Document processing method, system and device, storage medium and product

By using a large language model to analyze documents and fill in form content, combined with the verification mechanism, the problem of time-consuming and low efficiency of filling in form in the existing technology is solved, and an efficient and accurate document processing process is achieved.

CN118982008APending Publication Date: 2024-11-19SHANGHAI INVESTIGATION DESIGN & RES INST CO LTD
View PDF 0 Cites 4 Cited by

Patent Information

Application Number
CN202411040944.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-31
Publication Date
2024-11-19

AI Technical Summary

Technical Problem

In the existing document processing technology, the filling of forms relies on manual reading and data entry, which is time-consuming and error-prone. In the professional field extraction tools based on mechanical learning, require a large amount of data labeled, resulting in inefficient processing.

Method used

The first language model is used to analyze and extract the key data of the document, store it in the table library, the second language model extracts the table content and fills it in the preset form, combines the third language model to verify the data accuracy and realize automation Processing process, using weakly supervised learning for training.

Benefits of technology

It significantly improves the accuracy and efficiency of document processing, reaches a success rate of 95%, reduces training time costs, and improves the reliability of data processing.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118982008A_ABST
    Figure CN118982008A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of document processing, and discloses a document processing method, system and device, a storage medium and a product, and the method comprises the steps: carrying out the document analysis of a received document based on a first large language model, extracting key data, and storing the key data in a preset table library; and extracting table content of the preset table based on the second large language model, retrieving a matching item from the table library according to the table content, and filling the matching item into the preset table. According to the method, the first large language model and the second large language model are adopted to achieve document extraction and table filling respectively, an excellent filling effect can be achieved only through training of weak supervised learning, the time cost needed by training is greatly reduced, the success rate can reach up to 95%, and the training efficiency is improved. And the accuracy and efficiency of document processing are obviously improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of document processing, and in particular to a document processing method, system, device, storage medium and product. Background Art

[0002] In the current document processing field, document reading, analysis, content extraction, and form filling are usually tedious and time-consuming tasks.

[0003] At present, traditional document processing technology and form filling usually rely on manual reading and data entry, which is not only time-consuming but also prone to errors. Large language models such as BERT and transformer architecture can understand the deep semantics of language through pre-training and fine-tuning, making it possible to automatically read and analyze document content. However, there are currently no extraction tools developed based on artificial intelligence on the market. Most of them are analyzed by AI plus manual extraction, or are based on machine learning for professional fields and form filling. Extraction tools based on machine learning for professional fields often require a large amount of labeled data and manual intervention, and document processing efficiency is low. Summary of the invention

[0004] In view of this, the present invention provides a document processing method, system, device, storage medium and product to solve the technical problem of low processing efficiency of existing document processing methods.

[0005] In a first aspect, the present invention provides a document processing method, comprising: parsing a received document based on a first language model, extracting key data and storing the key data in a pre-set table library; extracting table content of a preset table based on a second language model, retrieving matching items from the table library according to the table content, and filling the matching items into the preset table.

[0006] A document processing method of the present invention parses a received document based on a first large language model, extracts key data and stores the key data in a preset table library, extracts table content of a preset table based on a second large language model, retrieves matching items from the table library according to the table content, and fills the matching items into the preset table, adopts a large language model to realize automatic reading and analysis of document content, and fills the extracted information into the preset table, utilizes the question-answering reading ability, automatic reading ability and supervised learning technology of the large language model, realizes an end-to-end automated processing flow, and compared with the traditional supervised learning method based on machine learning with manual error correction, feedback and parameter adjustment, the present invention adopts the first large language model and the second large language model to realize document extraction and table filling respectively, and only needs to be trained through weak supervised learning to achieve a relatively excellent filling effect, which not only greatly reduces the time cost required for training, but also can achieve a success rate of up to 95%, significantly improving the accuracy and efficiency of document processing.

[0007] Optionally, after filling the matching items into the preset form, it also includes: verifying the matching items filled in the preset form based on the third largest language model, if the verification fails, returning the failed matching items and the corresponding table content to the first largest language model, re-parsing the document and extracting key data through the first largest language model combined with the returned matching items and the corresponding table content, updating the data in the form library, and re-filling the preset form through the second largest language model.

[0008] In this method, the filled-in content is verified through the third language model to ensure the accuracy of the data filled in the preset form. If the verification fails, the data is automatically re-parsed and updated, further improving the reliability of data processing.

[0009] Optionally, the matching items filled in the preset form are verified based on a third language model, including: based on the third language model, using a preset data verification algorithm and a preset machine learning model to verify the matching items filled in the preset form.

[0010] By using preset data verification algorithms and preset machine learning models to comprehensively verify the filled-in data, the comprehensiveness and accuracy of the verification are enhanced.

[0011] Optionally, the preset data verification algorithm includes a regular expression matching algorithm, a range checking algorithm and a logical dependency analysis algorithm, and the preset machine learning model includes a classifier and an anomaly detection model.

[0012] Optionally, the received document is parsed based on the first language model, key data is extracted and the key data is stored in a pre-set table library, including: based on the first language model, the received document is parsed by natural language processing technology to obtain parsed content; based on the parsed content, the key data is retrieved and extracted from an external knowledge base using a retrieval enhancement generation mechanism, and the key data is stored in a pre-set table library.

[0013] By combining natural language processing technology and retrieval enhancement generation mechanisms, we can more deeply understand and analyze document content, extract more accurate and comprehensive key data from documents, and use external knowledge bases to enhance the relevance and richness of data.

[0014] Optionally, before extracting the table content of the preset table based on the second largest language model, retrieving matching items from the table library according to the table content, and filling the matching items into the preset table, it also includes: extracting the table in the received document based on the first largest language model to obtain the preset table, or reading the table created by the user in the target interface to obtain the preset table.

[0015] By identifying the table in the document uploaded by the user, or extracting the table directly created by the user on the target interface, a preset table is obtained to provide a filling target for subsequent table content filling.

[0016] In a second aspect, the present invention provides a document processing system, comprising: a first large language model, used to parse a received document, extract key data and store the key data in a preset table library; a second large language model, used to extract table content of a preset table, retrieve matching items from the table library according to the table content, and fill the matching items into the preset table.

[0017] In a third aspect, the present invention provides a computer device, comprising: a memory and a processor, the memory and the processor are communicatively connected to each other, computer instructions are stored in the memory, and the processor executes the document processing method of the first aspect or any corresponding embodiment thereof by executing the computer instructions.

[0018] In a fourth aspect, the present invention provides a computer-readable storage medium having computer instructions stored thereon, the computer instructions being used to enable a computer to execute the document processing method of the first aspect or any corresponding embodiment thereof.

[0019] In a fifth aspect, the present invention provides a computer program product, comprising computer instructions, wherein the computer instructions are used to enable a computer to execute the document processing method of the first aspect or any corresponding embodiment thereof. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0021] Figure 1 is a flowchart of a document processing method according to an embodiment of the present invention;

[0022] Figure 2 is a flowchart of another document processing method according to an embodiment of the present invention;

[0023] Figure 3 is a structural block diagram of a document processing system according to an embodiment of the present invention;

[0024] Figure 4 It is a schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0025] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0026] Filling out forms is cumbersome and time-consuming, and has a high labor cost. Currently, there are no extraction tools developed based on large language models on the market. Most of them are AI-based analysis plus manual extraction, or extraction tools based on machine learning for professional fields, which are not universal. Based on this, an embodiment of the present invention provides a document processing method that can perform relatively general document content extraction and form filling based on a small amount of training, so as to automatically fill out preset forms and improve document processing efficiency.

[0027] According to an embodiment of the present invention, a document processing method embodiment is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0028] In this embodiment, a document processing method is provided, which can be used in mobile terminals, such as mobile phones, tablet computers, etc. Figure 1 As shown, the process includes the following steps:

[0029] Step S101 , parsing the received document based on the first language model, extracting key data and storing the key data in a pre-set table library.

[0030] Specifically, the first Large Language Model (LLM) is a large language model for text analysis that is fine-tuned by self-training, similar to Baoyue on the market. Baoyue is a domestic document analysis tool that can upload documents for real-time analysis. It supports various file types, such as PDF, doc, docx, etc. By selecting document upload, you can quickly read papers, books, documents, and web pages, and can quickly analyze files.

[0031] The first language model receives documents uploaded by users, analyzes the documents, reads key data therein, and enters the key data into a table library in a structured manner.

[0032] The table library essentially serves as a data warehouse for subsequent processing, supporting efficient storage and retrieval of multiple file formats, such as PDF, Doc, CSV, Excel, etc.

[0033] Step S102: extracting the table content of the preset table based on the second largest language model, retrieving matching items from the table library according to the table content, and filling the matching items into the preset table.

[0034] Specifically, the preset table can be obtained by extracting a table in the received document based on the first language model to obtain the preset table, or by reading a table created by the user in the target interface to obtain the preset table.

[0035] In one example, a deep learning model, in particular a combination of a convolutional neural network and a recurrent neural network, is used to automatically identify the table structure of a preset table in a document.

[0036] The second language model needs to have strong data processing and filling capabilities. It is a table reading and input model that is trained and fine-tuned by itself, similar to the WPS AI currently on the market. WPS AI can analyze, disassemble and re-fill the content in the table, and fill in the data according to user needs.

[0037] The second largest language model extracts table content through a table recognition algorithm, accesses data in the table library, and automatically retrieves matching items from the database based on the extracted table content according to preset logic or user instructions, and accurately fills them into the corresponding position of the preset table on the user's screen.

[0038] For example, if the user specifies the "product name" and "inventory quantity" that need to be filled in the preset form, the second largest language model can retrieve the product names and current inventory quantities of these products from the database and fill them in automatically.

[0039] In one example, the second largest language model uses a pre-trained LLM model, combined with fine-tuning in specific fields to improve the understanding and processing capabilities of table data, and customizes logical processing, defining data filling rules by writing scripts or using rule engines (such as Drools). The second largest language model uses API calls or database query statements to retrieve matching data from the table library, applies natural language understanding technology, parses user instructions and table content, and maps to corresponding data filling operations.

[0040] A document processing method according to an embodiment of the present invention parses a received document based on a first large language model, extracts key data and stores the key data in a preset table library, extracts table content of a preset table based on a second large language model, retrieves matching items from the table library according to the table content, and fills the matching items into the preset table, adopts a large language model to realize automatic reading and analysis of the document content, and fills the extracted information into the preset table, utilizes the question-answering reading ability, automatic reading ability and supervised learning technology of the large language model, realizes an end-to-end automated processing flow, and compared with the traditional supervised learning method based on machine learning with manual error correction, feedback and parameter adjustment, the present invention adopts the first large language model and the second large language model to realize document extraction and table filling respectively, and only needs to be trained through weak supervised learning to achieve a relatively excellent filling effect, which not only greatly reduces the time cost required for training, but also can achieve a success rate of up to 95%, significantly improving the accuracy and efficiency of document processing.

[0041] In one embodiment, after filling the matching items into the preset table in step S102, the method further includes:

[0042] Step S103, based on the third largest language model, the matching items filled in the preset table are verified. If the verification fails, the failed matching items and the corresponding table contents are returned to the first largest language model. The document is re-parsed and key data is extracted through the first largest language model combined with the returned matching items and the corresponding table contents, the data in the table library is updated, and the preset form is re-filled through the second largest language model.

[0043] Specifically, based on the third language model, a preset data verification algorithm and a preset machine learning model are used to verify the matching items filled in the preset form.

[0044] The preset data verification algorithms include regular expression matching algorithm, range checking algorithm and logical dependency analysis algorithm, and the preset machine learning models include classifier and anomaly detection model. By using a variety of preset data verification algorithms and preset machine learning models to conduct comprehensive verification of the filled data, the comprehensiveness and accuracy of the verification are enhanced.

[0045] The third largest language model is a supervised model that is trained and fine-tuned by itself, similar to iFlytek Intelligent Inspection developed by iFlytek on the market. iFlytek Intelligent Inspection is a product that helps users perform AI intelligent text error correction. It supports batch review of plain text, Word, pictures, audio, and video. While saving the cost of manual review, it greatly improves the overall review efficiency. Compared with the traditional reinforcement learning algorithm based on human feedback information (Reinforcement Learning from Human Feedback, RLHF), the third largest language model has the ability of self-reinforcement learning to control the performance results.

[0046] The data validation phase of the third language model plays the role of quality gatekeeper. The third language model uses complex validation algorithms, including but not limited to data consistency checks, range validation (such as inventory should not be negative), and possible logical dependency checks.

[0047] For example, if the inventory of a product is detected to be abnormally high or low, beyond the normal range, or there is a logical contradiction with other data (such as sales far exceeding inventory), the third language model will mark these data as potential problems and may trigger a re-examination process. This process will notify the first language model to re-search and extract specific data points, and the updated data will be filled in by the second language model M again, and finally reviewed by the third language model until all data is verified or the user actively interrupts the process to obtain the current processing results. Through such a circular verification mechanism, not only the accuracy and completeness of the data are ensured, but also the processing efficiency and user experience are greatly improved.

[0048] In one example, a feedback loop is designed to trigger a review process when the third language model detects data problems. The verification process is managed by a workflow engine (such as Activiti or Camunda) to ensure data consistency and accuracy.

[0049] like Figure 2As shown, the second language model transmits the generated results to the third language model for data verification. The problematic data or data missing are marked and returned to the first language model for re-retrieval and extraction, and returned to the second language model for filling, and then verified by the third language model. The cycle will continue until the third language model confirms that it is correct, or the user interrupts to obtain the final output.

[0050] In this method, the filled-in content is verified through the third language model to ensure the accuracy of the data filled in the preset form. If the verification fails, the data is automatically re-parsed and updated, further improving the reliability of data processing.

[0051] It should be understood that the first language model, the second language model, and the third language model will select the best option from the current open source market based on their functionality. After the three major language models are combined, end-to-end document input and form input output can be achieved, greatly improving the work efficiency of business personnel.

[0052] According to a document processing method of an embodiment of the present invention, step S101 parses a received document based on a first language model, extracts key data and stores the key data in a pre-set table library, including:

[0053] Step S1012, based on the first language model, the received document is parsed by natural language processing technology to obtain parsed content;

[0054] Step S1013: According to the parsed content, the key data is retrieved and extracted from the external knowledge base using the retrieval enhancement generation mechanism, and the key data is stored in a pre-set table library.

[0055] Specifically, a cloud-based service architecture is adopted, and users upload documents through a web interface or an API interface. At the same time, asynchronous file transfer technology is used to ensure that users can continue to perform other operations during the document upload process, thereby improving the user experience.

[0056] After receiving the document, natural language processing technology, such as optical character recognition technology and document structure analysis technology, is used to parse the uploaded document to obtain the parsed content.

[0057] Combining the first language model with an external knowledge base, relevant information is retrieved from the knowledge base through a retrieval enhancement generation mechanism. This process involves vector search and semantic matching algorithms to ensure that the retrieved information is highly relevant to the document content.

[0058] The Retrieval Augmented Generation (RAG) mechanism enhances the performance of the large language model by integrating external knowledge of the enterprise into the generation process. As a cutting-edge technology, RAG greatly enhances the generation capability of the first large language model by introducing external knowledge of the enterprise. In this process, the first large language model not only plays the role of a document parser, but also makes full use of the RAG mechanism to retrieve relevant information from the knowledge base, conduct in-depth analysis of the document, and accurately extract key data.

[0059] In one example, a pre-set table library uses a NoSQL database (such as MongoDB) to store structured data, supports flexible data models, and after extracting key data, converts the extracted data into a unified format and loads it into the table library.

[0060] By combining natural language processing technology and retrieval enhancement generation mechanisms, we can more deeply understand and analyze document content, extract more accurate and comprehensive key data from documents, and use external knowledge bases to enhance the relevance and richness of data.

[0061] A document processing method according to an embodiment of the present invention has the following advantages:

[0062] 1. A large language model is used to automatically read and analyze the document content, and the extracted information is filled into the preset form. The question-and-answer reading ability, automatic reading ability and supervised learning technology of the large language model are used to realize an end-to-end automated processing flow. Compared with the traditional supervised learning method based on machine learning, manual error correction, feedback and parameter adjustment, the present invention adopts the first large language model and the second large language model to realize document extraction and form filling respectively. Only weak supervised learning is needed for training to achieve a relatively good filling effect. It not only greatly reduces the time cost required for training, but also can achieve a success rate of up to 95%, significantly improving the accuracy and efficiency of document processing.

[0063] 2. The filled-in content is verified through the third language model to ensure the accuracy of the data filled in the preset form. If the verification fails, the data is automatically re-parsed and updated, further improving the reliability of data processing.

[0064] 3. By combining natural language processing technology and retrieval enhancement generation mechanism, we can more deeply understand and analyze document content, extract more accurate and comprehensive key data from documents, and use external knowledge bases to enhance the relevance and richness of data.

[0065] 4. The embodiment of the present invention solves the problem of business personnel repeatedly filling out the same form, and helps business personnel to handle the complicated and repetitive form filling problems more quickly and conveniently. At the same time, the collected data will continue to train the first language model and the second language model, and relatively general extraction can be performed based on a small amount of training to improve accuracy. The archiving of data through the preset form library also solves the warehousing problem of data governance, and improves the user's information work more quickly and effectively.

[0066] The embodiment of the present invention also provides a document processing system, such as Figure 3 As shown, including:

[0067] The first language model is used to parse the received documents, extract key data and store the key data in a pre-set table library;

[0068] The second largest language model is used to extract the table content of the preset table, retrieve matching items from the table library according to the table content, and fill the matching items into the preset table.

[0069] The document processing system of the embodiment of the present invention parses the received document based on the first large language model, extracts key data and stores the key data in a preset table library, extracts the table content of the preset table based on the second large language model, retrieves matching items from the table library according to the table content, and fills the matching items into the preset table, adopts the large language model to realize automatic reading and analysis of the document content, and fills the extracted information into the preset table, and utilizes the question-answering reading ability, automatic reading ability and supervised learning technology of the large language model to realize end-to-end automated processing flow. Compared with the traditional supervised learning method based on manual error correction feedback and parameter adjustment based on machine learning, the present invention adopts the first large language model and the second large language model to realize document extraction and table filling respectively, and only needs to be trained through weak supervised learning to achieve a relatively good filling effect, which not only greatly reduces the time cost required for training, but also can achieve a success rate of up to 95%, significantly improving the accuracy and efficiency of document processing.

[0070] Furthermore, the document processing system also includes:

[0071] The third language model is used to verify the matching items filled in the preset table. If the verification fails, the failed matching items and the corresponding table content are returned to the first language model. The first language model is combined with the returned matching items and the corresponding table content to re-parse the document and extract key data, update the data in the table library, and re-fill the preset table through the second language model. Specifically, for information that is not in the database, the document will be re-extracted by the first language model first, and then filled in by the second language model.

[0072] Furthermore, the third language model is also used to verify the matching items filled in the preset form based on the third language model using a preset data verification algorithm and a preset machine learning model.

[0073] Furthermore, the preset data verification algorithms include a regular expression matching algorithm, a range checking algorithm, and a logical dependency analysis algorithm, and the preset machine learning models include a classifier and an anomaly detection model.

[0074] Furthermore, the first language model is also used to parse the received document through natural language processing technology to obtain parsed content; based on the parsed content, the retrieval enhancement generation mechanism is used to retrieve and extract key data from the external knowledge base, and the key data is stored in a pre-set table library.

[0075] Furthermore, the first largest language model is also used to extract a table in the received document based on the first largest language model to obtain a preset table, or to read a table created by a user in the target interface to obtain a preset table.

[0076] The embodiment of the present invention also provides a schematic diagram of the structure of a computer device, such as Figure 4 As shown, the computer device includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. Various components are connected to each other using different buses for communication, and can be installed on a common mainboard or installed in other ways as needed. The processor can process the instructions executed in the computer device, including instructions stored in or on the memory to display the graphical information of the GUI on an external input / output device (such as, a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Similarly, multiple computer devices can be connected, and each device provides some necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Figure 4 A processor 10 is taken as an example.

[0077] The processor 10 may be a central processing unit, a network processor or a combination thereof. The processor 10 may further include a hardware chip. The hardware chip may be a dedicated integrated circuit, a programmable logic device or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic or any combination thereof.

[0078] The memory 20 stores instructions executable by at least one processor 10, so that at least one processor 10 executes the method shown in the above embodiment.

[0079] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created according to the use of the computer device, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some optional embodiments, the memory 20 may optionally include a memory remotely arranged relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0080] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk or a solid state drive; the memory 20 may also include a combination of the above types of memory.

[0081] The computer device also includes an input device 30 and an output device 40. The processor 10, the memory 20, the input device 30 and the output device 40 may be connected via a bus or other means. Figure 4 The example of connecting through bus is taken in the following.

[0082] The input device 30 can receive input digital or character information, and generate key signal input related to the user settings and function control of the computer device, such as a touch screen, a keypad, a mouse, a track pad, a touch pad, an indicator bar, one or more mouse buttons, a trackball, a joystick, etc. The output device 40 may include a display device, an auxiliary lighting device (e.g., an LED) and a tactile feedback device (e.g., a vibration motor), etc. The above-mentioned display device includes but is not limited to a liquid crystal display, a light emitting diode, a display and a plasma display. In some optional embodiments, the display device can be a touch screen.

[0083] The embodiment of the present invention also provides a computer-readable storage medium. The method according to the embodiment of the present invention can be implemented in hardware, firmware, or can be implemented as a computer code that can be recorded in a storage medium, or can be implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and will be stored in a local storage medium through a network download, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only storage memory, a random access memory, a flash memory, a hard disk or a solid-state hard disk, etc.; further, the storage medium can also include a combination of the above types of memories. It can be understood that a computer, a processor, a microprocessor controller, or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor, or hardware, the method shown in the above embodiment is implemented.

[0084] A part of the present invention may be applied as a computer program product, such as a computer program instruction, which, when executed by a computer, can call or provide the method and / or technical solution according to the present invention through the operation of the computer. Those skilled in the art should understand that the existence of the computer program instruction in a computer-readable medium includes, but is not limited to, a source file, an executable file, an installation package file, etc., and accordingly, the way in which the computer program instruction is executed by the computer includes, but is not limited to: the computer directly executes the instruction, or the computer compiles the instruction and then executes the corresponding compiled program, or the computer reads and executes the instruction, or the computer reads and installs the instruction and then executes the corresponding installed program. Here, the computer-readable medium may be any available computer-readable storage medium or communication medium accessible to the computer.

[0085] Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention, and such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A document processing method, characterized in that: include: Parsing the received document based on the first language model, extracting key data and storing the key data in a pre-set table library; The table content of the preset table is extracted based on the second largest language model, matching items are retrieved from the table library according to the table content, and the matching items are filled into the preset table.

2. The document processing method according to claim 1, characterized in that: After filling the matching items into the preset table, the method further includes: The matching items filled in the preset form are verified based on the third largest language model. If the verification fails, the failed matching items and the corresponding table contents are returned to the first largest language model. The document is re-parsed and key data is extracted through the first largest language model combined with the returned matching items and the corresponding table contents, the data in the form library is updated, and the preset form is re-filled through the second largest language model.

3. The document processing method according to claim 2, characterized in that: Verifying the matching items filled in the preset table based on the third language model includes: Based on the third language model, a preset data verification algorithm and a preset machine learning model are used to verify the matching items filled in the preset form.

4. The document processing method according to claim 3, characterized in that: The preset data verification algorithm includes a regular expression matching algorithm, a range checking algorithm and a logical dependency analysis algorithm, and the preset machine learning model includes a classifier and an anomaly detection model.

5. The document processing method according to claim 1, characterized in that: Parsing the received document based on the first language model, extracting key data and storing the key data in a pre-set table library, including: Based on the first language model, the received document is parsed by natural language processing technology to obtain parsed content; According to the parsed content, the key data is retrieved and extracted from the external knowledge base using the retrieval enhancement generation mechanism, and the key data is stored in a pre-set table library.

6. The document processing method according to claim 1, characterized in that: Before extracting the table content of the preset table based on the second largest language model, retrieving matching items from the table library according to the table content, and filling the matching items into the preset table, the method further includes: A table in the received document is extracted based on the first language model to obtain a preset table, or a table created by a user in the target interface is read to obtain a preset table.

7. A document processing system, characterized in that: include: The first language model is used to parse the received document, extract key data and store the key data in a pre-set table library; The second language model is used to extract the table content of the preset table, retrieve matching items from the table library according to the table content, and fill the matching items into the preset table.

8. A computer device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the document processing method according to any one of claims 1 to 6 by executing the computer instructions.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the document processing method according to any one of claims 1 to 6.

10. A computer program product, characterized in that The method comprises computer instructions for causing a computer to execute the document processing method according to any one of claims 1 to 6.

Citation Information

Cited By

  • Battery performance data automatic extraction system based on large model

    CN119849618A

  • Table data extraction method and device, computer equipment and storage medium

    CN120030162A

  • Table data extraction method, device, computer equipment and storage medium

    CN120030162B

  • Component cataloguing method and management method, program product and electronic device

    CN120950114A