Interactive document structured information extraction method and device, electronic equipment and storage medium

Abstract

The invention discloses an interactive document structured information extraction method and device, electronic equipment and a storage medium, and particularly relates to the field of document processing. The interactive document structured information extraction method comprises the steps that a document is recognized, and a document format is obtained; matching the document format with a sample format stored in a knowledge base, selecting a matched format, and executing a corresponding task instruction to extract document structured information; the user performs error correction on the extraction result; outputting an accurate result and verifying the consistency of the extraction result and the accurate result; querying a similar task instruction in a knowledge base according to the accurate result, and generating a new task instruction; and executing the new task instruction on the document format to obtain a test result, and verifying the consistency of the test result and the accurate result. According to the method, dynamic learning and optimization can be realized, the accuracy and robustness of information extraction are remarkably improved, the method has strong instant learning ability and adaptivity, and the learning and training cost is reduced.

Description

Technical Field

[0001] The present invention relates to the field of document processing and information extraction, and in particular to an interactive document structured information extraction method, device, electronic device and storage medium. Background Art

[0002] In recent years, with the rapid development of artificial intelligence and natural language processing technology, document information extraction technology has received widespread attention and application. Currently, there are many products on the market for structured extraction of document information. These products mainly rely on machine learning and deep learning algorithms, combined with big data analysis technology, to extract valuable information from large amounts of unstructured data.

[0003] However, existing document information extraction systems often have difficulty maintaining high accuracy when dealing with complex and diverse documents. In particular, for data in non-standardized formats and containing noise, the performance of existing systems is often unsatisfactory. In addition, when faced with new domains or new types of documents, these systems may need to be retrained or adjusted on a large scale, resulting in insufficient robustness in practical applications. Moreover, most current document information extraction systems rely on a large amount of annotated data for training. This data dependence not only increases development and maintenance costs, but also limits the application of the system in data-scarce areas. Existing technologies and products are time-consuming and laborious to obtain and annotate high-quality data, especially for areas that require professional knowledge, and lack real-time learning and adaptive capabilities. Even some deep learning-based models are often regarded as "black boxes", and it is difficult to explain their decision-making process and the basis of the extraction results, which may be problematic in some scenarios with high requirements for interpretability.

[0004] Therefore, how to improve the instant learning ability of the document information extraction system and how to reduce the learning and training cost of the document information extraction system are urgent problems to be solved in the existing document information extraction system. Summary of the invention

[0005] The present invention provides an interactive document structured information extraction method, device, electronic device and storage medium, which are used to improve the instant learning ability of the existing document information extraction system, reduce training costs, and improve the accuracy and robustness of information extraction.

[0006] In order to achieve the above object, the present invention provides the following technical solutions: The method for extracting structured information from an interactive document comprises the following steps: S1: Identify the document, obtain the document format, and execute S2; S2: Match the document layout with the sample layout stored in the knowledge base, select the sample layout with the highest similarity as the matching layout of the document, and execute the corresponding task instruction to extract the document structured information to obtain the extraction result, and execute S3; S3: The user corrects the extraction result to obtain an accurate result, and then executes S4; S4: output the accurate result and verify the consistency between the extraction result and the accurate result. If they are inconsistent, execute S5; S5: searching similar task instructions in the knowledge base using word embedding query technology according to the accurate result, and generating new task instructions based on the similar task instructions, and executing S6; S6: Execute the new task instruction in S5 on the document layout to extract the document structured information to obtain the test result, and verify the consistency between the test result and the accurate result. If they are consistent, store the document layout and the corresponding task instruction as new sample layout and task instruction in the knowledge base and execute S2; if they are inconsistent, execute S5.

[0007] The present invention also proposes an interactive document structured information extraction device, comprising: A document recognition module is used to recognize documents and obtain document formats; Knowledge base, used to store sample layouts and task instructions; A layout matching module, used to match the document layout with the sample layout, select the sample layout with the highest similarity as the matching layout of the document, and execute the corresponding task instruction; The structured information extraction module is used to execute the task instruction to extract the document structured information to obtain the extraction result; A user interaction module is used for the user to correct the extraction result and obtain and output an accurate result; An instruction generation module, used for searching similar task instructions in the knowledge base using word embedding query technology according to the accurate result, and generating new task instructions based on the similar task instructions; A testing module, used for executing the new task instruction generated by the instruction generation module on the document format to extract document structured information and obtain a test result; A verification module is used to verify the consistency of the extraction result, the accurate result and the test result.

[0008] An electronic device comprises a processor, a memory and a computer program stored in the memory and executable on the processor, wherein the computer program implements the steps of the above-mentioned method for extracting structured information from interactive documents when executed by the processor.

[0009] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the above-mentioned method for extracting structured information from interactive documents are implemented.

[0010] In the above technical solution, the technical effects and advantages provided by the present invention are: 1. The present invention can generate accurate task instructions in different types of documents through the layout matching service, significantly reducing the randomness of the Large Language Model (LLM), reducing the erroneous output caused by hallucinations, and ensuring high accuracy of information extraction; 2. Based on Few-shot learning technology, the present invention can quickly learn and adapt to new tasks with only a small amount of labeled data, reducing the dependence on large-scale pre-training data and improving the flexibility and practicality of the system; 3. The present invention can continuously adapt to new document types and information extraction requirements, maintain efficient and accurate information extraction performance, meet diverse and complex practical application requirements, greatly improve user work efficiency, and improve the level of automation of data processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0012] Figure 1 A schematic diagram of a flow chart of a method for extracting structured information from an interactive document provided by an embodiment of the present invention; Figure 2 A schematic block diagram of an interactive document structured information extraction device provided by an embodiment of the present invention; Figure 3 A schematic block diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0013] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings 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 ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0014] In the description of the embodiments of the present invention, it should be noted that the terms "center", "longitudinal", "lateral", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the embodiments of the present invention and simplifying the description, and do not indicate or imply that the devices or elements referred to must have a specific orientation, be constructed and operate in a specific orientation, and therefore cannot be understood as limitations on the embodiments of the present invention. In addition, the terms "first", "second", and "third" are used for descriptive purposes only and cannot be understood as indicating or implying relative importance.

[0015] In the description of the embodiments of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "connected" and "connection" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium. For ordinary technicians in this field, the specific meanings of the above terms in the embodiments of the present invention can be understood according to specific circumstances.

[0016] In the embodiments of the present invention, unless otherwise clearly specified and limited, the first feature being "above" or "below" the second feature may mean that the first and second features are in direct contact, or the first and second features are in indirect contact through an intermediate medium. Moreover, the first feature being "above", "above" and "above" the second feature may mean that the first feature is directly above or obliquely above the second feature, or simply means that the first feature is higher in level than the second feature. The first feature being "below", "below" and "below" the second feature may mean that the first feature is directly below or obliquely below the second feature, or simply means that the first feature is lower in level than the second feature.

[0017] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the embodiment of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0018] Combine the following Figure 1-Figure 3 Embodiments of the present invention are described.

[0019] like Figure 1 As shown, this embodiment discloses an interactive document structured information extraction method, which combines human-computer interaction, instant learning and large model generation Few-shot (small sample) example technology to achieve structured extraction of document information. Specifically, it includes the following steps: S1: Identify the document, obtain the document format, and execute S2; S2: Match the document layout with the sample layout stored in the knowledge base, select the sample layout with the highest similarity as the matching layout of the document, and execute the corresponding task instructions to extract the document structured information to obtain the extraction result, and execute S3; S3: The user corrects the extraction result, obtains an accurate result, and executes S4; S4: Output the accurate result and verify the consistency between the extracted result and the accurate result. If they are inconsistent, execute S5; S5: Based on the accurate result, use the word embedding query technology to query similar task instructions in the knowledge base, and generate new task instructions based on the similar task instructions, and execute S6; S6: Execute the new task instructions in S5 on the document layout to extract the document structured information to obtain the test results, and verify the consistency between the test results and the accurate results. If they are consistent, the document layout and the corresponding task instructions are stored in the knowledge base as new sample layouts and task instructions, and execute S2; if they are inconsistent, execute S5.

[0020] Among them, this method can perform detailed analysis of documents through the layout matching service, find the same or similar layout from historical data records, and call task instructions that match the layout. These instructions can effectively guide the large language model (LLM) to accurately locate and extract key information in information extraction tasks. This method significantly reduces the randomness of LLM and reduces or even eliminates erroneous outputs caused by hallucinations, thereby meeting the high-accuracy business needs in specific vertical applications. Through layout matching, the system can maintain consistent high-precision performance in different types of documents, ensuring the reliability of information extraction.

[0021] In the process of document information processing, reviewing the output results and performing interactive error correction are key steps to ensure data accuracy. Users can accurately correct the erroneous information in the extraction results by selecting the image frame. This interactive method allows users to directly select the text area that needs to be modified on the interface. The front-end system will capture the user's operation, including the selected characters and their coordinate information, and send this data to the back-end system for storage and processing. This process not only improves the accuracy of information extraction, but also gradually optimizes the system's learning model, making it more accurate in future tasks. This iterative optimization method based on user feedback improves the system's adaptability, enabling it to more effectively cope with complex and changing document formats.

[0022] In addition, in order to improve the efficiency and accuracy of document information extraction, the embodiment provided by the present invention adopts Embedding query technology, which quickly finds similar task instructions in the historical knowledge base and generates Few-shot examples based on these instructions. This method can achieve good learning effects with only a small amount of training data, thereby having the ability to adapt to new tasks without large-scale pre-training. Through word embedding technology, the system can perform semantic search on the knowledge base, extract examples that are highly relevant to the current task, and ensure that the generated small sample examples are highly representative and practical, thereby significantly improving the efficiency and accuracy of information extraction.

[0023] Through the comprehensive application of the above technologies, the method provided in this embodiment has a powerful instant learning ability, which enables the system to quickly adapt to new document types and task requirements, and provide leading output reliability and accuracy. In practical applications, this system can not only process a variety of documents, but also continuously learn and optimize, gradually improving its performance in different application scenarios.

[0024] like Figure 2 As shown, this embodiment discloses an interactive document structured information extraction device, including: A document recognition module is used to recognize documents and obtain document formats; Knowledge base, used to store sample layouts and task instructions; The layout matching module is used to match the document layout with the sample layout, select the sample layout with the highest similarity as the matching layout of the document, and execute the corresponding task instructions; The structured information extraction module is used to execute the task instruction to extract the document structured information to obtain the extraction result; The user interaction module is used for the user to correct the extraction results and obtain and output accurate results; An instruction generation module is used to query similar task instructions in the knowledge base based on the accurate results using word embedding query technology, and generate new task instructions based on similar task instructions; The test module is used to extract the document structured information from the new task instruction generated by the document layout execution instruction generation module to obtain the test result; Verification module, used to verify the consistency of extraction results, accurate results and test results.

[0025] like Figure 3 As shown, in some embodiments, the method disclosed in the present invention can be applied to some electronic devices, which need to include: a processor and a memory for storing processor executable instructions. The processor is configured to implement the above method when executing the executable instructions.

[0026] In addition, the method disclosed in the present invention may also be implemented as a computer-readable storage medium, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to implement the above method when executed by a processor.

[0027] In summary, the interactive document structured information extraction method, device, electronic device and storage medium disclosed in the present invention realize the structured extraction of document information by combining advanced technologies such as human-computer interaction, instant learning and large model generation of small sample examples. Users can check and correct erroneous information on the system interface, enabling the system to dynamically learn and optimize, thereby significantly improving the accuracy and robustness of information extraction. Through the document layout matching service, the system can dynamically generate task instructions suitable for different document layouts to ensure high-precision information extraction in different types of documents. The comprehensive application of these technologies gives the system powerful instant learning capabilities and adaptability. It overcomes the problem that the industry usually requires a large amount of data annotation and training to improve the accuracy of the engine, and reduces the maintenance costs during business system preparation, implementation and production application.

[0028] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An interactive document structured information extraction method, characterized in that: The following steps are involved: S1: Identify the document, obtain the document format, and execute S2; S2: Match the document layout with the sample layout stored in the knowledge base, select the sample layout with the highest similarity as the matching layout of the document, and execute the corresponding task instruction to extract the document structured information to obtain the extraction result, and execute S3; S3: The user corrects the extraction result to obtain an accurate result, and then executes S4; S4: output the accurate result and verify the consistency between the extraction result and the accurate result. If they are inconsistent, execute S5; S5: searching similar task instructions in the knowledge base using word embedding query technology according to the accurate result, and generating new task instructions based on the similar task instructions, and executing S6; S6: Execute the new task instruction in S5 on the document layout to extract the document structured information to obtain the test result, and verify the consistency between the test result and the accurate result. If they are consistent, store the document layout and the corresponding task instruction as new sample layout and task instruction in the knowledge base and execute S2; if they are inconsistent, execute S5.

2. An interactive document structured information extraction device, characterized in that: include: A document recognition module is used to recognize documents and obtain document formats; Knowledge base, used to store sample layouts and task instructions; A layout matching module, used to match the document layout with the sample layout, select the sample layout with the highest similarity as the matching layout of the document, and execute the corresponding task instruction; The structured information extraction module is used to execute the task instruction to extract the document structured information to obtain the extraction result; A user interaction module is used for the user to correct the extraction result and obtain and output an accurate result; An instruction generation module, used to query similar task instructions in the knowledge base using word embedding query technology according to the accurate result, and generate new task instructions based on the similar task instructions; A testing module, used for executing the new task instruction generated by the instruction generation module on the document format to extract document structured information and obtain a test result; A verification module is used to verify the consistency of the extraction result, the accurate result and the test result.

3. An electronic device, characterized in that: The invention comprises a processor, a memory and a computer program stored in the memory and executable on the processor, wherein the computer program implements the steps of the method for extracting structured information from an interactive document as claimed in claim 1 when executed by the processor.

4. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the interactive document structured information extraction method according to claim 1 are implemented.

Images