies face high-frequency scenarios such as cross-departmental material submission, and industries such as healthcare and finance also have specific data reporting needs. Furthermore, data formats are becoming increasingly complex, requiring users to adapt to multiple terminals, such as batch import on computers and real-time filling on mobile devices. Traditional reporting methods struggle to meet these demands: manual data entry involves repetitive work and is prone to errors. Integrating intelligent speech-to-text and OCR automatic recognition technologies can replace repetitive manual data entry, reducing the error rate to below 1%. Inconsistent data formats across systems and reliance on manual field mapping can lead to workflow disruptions. Standardized data interfaces and automatic field mapping engines can achieve automatic compatibility and seamless data synchronization across different system form formats, eliminating workflow disruptions caused by manual mapping. Existing tools have weak capabilities in parsing unstructured data. Introducing a deep learning-driven file parsing model can achieve accurate extraction and standardized conversion of unstructured data such as PDFs, scanned documents, and handwritten text. Therefore, intelligent data entry technology, which has the ability to parse unstructured data, adapt across systems, and be compatible with multiple terminals, has become a key requirement for breaking through bottlenecks.

[0003] However, while existing intelligent form filling applications have achieved technological breakthroughs in some scenarios, they still have significant shortcomings in core scenarios such as voice interaction, process adaptation, and file parsing, making it difficult to meet diverse form filling needs. Summary of the Invention

[0004] The purpose of this invention is to solve the problem of low data entry efficiency in existing intelligent data entry systems. It provides an intelligent data entry method based on natural language processing, comprising:

[0005] Step 1: Receive the form filling task triggered by the user.

[0006] The triggering conditions for receiving the user-triggered data entry task in step one include:

[0007] Users can wake up the system via voice to initiate a data entry task triggered by the user.

[0008] Alternatively, users can initiate a user-triggered data entry task by clicking the built-in voice entry button.

[0009] The simple process of voice wake-up is as follows: before the user begins to fill in the form, the system will continuously listen for a specific wake-up word;

[0010] The wake-up words include, for example, "Start filling out the form" or "Hello, assistant".

[0011] Upon recognizing the wake word, the component receives the user-triggered data entry task and begins processing the user's instructions. The component collects the user's voice signal in real time and automatically initiates noise reduction processing to filter out environmental noise and ensure the clarity of the voice signal.

[0012] Step 2: Based on the type of data entry task described in Step 1, determine the data entry mode from a variety of preset data entry modes;

[0013] The filling modes include: voice interaction filling mode, file interpretation filling mode, sequential filling mode, and individual filling mode;

[0014] The intelligent data entry application developed based on the method of this invention integrates core data entry function modules and selects appropriate components to implement the functions. The components used include: voice interaction component, file upload component, instant data entry component, and individual editing component. After the components are integrated to form a complete data entry page, no additional configuration is required. It directly supports users to call functions such as voice data entry, file interpretation, sequential data entry, and individual modification to complete the data entry operation.

[0015] The voice entry component is the core operating component supporting both voice entry and standalone entry functions. It can be directly invoked without additional user configuration. Its core function is to convert voice information into standardized entry data, reducing the cost of manual input. When using the voice entry function, users can activate the component by clicking the built-in "Voice Entry" button or speaking a wake word. The component will collect the user's voice signal in real time and automatically initiate noise reduction processing to filter out ambient noise and ensure the clarity of the voice signal. Subsequently, by calling models such as NLP semantic parsing and ASR speech recognition, the voice content is accurately converted into text, and semantic optimization is performed for professional terminology in different entry scenarios to avoid recognition errors. After conversion, the component will automatically match the text data to the current field to be filled. If there is a recognition error, the user can directly click on the field to manually correct it or trigger voice input again to re-record. The entire process does not require jumping to other pages or setting additional recognition parameters, greatly improving the efficiency of field entry.

[0016] The file upload component is specifically designed for file interpretation and data entry functions, serving as the key entry point for converting file data into data entry information. The component has pre-defined logic for adapting to multiple file formats, allowing users to use it directly without needing to configure formats. Operationally, users simply click on the component area or drag and drop files into the designated area to upload various common file formats. The component automatically verifies the size and validity. After file upload, the component automatically connects to the backend file parsing logic, eliminating the need for manual parsing by the user. Once parsing is complete, the component automatically fills the corresponding fields with the extracted file data. The entire process requires no complex parsing skills from the user, providing a convenient "upload and parse, parse and fill" experience.

[0017] The standalone editing component is a dedicated component for implementing "partial field modification / filling" functionality. It can operate independently without being linked to the entire form filling process. Its core advantage lies in its ability to flexibly locate target fields and achieve partial data updates. When using it, there is no need to enter a specific editing page. Simply click on any field in the current form, and the component will activate the voice fill function to recognize and fill in the voice input for that field. After the input is entered, only the data in the current field is updated, without affecting other already filled fields.

[0018] Step 3: Obtain the original information entered by the user in the filling mode;

[0019] Step 4: Input the original information into the intelligent data entry model based on natural language processing for processing to obtain the data entry result;

[0020] If there is a recognition error, users can directly click on the field to manually correct it, or trigger voice input again to re-record. The whole process does not require jumping to other pages or setting additional recognition parameters, which greatly improves the efficiency of field filling.

[0021] The original information entered by the user in step three includes: voice data entered by the user and file data uploaded by the user.

[0022] User-uploaded file data includes common data types such as Word documents, PDFs, and scanned documents.

[0023] In the voice-interactive form filling mode, users input information via voice. The voice-interactive form filling can convert the voice content into structured text and automatically match it to the corresponding form field. If there are errors in the recognized content, users can directly make adjustments via voice interaction.

[0024] In the document interpretation and data entry mode, users upload and parse various types of files, covering common types such as PDFs and scanned documents. The document interpretation and data entry mode utilizes OCR image and text recognition, NLP semantic parsing, and table structure technology to automatically extract key data from uploaded files, simplifying the data collection process for data entry.

[0025] In sequential form filling mode, the fields to be filled are output in the order of preset field rules; in sequential form filling, the fields to be filled are output in the order of preset field rules, and the user automatically jumps to the next question after answering the current question by voice, thus completing the entire form filling process step by step.

[0026] Individual form filling mode focuses on voice input for a single field; individual form filling can directly locate a single field in the form and trigger the dedicated voice input function for that field;

[0027] The intelligent form filling application developed based on the method of this invention can be directly used by embedding it into the target page. When embedded into the page, the page contains interactive units corresponding to the four core functions of the application: a voice input button for voice interactive form filling, a file upload area for file interpretation form filling, a step-by-step display area for fields for sequential form filling, and a quick editing entry for fields for individual form filling. These interactive units are pre-laid out according to user operating habits, forming a combined interface with interconnected functions on the page. Users do not need to make any settings; they can directly click on the corresponding unit to call up the required functions and complete the form filling operation.

[0028] The intelligent data entry model based on natural language processing includes: a voice interaction data entry module and a document interpretation data entry module.

[0029] The specific process for obtaining the original information entered by the user in step three is as follows:

[0030] When the user inputs voice data, the system retrieves the original information entered by the user from the mobile / computer device. The specific process is as follows:

[0031] A1: Establish a connection with the microphone on the mobile / computer device via a preset device interface protocol, and utilize the microphone on the mobile / computer device.

[0032] A2: The microphone collects user-input voice data in real time according to the device parameters of 16Hz sampling rate and 16-bit depth;

[0033] In addition, the collected real-time audio signals are transmitted to the voice interaction data entry module in real time via instructions encapsulated in the protocol.

[0034] Ensure stable signal transmission between devices and applications to avoid data loss or delay.

[0035] In step four, the original information is input into a natural language processing-based intelligent data entry model for processing to obtain the data entry result. The specific process is as follows:

[0036] When the original information to be filled in is voice data input by the user, the voice data input by the user is input into the voice interaction filling module for processing to obtain the voice filling result;

[0037] When the original information is file data uploaded by the user, the file data uploaded by the user is input into the file interpretation and filling module to obtain the file filling result.

[0038] The specific process of inputting user-inputted voice data into the voice interaction filling module for processing to obtain the voice filling result is as follows:

[0039] S1: Preprocess the user-input voice data to obtain preprocessed voice data;

[0040] S2: According to the preset segmentation duration, the preprocessed speech data is segmented to obtain frame sequence data;

[0041] S3: Use a feature extraction algorithm to extract speech features from the frame sequence data to obtain speech features;

[0042] S4: Use an automatic speech recognition algorithm to identify and process speech features to obtain the text sequence corresponding to the speech;

[0043] The automatic speech recognition algorithm is based on an end-to-end deep neural network model;

[0044] S5: Perform syntax verification and error correction on the text sequence corresponding to the speech to obtain the text sequence after syntax verification and error correction.

[0045] S6: Based on the field format requirements of the form, perform structured processing on the text sequence after syntax checking and error correction to obtain the voice filling result; automatically match the format specification of the current field to be filled to complete the adaptation of text and field data formats;

[0046] A computer storage medium, characterized in that the storage medium stores at least one instruction, which is loaded and executed by a processor to implement the intelligent data entry method based on natural language processing.

[0047] A natural language processing-based intelligent data entry device is characterized in that the device includes a processor and a memory, the memory storing at least one instruction, the at least one instruction being loaded and executed by the processor to implement the natural language processing-based intelligent data entry method.

[0048] A multi-terminal collaborative intelligent data entry service system, characterized in that the system includes a task receiving module for receiving data entry tasks triggered by users;

[0049] The mode determination module is used to determine the filling mode from a variety of preset filling modes based on the type of the filling task; the filling modes include: voice interaction filling mode, file interpretation filling mode, sequential filling mode and individual filling mode;

[0050] The information acquisition module is used to acquire the original information entered by the user in the filling mode;

[0051] The intelligent processing module is used to process the original information entered into the intelligent filling model based on natural language processing to obtain the filling results. The system is compatible with access from devices such as computers, mobile devices, and government self-service terminals, executes the intelligent filling method based on natural language processing, receives filling requests from each terminal, and feeds back the filling results.

[0052] The beneficial effects of this invention are as follows:

[0053] Firstly, this invention solves the technical problems of low efficiency in manual data entry and low automation in file parsing in the prior art. It automatically extracts key data from multiple types of files by relying on OCR image and text recognition and NLP semantic parsing technology, and combines voice interaction technology to replace traditional manual data entry, which greatly reduces the data collection and filling cycle and significantly improves the efficiency of intelligent filling.

[0054] Secondly, this invention solves the technical problems of missing, incorrect, and logically confused data in the prior art. By using the logical guidance of sequentially filled fields to verify the validity of voice information, it avoids missing and data confusion during the filling process. Combined with the standardized format examples provided by the sample form viewing module, it reduces human error in filling and effectively ensures the compliance and logical consistency of the filled data.

[0055] Third, this invention solves the technical problems of rigid processes and cumbersome partial data filling operations in the prior art. Through the single-field voice modification function for separate filling, it supports independent operation of partial data without linking the entire process filling steps. It can flexibly adapt to scenarios such as temporary filling and partial data updates, significantly enhancing the flexibility of intelligent filling.

[0056] Fourth, this invention solves the technical problems of high usage threshold and limited user coverage in the prior art. By combining voice interaction with sample form assistance, it breaks the dependence of digital operation on user input skills, reduces the user threshold, expands the user coverage of intelligent form filling, and realizes the dual implementation of technology accessibility and high-quality form filling experience.

[0057] Fifth, this invention solves the technical problems of low speech recognition accuracy, lack of closed-loop error correction, and non-automatic file parsing and field mapping in the prior art. By using the NLP semantic parsing and multi-round speech error correction capabilities of the speech interaction module, the accuracy of speech recognition and field matching is improved. Combined with the multi-technology fusion parsing and automatic field mapping of the file interpretation module, a fully automated closed loop of "parsing-filling-verification-error correction" is formed, further improving the practicality and reliability of the technical solution.

[0058] Sixth, this invention solves the problem of traditional form filling systems having a single terminal access and being unable to continue processing across devices. By building a unified multi-terminal service architecture, it is compatible with access from devices such as computers, mobile devices, and government self-service terminals, and supports the synchronous flow of form filling progress and draft data across terminals. This meets the needs of intelligent form filling for government affairs anytime and anywhere in different scenarios, and broadens the applicable scenarios of the system. Attached Figure Description

[0059] Figure 1 This is a flowchart illustrating the method of the present invention. Detailed Implementation

[0060] Specific implementation method one: Combining Figure 1 This invention is described;

[0061] Step 1: Receive the form filling task triggered by the user.

[0062] The triggering conditions for receiving the user-triggered data entry task in step one include:

[0063] The user wakes up the device via voice to initiate a data entry task triggered by the user.

[0064] Alternatively, users can initiate a user-triggered data entry task by clicking the built-in voice entry button.

[0065] The simple process of voice wake-up is as follows: before the user begins to fill in the form, the system will continuously listen for a specific wake-up word;

[0066] The wake-up words include, for example, "Start filling out the form" or "Hello, assistant".

[0067] Upon recognizing the wake word, the component receives the user-triggered data entry task and begins processing the user's instructions. The component collects the user's voice signal in real time and automatically initiates noise reduction processing to filter out environmental noise and ensure the clarity of the voice signal.

[0068] Step 2: Based on the type of data entry task described in Step 1, determine the data entry mode from a variety of preset data entry modes;

[0069] The filling modes include: voice interaction filling mode, file interpretation filling mode, sequential filling mode, and individual filling mode;

[0070] The intelligent data entry application developed based on the method of this invention integrates core data entry function modules and selects appropriate components to implement the functions. The components used include: voice interaction component, file upload component, instant data entry component, and individual editing component. After the components are integrated to form a complete data entry page, no additional configuration is required. It directly supports users to call functions such as voice data entry, file interpretation, sequential data entry, and individual modification to complete the data entry operation.

[0071] The voice entry component is the core operating component that supports voice entry and individual entry functions. It can be directly called without additional configuration by the user. Its core function is to convert voice information into standardized entry data, reducing the cost of manual input.

[0072] When using the voice entry function, users can activate the component by clicking the built-in "Voice Entry" button or speaking the wake word. The component will collect the user's voice signal in real time and automatically start noise reduction processing to filter out ambient noise and ensure the clarity of the voice signal. Then, by calling models such as NLP semantic parsing and ASR speech recognition, the voice content is accurately converted into text, and semantic optimization is performed for professional terms in different entry scenarios to avoid recognition errors. After the conversion is completed, the component will automatically match the text data to the current field to be filled. If there is a recognition error, the user can directly click on the field to manually correct it or trigger voice input again to re-record. The whole process does not require jumping to other pages or setting additional recognition parameters, which greatly improves the efficiency of field filling.

[0073] The file upload component is specifically designed for file interpretation and data entry functions, serving as the key entry point for converting file data into data entry information. The component has pre-defined logic for adapting to multiple file formats, allowing users to use it directly without needing to configure formats. Operationally, users simply click on the component area or drag and drop files into the designated area to upload various common file formats. The component automatically verifies the size and validity. After file upload, the component automatically connects to the backend file parsing logic, eliminating the need for manual parsing by the user. Once parsing is complete, the component automatically fills the corresponding fields with the extracted file data. The entire process requires no complex parsing skills from the user, providing a convenient "upload and parse, parse and fill" experience.

[0074] The standalone editing component is a dedicated component for implementing "partial field modification / filling" functionality. It can operate independently without being linked to the entire form filling process. Its core advantage lies in its ability to flexibly locate target fields and achieve partial data updates. When using it, there is no need to enter a specific editing page. Simply click on any field in the current form, and the component will activate the voice fill function to recognize and fill in the voice input for that field. After the input is entered, only the data in the current field is updated, without affecting other already filled fields.

[0075] Step 3: Obtain the original information entered by the user in the filling mode;

[0076] Step 4: Input the original information into the intelligent data entry model based on natural language processing for processing to obtain the data entry result;

[0077] If there is a recognition error, users can directly click on the field to manually correct it, or trigger voice input again to re-record. The whole process does not require jumping to other pages or setting additional recognition parameters, which greatly improves the efficiency of field filling.

[0078] Specific Implementation Method Two: The difference between this implementation method and Specific Implementation Method One is that:

[0079] The original information entered by the user in step three includes: voice data entered by the user and file data uploaded by the user.

[0080] User-uploaded file data includes common data types such as Word documents, PDFs, and scanned documents;

[0081] In the voice-interactive form filling mode, users input information via voice. The voice-interactive form filling can convert the voice content into structured text and automatically match it to the corresponding form field. If there are errors in the recognized content, users can directly make adjustments via voice interaction.

[0082] In the document interpretation and data entry mode, users upload and parse various types of files, covering common types such as PDFs and scanned documents. The document interpretation and data entry mode utilizes OCR image and text recognition, NLP semantic parsing, and table structure technology to automatically extract key data from uploaded files, simplifying the data collection process for data entry.

[0083] In sequential form filling mode, the fields to be filled are output in the order of preset field rules; in sequential form filling, the fields to be filled are output in the order of preset field rules, and the user automatically jumps to the next question after answering the current question by voice, thus completing the entire form filling process step by step.

[0084] Individual form filling mode focuses on voice input for a single field; individual form filling can directly locate a single field in the form and trigger the dedicated voice input function for that field;

[0085] The intelligent data entry application developed based on the method of this invention can be directly used by embedding it into the target page. When embedded into the page, the page contains interactive units corresponding to the four core functions of the application: a voice input button for voice interaction data entry, a file upload area for file interpretation data entry, a step-by-step display area for sequential data entry, and a quick editing entry for individual data entry. These interactive units are pre-laid out according to user operating habits, forming a combined interface with interconnected functions on the page. Users do not need to make any settings; they can directly click on the corresponding unit to call up the required functions and complete the data entry operation. Other steps and parameters are the same as in Specific Implementation Method 1.

[0086] Specific Implementation Method Three: The difference between this implementation method and Specific Implementation Methods One and Two is that:

[0087] The specific process for obtaining the original information entered by the user in step three is as follows:

[0088] When the user inputs voice data, the system retrieves the original information entered by the user from the mobile / computer device. The specific process is as follows:

[0089] A1: Establish a connection with the microphone on the mobile / computer device through the preset device interface protocol and call the microphone on the mobile / computer device;

[0090] A2: The microphone collects user-input voice data in real time according to the device parameters of 16Hz sampling rate and 16-bit depth;

[0091] In addition, the collected real-time audio signals are transmitted to the voice interaction data entry module in real time via instructions encapsulated in the protocol.

[0092] Ensure stable signal transmission between devices and applications to avoid data loss or delay;

[0093] The intelligent data entry model based on natural language processing includes: a voice interaction data entry module and a document interpretation data entry module. Other steps and parameters are the same as in one of the specific implementation methods one or two.

[0094] Specific Implementation Method Four: This implementation method differs from Specific Implementation Methods One to Three in that:

[0095] In step four, the original information is input into a natural language processing-based intelligent data entry model for processing to obtain the data entry result. The specific process is as follows:

[0096] When the original information to be filled in is voice data input by the user, the voice data input by the user is input into the voice interaction filling module for processing to obtain the voice filling result;

[0097] When the original information is file data uploaded by the user, the file data uploaded by the user is input into the file interpretation and filling module to obtain the file filling result;

[0098] The other steps and parameters are the same as those in one of the specific implementation methods one to three.

[0099] Specific Implementation Method Five: The difference between this implementation method and Specific Implementation Methods One to Four is that:

[0100] The specific process of inputting user-inputted voice data into the voice interaction filling module for processing to obtain the voice filling result is as follows:

[0101] S1: Preprocess the user-input voice data to obtain preprocessed voice data;

[0102] S2: According to the preset segmentation duration, the preprocessed speech data is segmented to obtain frame sequence data;

[0103] S3: Use the Mel Frequency Cepstral Coefficient (MFCC) algorithm to extract speech features from the frame sequence data to obtain speech features;

[0104] S4: Use an automatic speech recognition algorithm to identify and process speech features to obtain the text sequence corresponding to the speech;

[0105] The automatic speech recognition algorithm is based on an end-to-end deep neural network model; its core architecture includes, but is not limited to, encoder-decoder architecture, CTC (connectionist temporal classification) architecture, RNNT (recurrent neural network transducer) architecture, Conformer architecture, Transformer architecture, and any combination thereof.

[0106] S5: Perform syntax verification and error correction on the text sequence corresponding to the speech to obtain the text sequence after syntax verification and error correction.

[0107] S6: Based on the field format requirements of the form, perform structured processing on the text sequence after syntax verification and error correction to obtain the voice filling result; automatically match the format specification of the current field to be filled to complete the adaptation of text and field data format; other steps and parameters are the same as in one of the specific implementation methods one to four.

[0108] Specific Implementation Method Six: The difference between this implementation method and Specific Implementation Methods One through Five is that:

[0109] In step S1, the user-input voice data is preprocessed to obtain preprocessed voice data; the specific process is as follows:

[0110] S1.1: The user-input voice data is initially filtered using an adaptive noise suppression algorithm to obtain the initially filtered voice data;

[0111] The initial filtering is to filter out environmental interference signals.

[0112] S1.2: The speech data after the initial filtering is processed by the echo cancellation algorithm to obtain the speech data after secondary filtering.

[0113] The secondary filtering process is to eliminate interference caused by the device's own audio feedback.

[0114] S1.3: The signal gain adjustment and amplitude normalization of the secondary filtered speech data are performed by the root mean square normalization algorithm to obtain the preprocessed speech data;

[0115] Normalize to fit the range of subsequent processing to reduce transcription errors caused by differences in input volume;

[0116] The other steps and parameters are the same as those in any of the specific implementation methods one to five.

[0117] Specific Implementation Method Seven: The difference between this implementation method and Specific Implementation Methods One through Six is ​​that:

[0118] The file interpretation and reporting module includes: a file parsing unit, a preprocessing unit, a content extraction unit, and a data structuring unit;

[0119] The process involves inputting the user-uploaded file data into the file interpretation and reporting module to obtain the file reporting results; the specific process is as follows:

[0120] B1: The user-uploaded file data is input into the file parsing unit for processing to obtain file metadata; the specific process is as follows:

[0121] Based on a standardized file transfer protocol, the file data uploaded by the user is parsed into a binary stream to obtain file metadata;

[0122] B2: Input the file metadata into the preprocessing unit for processing to obtain preprocessed file data; the specific process is as follows:

[0123] According to the file type, the file metadata is preprocessed to obtain the preprocessed file data;

[0124] The file types include scanned PDF and image file types, editable PDF and Office file types;

[0125] For scanned PDFs and image files, an image preprocessing workflow is initiated.

[0126] First, an adaptive binarization algorithm is used to perform image grayscale conversion and noise removal on scanned PDF and image files to obtain scanned PDF and image files after noise removal.

[0127] Then, by combining perspective transformation, tilt correction image processing is performed on scanned PDF and image files to optimize character edge contours, resulting in preprocessed scanned PDF and image files;

[0128] For editable PDFs and Office files, initiate the editable data preprocessing workflow.

[0129] First, the raw data of elements such as text, tables, and images is extracted by parsing the internal data stream structure of the file.

[0130] Then, redundant format information in the original data file is removed, and a standardized intermediate data format is generated to obtain pre-processed editable PDF and Office files, laying the foundation for subsequent content extraction.

[0131] B3: Input the preprocessed file data into the content extraction unit for processing to obtain the extracted content.

[0132] The extracted content includes: text extraction, table extraction, and special symbol extraction.

[0133] The specific process by which the preprocessed file data in B3 is input into the content extraction unit for processing to obtain the extracted text content is as follows:

[0134] When the preprocessed file data consists of preprocessed scanned PDF documents and image files...

[0135] An OCR engine based on an LSTM neural network model is used to extract text from preprocessed scanned PDF and image files to obtain the extracted text content.

[0136] When the preprocessed file data is a preprocessed editable PDF or Office file, the corresponding method is used to obtain the first extraction result.

[0137] The Adobe Acrobat SDK is used to parse the original file streams of editable PDFs and Office files, which are then used as the second extraction result.

[0138] The first extraction result and the second extraction result are fused together using a linear weighted fusion algorithm to obtain the final file extraction result, thereby improving the accuracy of file extraction.

[0139] The specific process by which the preprocessed file data in B3 is input into the content extraction unit for processing to obtain the table content is as follows:

[0140] For table extraction, a deep learning-based table detection model is used to perform table detection processing on the document image to obtain the table regions in the document. This table detection processing employs OCR image recognition, with the YOLO series of models being the preferred OCR image recognition algorithm. YOLO is an optional implementation scheme for one step in the OCR process.

[0141] First, locate the table area in the file; then, identify the table's row and column structure using a cell segmentation algorithm; finally, extract the file data within the cells and establish row and column relationships to obtain the extracted table content.

[0142] The specific process by which the preprocessed file data in B3 is input into the content extraction unit for processing to obtain the extracted special symbols is as follows:

[0143] The special symbol extraction engine is a feature-matching-based symbol recognition engine that uses a pre-trained symbol feature library to classify and locate target symbols.

[0144] Preferably, the present invention performs special symbol extraction processing on document images to obtain the category and location information of special symbols; the special symbol extraction processing adopts the ORB feature matching algorithm, and realizes the classification and positioning of target symbols through a pre-trained symbol feature library.

[0145] B4: Input the extracted content obtained in B3 into the data structuring unit for processing to obtain the file filling result; the specific process is as follows:

[0146] The data structuring unit achieves standardized transformation of extracted data based on a pre-built field mapping template library, a rule engine, and semantic mapping technology.

[0147] B4.1: Load the configuration file from the preset field mapping template library that matches the extracted content obtained in B3.

[0148] The configuration file defines the mapping relationship between source data fields and target form fields;

[0149] The configuration file is in XML or JSON format.

[0150] B4.2: The Conditional Random Field (CRF) algorithm is used to perform key entity recognition processing on the text information to obtain key entity information in the text.

[0151] The key entities are formatted using a regular expression engine and a pre-built dictionary to obtain standardized key entities.

[0152] B4.3: For the standardized key entities, the Naive Bayes algorithm is used to classify them into the corresponding form fields to obtain the document filling results.

[0153] The final generated file entry result is structured data conforming to the JSON specification, which is synchronously transmitted to the form data storage module via API to complete the automated processing of the entire file parsing process.

[0154] Preferably, the text classification algorithm of the present invention is the TextCNN algorithm; other steps and parameters are the same as in one of the specific embodiments one to six.

[0155] Specific Implementation Method Eight: This implementation method is a computer storage medium that stores at least one instruction. The at least one instruction is loaded and executed by a processor to implement the intelligent data entry method based on natural language processing.

[0156] It should be understood that the instructions include computer program products, software, or computerized methods corresponding to any method described in this invention; the instructions can be used to program computer systems or other electronic devices. Computer storage media may include readable media on which instructions are stored, and may include, but are not limited to, magnetic storage media, optical storage media; magneto-optical storage media include read-only memory (ROM), random access memory (RAM), erasable programmable memory (e.g., EPROM and EEPROM), and flash memory layers, or other types of media suitable for storing electronic instructions.

[0157] Specific Implementation Method Nine: This implementation method is an intelligent data entry device based on natural language processing. The device includes a processor and a memory. It should be understood that it includes any device including a processor and a memory described in this invention. The device may also include other units and modules that perform display, interaction, processing, control and other functions through signals or instructions.

[0158] The memory stores at least one instruction, which is loaded and executed by the processor to implement the intelligent data entry method based on natural language processing.

[0159] Specific Implementation Method 10: This implementation method is a multi-terminal adapted intelligent data entry service system. The system is characterized by its compatibility with devices such as computers, mobile devices, and government self-service terminals. It executes the intelligent data entry method based on natural language processing described in any one of Specific Implementation Methods 1 to 7, receives data entry requests from each terminal, and provides feedback on the data entry results. The system includes:

[0160] The task receiving module is used to receive user-triggered data entry tasks.

[0161] The mode determination module is used to determine the filling mode from a variety of preset filling modes based on the type of the filling task; the filling modes include: voice interaction filling mode, file interpretation filling mode, sequential filling mode and individual filling mode;

[0162] The information acquisition module is used to acquire the original information entered by the user in the filling mode;

[0163] The intelligent processing module is used to process the raw information entered into the intelligent data entry model based on natural language processing to obtain the data entry result.

[0164] The above description is merely of preferred embodiments of the present invention. It should be understood that the present invention is not limited to the specific embodiments described above. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the disclosed technical content without departing from the scope of the present invention, forming equivalent embodiments with equivalent changes. Any simple modifications, equivalent substitutions, and improvements made to the above embodiments without departing from the scope of the present invention, based on the technical essence of the present invention and within the spirit and principles of the present invention, shall still fall within the protection scope of the present invention.

Claims

1. An intelligent data entry method based on natural language processing, characterized in that, include: Step 1: Receive the form filling task triggered by the user. Step 2: Based on the type of data entry task described in Step 1, determine the data entry mode from a variety of preset data entry modes; The filling modes include: voice interaction filling mode, file interpretation filling mode, sequential filling mode, and individual filling mode; Step 3: Obtain the original information entered by the user in the filling mode; Step 4: Input the original information into the intelligent data entry model based on natural language processing for processing to obtain the data entry result.

2. The intelligent data entry method based on natural language processing according to claim 1, characterized in that, The original information entered by the user in step three includes: voice data entered by the user and file data uploaded by the user. The intelligent data entry model based on natural language processing includes: a voice interaction data entry module and a document interpretation data entry module.

3. The intelligent data entry method based on natural language processing according to claim 2, characterized in that, The specific process for obtaining the original information entered by the user in step three is as follows: When the user inputs voice data, the system retrieves the original information entered by the user from the mobile / computer device. The specific process is as follows: A1: Establish a connection with the microphone on the mobile / computer device through the preset device interface protocol and call the microphone on the mobile / computer device; A2: The microphone collects user-input voice data in real time according to device parameters of 16Hz sampling rate and 16-bit depth.

4. The intelligent data entry method based on natural language processing according to claim 3, characterized in that, In step four, the original information is input into a natural language processing-based intelligent data entry model for processing to obtain the data entry result. The specific process is as follows: When the original information to be filled in is voice data input by the user, the voice data input by the user is input into the voice interaction filling module for processing to obtain the voice filling result; When the original information is file data uploaded by the user, the file data uploaded by the user is input into the file interpretation and filling module to obtain the file filling result.

5. The intelligent data entry method based on natural language processing according to claim 4, characterized in that, The specific process of inputting user-supplied voice data into the voice interaction filling module for processing to obtain the voice filling result is as follows: S1: Preprocess the user-input voice data to obtain preprocessed voice data; S2: According to the preset segmentation duration, the preprocessed speech data is segmented to obtain frame sequence data; S3: Use the Mel Frequency Cepstral Coefficient (MFCC) algorithm to extract speech features from the frame sequence data to obtain speech features; S4: Use an automatic speech recognition algorithm to identify and process speech features to obtain the text sequence corresponding to the speech; S5: Perform syntax verification and error correction on the text sequence corresponding to the speech to obtain the text sequence after syntax verification and error correction. S6: Based on the field format requirements of the form, perform structured processing on the text sequence after syntax validation and error correction to obtain the voice entry result.

6. The intelligent data entry method based on natural language processing according to claim 5, In step S1, the user-input voice data is preprocessed to obtain preprocessed voice data; the specific process is as follows: S1.1: The user-input voice data is initially filtered using an adaptive noise suppression algorithm to obtain the initially filtered voice data; S1.2: The speech data after the initial filtering is processed by the echo cancellation algorithm to obtain the speech data after secondary filtering. S1.3: The signal gain of the secondary filtered speech data is adjusted and the amplitude is normalized by the root mean square normalization algorithm to obtain the preprocessed speech data.

7. The intelligent data entry method based on natural language processing according to claim 6, characterized in that, The file interpretation and reporting module includes: a file parsing unit, a preprocessing unit, a content extraction unit, and a data structuring unit; The process involves inputting the user-uploaded file data into the file interpretation and reporting module to obtain the file reporting results; the specific process is as follows: B1: The user-uploaded file data is input into the file parsing unit for processing to obtain file metadata; the specific process is as follows: Based on a standardized file transfer protocol, the file data uploaded by the user is parsed into a binary stream to obtain file metadata; B2: Input the file metadata into the preprocessing unit for processing to obtain preprocessed file data; B3: Input the preprocessed file data into the content extraction unit for processing to obtain the extracted content. The extracted content includes: text content, table content, and special character content; B4: Input the extracted content obtained in B3 into the data structuring unit for processing to obtain the file filling result; the specific process is as follows: B4.1: Load the configuration file from the preset field mapping template library that matches the extracted content obtained in B3. B4.2: The Conditional Random Field (CRF) algorithm is used to perform key entity recognition processing on the text information to obtain key entity information in the text. The key entities are formatted using a regular expression engine and a pre-built dictionary to obtain standardized key entities. B4.3: For the standardized key entities, the Naive Bayes algorithm is used to classify them into the corresponding form fields to obtain the document filling results.

8. A computer storage medium, characterized in that, The storage medium stores at least one instruction, which is loaded and executed by a processor to implement the intelligent data entry method based on natural language processing as described in any one of claims 1 to 7.

9. An intelligent data entry device based on natural language processing, characterized in that, The device includes a processor and a memory, the memory storing at least one instruction, which is loaded and executed by the processor to implement a natural language processing-based intelligent data entry method as described in any one of claims 1 to 7.

10. A multi-terminal adapted intelligent data entry service system, characterized in that, The system is compatible with access from devices such as computers, mobile devices, and government self-service terminals. It executes the intelligent data entry method based on natural language processing as described in any one of claims 1 to 7, receives data entry requests from various terminals, and provides feedback on the data entry results. The system includes: The task receiving module is used to receive user-triggered data entry tasks. The mode determination module is used to determine the filling mode from a variety of preset filling modes based on the type of the filling task; the filling modes include: voice interaction filling mode, file interpretation filling mode, sequential filling mode and individual filling mode; The information acquisition module is used to acquire the original information entered by the user in the filling mode; The intelligent processing module is used to process the raw information entered into the intelligent data entry model based on natural language processing to obtain the data entry result.

Images