Multi-table incidence relation document auditing method and device

Abstract

The invention discloses a multi-table incidence relation document auditing method and device, and the method comprises the steps: obtaining a Word or Excel document, and carrying out the analysis; extracting table visual features of the analyzed document according to the layout visual model, and combining the table visual features with original marks in the document to obtain a structured table; performing semantic analysis on the fields of the structured table and the text description in the document, and unifying the fields and the text description to obtain a mapping relationship between the fields and the text description; establishing an association relationship between the general table and the sub-table according to a mapping relationship between the fields and the text description to obtain a cross-table association relationship; performing consistency verification on the cross-table association relationship to obtain a verification result; and generating a visual auditing report according to the verification result. According to the method and the device, automatic analysis, cross-table proofreading and consistency verification are carried out on the document containing a plurality of tables and text description, and the document auditing efficiency and accuracy are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of document processing, in particular to a multi-table correlation relationship document auditing method and device. BACKGROUND

[0002] In the management process of scientific research projects, scientific research fund declaration, project budget preparation and contract execution are crucial links. In order to ensure the rational use of funds and the smooth implementation of the project, relevant departments usually need to provide detailed financial and budget data. Such documents usually contain multiple tables, involving budget items, fund use plans, contract terms and other content. These tables not only need to accurately reflect the financial status of the project, but also need to meet the requirements of relevant policies and approval processes.

[0003] For example: the document may need to include a total budget table, a sub-unit budget table, and specific item explanations for each unit; each sub-table and its textual explanation must be consistent with the data in the total table, and ensure that the data between multiple tables meet certain constraint relationships (such as the sum of each item equals the total item). These requirements pose high standards for the accuracy and consistency of the document.

[0004] However, in actual operation, the existing technology still faces many problems in processing multi-table documents in scientific research fund declaration, project budget preparation and contract execution:

[0005] 1. Reliance on manual inspection: manual inspection is time-consuming and prone to errors, which cannot effectively guarantee the accuracy of the data;

[0006] 2. Rule-driven limitations: traditional rule-based parsing methods are difficult to adapt to documents with complex formats or merged cells;

[0007] 3. Insufficient cross-table verification: most existing technologies only support internal logical correction of single tables, lack consistency checking across tables and texts, and are difficult to fully guarantee the data consistency between tables;

[0008] 4. Unable to identify visual features: hidden borders, merged cells and other visual features commonly found in Word or Excel documents are easily missed, leading to data parsing errors.

[0009] In summary, due to reliance on manual inspection, rule-driven limitations, insufficient cross-table verification and inability to identify visual features, the existing technology results in low efficiency and poor accuracy in auditing documents containing multiple tables. SUMMARY

[0010] Therefore, the present application provides a multi-table correlation relationship document auditing method and device to solve the problem of low efficiency and poor accuracy in auditing multi-table documents in the existing technology.

[0011] To achieve the above objectives, this application provides the following technical solution:

[0012] Firstly, a method for reviewing documents with multiple table relationships includes:

[0013] Step 1: Obtain the Word or Excel document and parse it;

[0014] Step 2: Extract the table visual features of the parsed Word or Excel document based on the layout visual model;

[0015] Step 3: Combine the visual features of the table with the existing tags inside the Word or Excel document to obtain a structured table;

[0016] Step 4: Perform semantic parsing on the fields of the structured table and the text descriptions in the Word or Excel document, and unify the fields and text descriptions to obtain the mapping relationship between fields and text descriptions;

[0017] Step 5: Establish the relationship between the master table and the sub-tables based on the mapping relationship between the fields and the text descriptions, and obtain the cross-table relationship;

[0018] Step 6: Perform a consistency check on the cross-table relationships and obtain the check results;

[0019] Step 7: Generate a visual audit report based on the verification results.

[0020] Preferably, in step 2, the visual features of the table include table boundaries, cell border styles, page crossover markers, and merged cell information.

[0021] Preferably, step 4 specifically includes: performing semantic parsing on the fields of the structured table and the text descriptions in the Word or Excel document, and converting them into semantic vectors after word segmentation; unifying the semantic vectors of the fields and text descriptions based on synonyms and a pre-trained language model to obtain the mapping relationship between the fields and text descriptions.

[0022] Preferably, when performing semantic parsing on the fields of the structured table and the text descriptions in the Word or Excel document, the monetary units and units of measurement are unified.

[0023] Preferably, step 5 specifically includes: establishing constraint edges between each table field node and text description node based on the graph structure and the mapping relationship between the fields and text descriptions, and establishing the association relationship between the master table and the sub-tables based on the predefined constraint rules to obtain the cross-table association relationship.

[0024] Preferably, when establishing the association between the master table and the sub-tables according to the predefined constraint rules, numerical normalization and constraint checking algorithms are used.

[0025] Preferably, the constraint rule is: total table amount = sum of sub-table amounts; table amount = amount mentioned in the text description.

[0026] Preferably, in step 6, numerical comparison and differential calculation are used to perform consistency verification on the cross-table association relationship.

[0027] Preferably, in step 6, the verification results include inconsistent table positions, cell coordinates, and text paragraphs.

[0028] Secondly, a document review device with multiple table relationships includes:

[0029] The document parsing module is used to acquire Word or Excel documents and parse them.

[0030] The table recognition module is used to extract the table visual features of the parsed Word or Excel document based on the layout visual model, and combine the table visual features with the original marks inside the Word or Excel document to obtain a structured table.

[0031] The semantic matching module is used to perform semantic parsing on the fields of the structured table and the text descriptions in the Word or Excel document, and to unify the fields and text descriptions to obtain the mapping relationship between the fields and text descriptions.

[0032] The association modeling module is used to establish the association relationship between the master table and the sub-table based on the mapping relationship between the fields and the text descriptions, so as to obtain the cross-table association relationship;

[0033] The verification module is used to perform consistency verification on the cross-table association relationship and obtain the verification result.

[0034] The result output module is used to generate a visual audit report based on the verification results.

[0035] Compared with the prior art, this application has at least the following beneficial effects:

[0036] This application provides a method for reviewing documents with multiple table relationships. The method involves acquiring and parsing a Word or Excel document; extracting the visual features of the tables from the parsed Word or Excel document based on a layout visual model; combining these visual features with existing internal markup in the Word or Excel document to obtain a structured table; performing semantic analysis on the fields of the structured table and the text descriptions in the Word or Excel document, and unifying the fields and text descriptions to obtain a mapping relationship between fields and text descriptions; establishing relationships between the master table and sub-tables based on the mapping relationship to obtain cross-table relationships; performing consistency checks on the cross-table relationships to obtain the check results; and generating a visual review report based on the check results. This method for reviewing documents with multiple table relationships enables automatic parsing, cross-table verification, and consistency validation of documents containing multiple tables and text descriptions, improving the efficiency and accuracy of document review. Attached Figure Description

[0037] To more intuitively illustrate the prior art and this application, exemplary drawings are provided below. It should be understood that the specific shapes and structures shown in the drawings should not generally be regarded as limiting conditions for implementing this application; for example, based on the technical concept disclosed in this application and the exemplary drawings, those skilled in the art are able to easily make conventional adjustments or further optimizations to the addition / reduction / classification, specific shapes, positional relationships, connection methods, size ratios, etc. of certain units (components).

[0038] Figure 1 A flowchart of a multi-table related document review method provided in Embodiment 1 of this application;

[0039] Figure 2 This is a schematic diagram of the structure of a document review method with multiple table relationships provided in Embodiment 1 of this application;

[0040] Figure 3 This is a schematic diagram of the structured table recognition structure provided in Embodiment 1 of this application;

[0041] Figure 4 This is a schematic diagram of the cross-table relationship modeling structure provided in Embodiment 1 of this application;

[0042] Figure 5 The consistency verification flowchart provided in Embodiment 1 of this application;

[0043] Figure 6 This is a schematic diagram of a multi-table document review device provided in Embodiment 2 of this application. Detailed Implementation

[0044] The present application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0045] In the description of this application: unless otherwise stated, "a plurality of" means two or more. The terms "first," "second," "third," etc., in this application are intended to distinguish the objects referred to and do not have any special meaning in terms of technical connotation (e.g., they should not be construed as an emphasis on importance or order). Expressions such as "including," "comprising," and "having" also mean "not limited to" (certain units, components, materials, steps, etc.).

[0046] The terms used in this application, such as "upper," "lower," "left," "right," and "middle," are generally used to indicate the general relative positional relationship for the purpose of intuitive understanding by referring to the accompanying drawings, and are not absolute limitations on the positional relationship in the actual product.

[0047] Example 1

[0048] Please see Figure 1 and Figure 2 This embodiment provides a method for reviewing documents with multiple table relationships, applicable to scenarios requiring consistency verification of multiple tables, such as budget application materials, contract terms, and research funding reports. The method includes:

[0049] S1: Obtain a Word or Excel document and parse it;

[0050] S2: Extract the table visual features of the parsed Word or Excel document based on the layout visual model;

[0051] For details, please refer to Figure 3 This step extracts table visual features based on the layout visual model. These features include table boundaries, cell border styles, page break markers, and merged cell information. By introducing a fusion mechanism of visual feature recognition and traditional document parsing, this step not only relies on the document's built-in table objects but also utilizes visual features such as font size, line thickness, and relative cell positions to recover hidden or non-standardized table structures. This avoids parsing failures encountered by existing methods when dealing with "hidden borders" or "page break tables."

[0052] S3: Combine the visual features of the table with the original markup inside the Word or Excel document to obtain a structured table;

[0053] S4: Perform semantic parsing on the fields of the structured table and the text descriptions in the Word or Excel document, and unify the fields and text descriptions to obtain the mapping relationship between the fields and text descriptions;

[0054] Specifically, this step involves semantic parsing of the fields in the structured table and the text descriptions in Word or Excel documents, followed by word segmentation and conversion into semantic vectors. Based on synonyms and a pre-trained language model, the semantic vectors of the fields and text descriptions are unified to obtain the mapping relationship between the fields and the text descriptions. During the semantic parsing process, monetary units (yuan, ten thousand yuan) and units of measurement (unit, set) are unified.

[0055] In other words, this step involves word segmentation and embedding representation of the table header, field names, and text descriptions, followed by field name normalization based on synonyms and a pre-trained language model (e.g., "equipment fee" is equivalent to "instrument equipment fee"). This step uses semantic vector representation combined with synonym normalization, overcoming the limitation of traditional rule matching in handling semantic diversity.

[0056] S5: Establish the relationship between the master table and the sub-tables based on the mapping relationship between fields and text descriptions to obtain cross-table relationships;

[0057] For details, please refer to Figure 4 This step establishes constraint edges between each table field node and text description node based on the graph structure and the mapping relationship between fields and text descriptions. It also establishes the association relationship between the master table and the sub-tables according to the predefined constraint rules, thus obtaining the cross-table association relationship (i.e., realizing cross-table and cross-text mapping).

[0058] The relationship between the master table and sub-tables is established based on predefined constraint rules using numerical normalization and constraint checking algorithms. Predefined constraint rules include: Master table amount = sum of sub-table amounts; Table amount = amount mentioned in the text description.

[0059] This step utilizes a two-tiered cross-table modeling framework of "semantic constraints + numerical constraints." It not only establishes logical mappings between fields through semantic similarity matching but also strengthens constraint consistency through numerical verification. For example, in the master table, "Equipment Fee = 500,000 yuan," this embodiment automatically locates the "Equipment Fee" entry in the sub-table and verifies whether the total amount matches. This two-tiered constraint greatly enhances the robustness of cross-table consistency verification.

[0060] S6: Perform consistency checks on cross-table relationships and obtain the check results;

[0061] For details, please refer to Figure 5 When performing consistency checks on cross-table relationships, numerical comparison and differential calculation are used. The check result is an error report, which includes inconsistent table positions, cell coordinates, and text paragraphs. The error report also provides semantic hints and an analysis of the source of the conflict.

[0062] S7: Generate a visual audit report based on the verification results.

[0063] Specifically, this step generates a visual audit report based on the verification results. Users can click on highlighted items on the interface to directly jump to the table or text location. This embodiment also provides possible correction suggestions based on AI inference results. Employing an interactive audit and correction mechanism, users can not only view errors but also adopt recommended correction solutions with a single click, achieving an organic combination of automated verification and human decision-making.

[0064] The following example, "Research Funding Budget Document Review," illustrates the method provided in this embodiment:

[0065] S1: Obtain a Word document containing a research project budget application and parse it;

[0066] S2: Extract the table visual features of the analyzed research project budget application Word document based on the layout visual model;

[0067] S3: Combine the visual features of the table with the original tags inside the Word document for the research project budget application to obtain a structured table (including the total budget table and the sub-unit budget table).

[0068] S4: Parse the "Equipment Costs" in the master budget table and automatically match the equipment cost items of each unit in the sub-unit budget tables;

[0069] S5: Compare the totals in the sub-table with the values ​​in the master table;

[0070] S6: Verify whether the "equipment cost budget amount" of 500,000 yuan in the description text is consistent with the table;

[0071] S7: Generate an audit report, indicating "The total amount in the sub-table does not match the total amount in the master table, with a difference of 20,000 yuan".

[0072] The document review method with multiple table relationships provided in this embodiment has the following advantages:

[0073] 1. Automation: Reduces the workload of manual verification;

[0074] 2. Accuracy: Combining visual features with semantic understanding improves the accuracy of parsing and verification;

[0075] 3. Cross-table capability: Enables multi-level consistency checks between the master table, sub-tables, and descriptive text;

[0076] 4. High adaptability: It can handle non-standard format documents and supports complex situations such as merging cells and hiding rows.

[0077] In summary, the document review method for multiple tables provided in this embodiment can automatically parse, cross-table verify, and confirm consistency of documents containing multiple tables and text descriptions, thereby improving the efficiency and accuracy of document review.

[0078] Example 2

[0079] Please see Figure 6 This embodiment provides a document review device with multiple table relationships, including:

[0080] The document parsing module is used to acquire Word or Excel documents and parse them.

[0081] The table recognition module is used to extract the table visual features of the parsed Word or Excel document based on the layout visual model, and combine the table visual features with the original marks inside the Word or Excel document to obtain a structured table.

[0082] The semantic matching module is used to perform semantic parsing on the fields of the structured table and the text descriptions in the Word or Excel document, and to unify the fields and text descriptions to obtain the mapping relationship between the fields and text descriptions.

[0083] The association modeling module is used to establish the association relationship between the master table and the sub-table based on the mapping relationship between the fields and the text descriptions, so as to obtain the cross-table association relationship;

[0084] The verification module is used to perform consistency verification on the cross-table association relationship and obtain the verification result.

[0085] The result output module is used to generate a visual audit report based on the verification results.

[0086] For details on the implementation of each module in a multi-table document review device, please refer to the above description of the limitations of a multi-table document review method, which will not be repeated here.

[0087] The technical features of the above embodiments can be combined in any way (as long as there is no contradiction in the combination of these technical features). For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described; these embodiments not explicitly written should also be considered to be within the scope of this specification.

Claims

1. A method for reviewing documents with multiple table relationships, characterized in that, include: Step 1: Obtain the Word or Excel document and parse it; Step 2: Extract the table visual features of the parsed Word or Excel document based on the layout visual model; Step 3: Combine the visual features of the table with the existing tags inside the Word or Excel document to obtain a structured table; Step 4: Perform semantic parsing on the fields of the structured table and the text descriptions in the Word or Excel document, and unify the fields and text descriptions to obtain the mapping relationship between fields and text descriptions; Step 5: Establish the relationship between the master table and the sub-tables based on the mapping relationship between the fields and the text descriptions, and obtain the cross-table relationship; Step 6: Perform a consistency check on the cross-table relationships and obtain the check results; Step 7: Generate a visual audit report based on the verification results.

2. The method for reviewing documents with multiple table relationships according to claim 1, characterized in that, In step 2, the visual features of the table include table boundaries, cell border styles, page crossover markers, and merged cell information.

3. The document review method with multiple table relationships according to claim 1, characterized in that, Step 4 specifically includes: performing semantic parsing on the fields of the structured table and the text descriptions in the Word or Excel document, and converting them into semantic vectors after word segmentation; unifying the semantic vectors of the fields and text descriptions based on synonyms and a pre-trained language model to obtain the mapping relationship between the fields and text descriptions.

4. The document review method with multiple table relationships according to claim 3, characterized in that, When performing semantic parsing on the fields of the structured table and the text descriptions in the Word or Excel document, the monetary units and units of measurement are unified.

5. The method for reviewing documents with multiple table relationships according to claim 1, characterized in that, Step 5 specifically includes: establishing constraint edges between each table field node and text description node based on the graph structure and the mapping relationship between the fields and text descriptions, and establishing the association relationship between the master table and the sub-tables according to the predefined constraint rules to obtain the cross-table association relationship.

6. The method for reviewing documents with multiple table relationships according to claim 5, characterized in that, The process of establishing the association between the master table and the sub-tables based on predefined constraint rules employs numerical normalization and constraint checking algorithms.

7. The method for reviewing documents with multiple table relationships according to claim 5, characterized in that, The constraint rules are as follows: Total table amount = sum of sub-table amounts; Table amount = amount mentioned in the text description.

8. The method for reviewing documents with multiple table relationships according to claim 1, characterized in that, In step 6, numerical comparison and differential calculation are used to perform consistency verification on the cross-table association relationship.

9. The method for reviewing documents with multiple table relationships according to claim 1, characterized in that, In step 6, the verification results include inconsistent table positions, cell coordinates, and text paragraphs.

10. A document review device with multiple table relationships, characterized in that, include: The document parsing module is used to acquire Word or Excel documents and parse them. The table recognition module is used to extract the table visual features of the parsed Word or Excel document based on the layout visual model, and combine the table visual features with the original marks inside the Word or Excel document to obtain a structured table. The semantic matching module is used to perform semantic parsing on the fields of the structured table and the text descriptions in the Word or Excel document, and to unify the fields and text descriptions to obtain the mapping relationship between the fields and text descriptions. The association modeling module is used to establish the association relationship between the master table and the sub-table based on the mapping relationship between the fields and the text descriptions, so as to obtain the cross-table association relationship; The verification module is used to perform consistency verification on the cross-table association relationship and obtain the verification result. The result output module is used to generate a visual audit report based on the verification results.

Images