Method and apparatus for evaluating and processing generated tabular documents for retrieval

Abstract

This application relates to the technical fields of data processing and artificial intelligence, and discloses a method and apparatus for evaluating and processing retrieved tabular documents. The method includes: parsing the tabular document to obtain at least its table structure information, data content, and metadata information; performing multi-dimensional evaluation based on the table structure information, data content, and metadata information to obtain at least two of the following: table standardization evaluation result, metadata integrity evaluation result, data clarity evaluation result, and contextual relevance evaluation result; and weighting at least two of these results to obtain a comprehensive evaluation result for the tabular document. This method allows for a comprehensive evaluation of tabular documents, providing users with clear optimization guidance.

Description

Technical Field

[0001] This application relates to the fields of data processing, artificial intelligence and other technical fields, and in particular to a method and apparatus for evaluating and processing retrieved tabular documents. Background Technology

[0002] Currently, spreadsheets are among the most important data organization and analysis tools in enterprises, with a large amount of key business data, statistical reports, and bills of materials existing in spreadsheet format. When building enterprise-level retrieval and generation systems, enabling AI (Artificial Intelligence) to accurately retrieve, understand, and answer data-related questions from these spreadsheets is crucial. However, many spreadsheet files, for user convenience, use numerous unstructured or semi-structured formats such as merged cells, multi-level headers, comments, and special color markings. This makes it difficult for AI to correctly parse the row and column relationships when attempting to use them as knowledge sources, thus hindering the extraction of accurate data. Traditional technologies lack an effective method to evaluate whether a spreadsheet file is "AI-friendly," and cannot provide targeted optimization suggestions. Furthermore, the structured databases obtained through data cleaning and import tools cannot meet the needs of large-scale, routine knowledge base document optimization, and it is even more difficult to provide users with clear optimization suggestions for spreadsheet creation. Summary of the Invention

[0003] The embodiments of this application aim to at least partially address one of the technical problems in the related art. To this end, embodiments of this application propose a method and apparatus for evaluating and processing generated tabular documents.

[0004] The embodiments of this application provide a method for evaluating and processing retrieved tabular documents. This method includes: parsing the tabular document to obtain at least its table structure information, data content, and metadata information; performing multi-dimensional evaluation based on the table structure information, data content, and metadata information to obtain at least two of the following: table standardization evaluation result, metadata integrity evaluation result, data clarity evaluation result, and contextual relevance evaluation result; and weighting at least two of the following evaluation results to obtain a comprehensive evaluation result for the tabular document.

[0005] In some implementations, at least one of the table structure information, data content, and metadata information is subjected to multi-dimensional evaluation processing to obtain a table standardization evaluation result, including: determining merged cells in the table document based at least on the table structure information to obtain a cell merging evaluation result; determining the nesting level of the table header structure in the table document based at least on the table structure information to obtain a table header structure evaluation result; determining whether the data format of the same category of data in the table document is consistent based at least on the data content to obtain a data consistency evaluation result; and obtaining a table standardization evaluation result based on the cell merging evaluation result, the table header structure evaluation result, and the data consistency evaluation result.

[0006] In some implementations, the method further includes: parsing the table document to obtain the associated text of the table document; performing multi-dimensional evaluation processing based on at least one of the table structure information, data content, metadata information, and associated text to obtain a metadata integrity evaluation result, including: at least based on the metadata information, performing natural language processing to recognize the clarity of the table header name to obtain a table header name evaluation result; at least based on the metadata information, determining the meaning of the worksheet name in the table document to obtain a worksheet naming evaluation result; at least based on the metadata information and associated text, determining whether the metadata information contains annotation information for the target content or whether the associated text contains explanatory information for the target content to obtain an annotation / explanation evaluation result; and obtaining a data integrity evaluation result based on the table header name evaluation result, the worksheet naming evaluation result, and the annotation / explanation evaluation result.

[0007] In some implementations, at least one of the table structure information, data content, and metadata information is subjected to multi-dimensional evaluation processing to obtain a data clarity evaluation result, including: determining, based at least on the data content and metadata information, whether the table header or the annotation information in the metadata information contains unit information for numerical values, to obtain a data unit evaluation result; determining, based at least on the data content, meaningless characters in the table document, to obtain a character evaluation result; and obtaining a data clarity evaluation result based on the data unit evaluation result and the character evaluation result.

[0008] In some implementations, at least one of the table structure information, data content, metadata information, and associated text is subjected to multi-dimensional evaluation processing to obtain a contextual relevance evaluation result, including: determining whether the associated text interprets the data content based at least on the data content and associated text, thereby obtaining a contextual relevance evaluation result.

[0009] In some implementations, the method further includes: determining the content to be optimized in the table document based on the comprehensive evaluation results, and generating document optimization suggestions for the content to be optimized; generating a document evaluation report based on the comprehensive evaluation results and the document optimization suggestions, wherein the document evaluation report includes a multi-dimensional radar chart of the table document, a scorecard of the table document, document optimization suggestions, and location information of the content to be optimized in the table document.

[0010] In some implementations, the tabular document includes at least one worksheet; the method further includes: determining a worksheet to be optimized from the at least one worksheet based on a comprehensive evaluation result; outputting optimization and refactoring instruction information based on the worksheet to be optimized; and in response to receiving an optimization and refactoring instruction for the optimization and refactoring instruction information, refactoring the worksheet to be optimized into a one-dimensional table, and generating an optimized worksheet based on the one-dimensional table to obtain an optimized tabular document.

[0011] In some implementations, the tabular document includes at least one worksheet; the method further includes generating a data dictionary for the tabular document, wherein the data dictionary describes each worksheet and the contents of each worksheet in natural language.

[0012] In some implementations, the method further includes: in response to receiving a user-input question, retrieving a first answer from a form document based on the question, retrieving a second answer from an optimized form document based on the question, and outputting the first and second answers. The embodiments of this application provide an evaluation processing apparatus for retrieved generated table documents. The apparatus includes: a parsing module for parsing the table document to obtain at least table structure information, data content, and metadata information; an evaluation module for performing multi-dimensional evaluation processing based on the table structure information, data content, and metadata information to obtain at least two of the following: table standardization evaluation result, metadata integrity evaluation result, data clarity evaluation result, and contextual relevance evaluation result; and a processing module for weighting at least two of the following evaluation results to obtain a comprehensive evaluation result for the table document.

[0013] An embodiment of this application provides an electronic device, which includes: a memory, and one or more processors communicatively connected to the memory; the memory stores instructions executable by one or more processors, which are executed by one or more processors to cause the one or more processors to implement the steps of the method of any of the above embodiments.

[0014] Embodiments of this application provide a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the method of any of the above embodiments.

[0015] In the solution provided in this application, by parsing and extracting content from the table document, key scoring information that conforms to the characteristics of the table document can be accurately obtained. With the help of an AI scoring engine, the key scoring information can be evaluated in multiple dimensions. From an AI perspective, the parsability, comprehensibility and other content of a table document can be comprehensively evaluated, and optimization guidance can be provided to users based on the comprehensive evaluation results of the table document. Attached Figure Description

[0016] Figure 1 This application provides a flowchart illustrating a method for evaluating and processing generated tabular documents. Figure 2 A flowchart illustrating a method for evaluating and processing generated tabular documents for retrieval, provided as an embodiment of this application; Figure 3 A schematic diagram of an evaluation processing apparatus for retrieving generated tabular documents provided in this application; Figure 4 A block diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0017] The embodiments of this application are described in detail below. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0018] Currently, spreadsheets are one of the most important data organization and analysis tools in enterprises, with a large amount of key business data, statistical reports, bills of materials, etc., existing in spreadsheet format. When building an enterprise-level RAG (Retrieval-Augmented Generation) question-answering system, enabling AI to accurately retrieve, understand, and answer questions about the data from these spreadsheets, such as which product had the highest sales in the last quarter, is of great significance to the company's development.

[0019] Enterprises using AI question-answering systems based on RAG technology, especially those that need to extract and understand data from spreadsheets, typically face the following technical challenges: First, many spreadsheets, for the sake of human readability, use a large number of unstructured or semi-structured formats such as merged cells, multi-level headers, comments, and special color markings. This makes it difficult for AI to correctly parse the row and column relationships of the tables when attempting to use them as a knowledge source, thus preventing it from extracting accurate data.

[0020] Secondly, many spreadsheets contain only raw data, lacking necessary metadata descriptions. For example, the header of a data column might only be a vague quantity, making it impossible for AI to determine whether it represents order quantities or inventory quantities. This lack of context prevents AI from accurately answering questions about the data.

[0021] Finally, similar to documents, businesses completely lack an effective method to assess whether a spreadsheet is an AI-friendly document, and they also lack targeted optimization suggestions to transform massive amounts of human-readable spreadsheets into high-quality, machine-understandable knowledge assets.

[0022] In some embodiments, data import and cleaning tools, such as ETL tools (Extract, Transform, Load), can be used. These tools allow data analysts to write complex rules and scripts to load an irregular tabular document into a structured database through a series of transformations.

[0023] However, ETL tools also have some problems in practical use. For example, ETL tools require professional programming skills, making them extremely difficult to use. ETL tools are designed for professional data engineers and require writing code or performing complex graphical configurations. Ordinary business personnel or knowledge base administrators cannot use them at all, and they cannot meet the needs of large-scale, daily knowledge base document optimization.

[0024] For example, the purpose of ETL is data migration, not quality assessment of the source file itself. It cannot tell users what aspects of the spreadsheet document are good or bad, nor can it provide general optimization suggestions to guide future spreadsheet creation. For instance, ETL tools focus on the structural transformation of tabular data and are completely unable to handle equally important textual information in spreadsheet documents, such as chart titles, cell comments, or text boxes used to explain the background, which are crucial for AI to understand the context of the table.

[0025] Therefore, in view of the above, this application provides a method for evaluating and processing retrieved tabular documents. This method does not require programming and uses an automated intelligent evaluation approach to comprehensively evaluate the parsability and comprehensibility of a tabular document from an AI perspective, resulting in a clear and comprehensive evaluation result.

[0026] Figure 1 This application provides a flowchart illustrating a method for evaluating and processing generated tabular documents.

[0027] like Figure 1 As shown, the evaluation processing method 100 for retrieving generated tabular documents provided in this application includes steps S110-S130.

[0028] Step S110: Parse the table document to obtain at least the table structure information, data content, and metadata information of the table document.

[0029] Spreadsheet documents include Excel documents. Parsing a spreadsheet document involves, for example, parsing the row and column structure of the data range within the spreadsheet document; the table structure information of the spreadsheet document includes, for example, merged cells, multi-level (nested) headers, blank rows or columns, etc.; the data content includes, for example, the data in each cell, and it is also possible to determine the basic data type of each cell, such as numeric, text, date, etc.

[0030] Metadata information includes, for example, the worksheet name, header, cell comments, chart titles and descriptions extracted from a table document. Additionally, it may include associated text information, such as text boxes surrounding the extracted table and explanatory text.

[0031] Step S120: Based on the table structure information, data content and metadata information, perform multi-dimensional evaluation processing to obtain at least two of the following: table standardization evaluation result, metadata integrity evaluation result, data clarity evaluation result and contextual relevance evaluation result.

[0032] A multi-dimensional evaluation standard was designed to address the characteristics of tabular documents, including dimensions such as table standardization, metadata integrity, data clarity, and contextual relevance, with different weight values ​​assigned to each dimension. The evaluation dimension for table standardization specifically includes assessments of merged cells, table header structure, and data consistency; the evaluation dimension for metadata integrity specifically includes assessments of table header clarity, worksheet naming, and comments and explanations; the evaluation dimension for data clarity specifically includes assessments of the clarity of data units and the presence of meaningless characters; and the evaluation dimension for contextual relevance specifically includes assessments of the relationship between tables and text.

[0033] Then, based on the AI ​​scoring engine, the document tables are evaluated and scored in multiple dimensions to obtain at least two of the following: table standardization evaluation results, metadata integrity evaluation results, data clarity evaluation results, and contextual relevance evaluation results.

[0034] Step S130: Weight at least two of the table standardization assessment results, metadata integrity assessment results, data clarity assessment results, and context relevance assessment results to obtain a comprehensive assessment result of the table document.

[0035] Based on the weight values ​​of different evaluation dimensions, at least two of the evaluation results of table standardization, metadata integrity, data clarity, and contextual relevance are weighted to obtain the comprehensive evaluation result of the table document.

[0036] As can be seen, in the solution provided in this application, by parsing and extracting content from the table document, key scoring information that conforms to the characteristics of the table document can be accurately obtained. With the help of the AI ​​scoring engine, the key scoring information can be evaluated in multiple dimensions. From the AI ​​perspective, the parsability, comprehensibility and other content of a table document can be comprehensively evaluated, and optimization guidance can be provided to users based on the comprehensive evaluation results of the table document.

[0037] Figure 2 This is a flowchart illustrating a method for evaluating and processing generated tabular documents, provided as an embodiment of this application.

[0038] like Figure 2 As shown, the evaluation processing method 200 for retrieving generated tabular documents provided in this application embodiment includes steps S210-S240.

[0039] Step S210: Obtain the electronic document uploaded by the user and perform in-depth analysis on it.

[0040] Step S220: Evaluate based on AI friendliness scoring engine.

[0041] Step S230: Generate a comprehensive score and optimization suggestions.

[0042] Step S240: Present a visual report.

[0043] Combination Figure 1 , Figure 2 The content provides a detailed description of the solution provided in this application.

[0044] In one embodiment, a multi-dimensional evaluation process is performed based on at least one of table structure information, data content, and metadata information to obtain a table standardization evaluation result, including: determining merged cells in the table document based at least on table structure information to obtain a cell merging evaluation result; determining the nesting level of the table header structure in the table document based at least on table structure information to obtain a table header structure evaluation result; determining whether the data format of the same category of data in the table document is consistent based at least on data content to obtain a data consistency evaluation result; and obtaining a table standardization evaluation result based on the cell merging evaluation result, the table header structure evaluation result, and the data consistency evaluation result.

[0045] Based at least on the table structure information, merged cells in the table document are identified to obtain the cell merging evaluation results. For example, detecting and counting the number of merged cells is the biggest obstacle affecting AI parsing, and therefore it is given the highest weight.

[0046] Based at least on the table structure information, the nesting level of the table header structure in the table document is determined to obtain the table header structure evaluation result. For example, the system evaluates the clarity of the table header structure; a flat, one-dimensional table header structure can obtain a higher score, while a complex, multi-level table header will be penalized accordingly.

[0047] At least based on the data content, the engine determines whether the data format of the same category of data in the table document is consistent, thus obtaining the data consistency assessment result. For example, the engine will check whether the data format within a single column is consistent, such as whether there are mixed formats such as text and numbers in the same column. Inconsistent formatting in the table document will affect the final score.

[0048] Based on the cell merging evaluation results, header structure evaluation results, and data consistency evaluation results, the table standardization evaluation result is obtained. Specifically, the evaluation results of these three dimensions are weighted and calculated to generate the final table standardization evaluation result. The table standardization evaluation result accounts for 40% of the overall score, reflecting that table structure is a key factor in AI parsing table documents.

[0049] As can be seen, the embodiments of this application provide a quantitative standard that can be used to evaluate the "AI friendliness" of spreadsheet documents such as Excel documents. In particular, it focuses on evaluating pain points that have long plagued machine parsing, such as merged cells and multi-level headers, effectively improving the applicability and accuracy of machine evaluation.

[0050] In another embodiment, the method provided by this application further includes: parsing the table document to obtain the associated text of the table document; and performing multi-dimensional evaluation processing based on at least one of the table structure information, data content, metadata information, and associated text to obtain a metadata integrity evaluation result, including: Based at least on metadata information, the system uses natural language processing to assess the clarity of table header names and obtain evaluation results. For example, the system uses an NLP (Natural Language Processing) model to evaluate whether the table header names are clear, unambiguous, and accurately describe the data in the corresponding column. Table headers using ambiguous names, such as "Remarks" or "Data 1," will be identified and points will be deducted accordingly.

[0051] Based at least on metadata information, the meaning of worksheet names in a spreadsheet document is determined, resulting in a worksheet naming evaluation result. For example, the system might analyze whether the worksheet name has business meaning, and derive the worksheet naming evaluation result accordingly.

[0052] Based at least on metadata information and associated text, the system determines whether the metadata information contains annotations related to the target content, or whether the associated text contains explanatory information related to the target content, thus obtaining an annotation evaluation result. The target content may be, for example, key data or complex formulas. For instance, the system may also check whether the key data or complex formulas have accompanying annotations, including cell comments or explanatory information in text boxes, thereby obtaining an annotation evaluation result.

[0053] Based on the evaluation results of the table header name, worksheet naming, and annotation descriptions, a data integrity evaluation result is obtained. These three sub-dimensions are then comprehensively calculated to generate the final data integrity evaluation result. This data integrity evaluation result is another important evaluation dimension of the AI ​​scoring engine, ensuring the comprehensibility of the table document content; its weight is, for example, 25%.

[0054] In one embodiment, a multi-dimensional evaluation process is performed based on at least one of table structure information, data content, and metadata information to obtain a data clarity evaluation result, including: Based at least on the data content and metadata information, determine whether the header or comments in the metadata of the table document contain unit information for the numerical values, and obtain the data unit evaluation result. For example, the system can perform explicit analysis of data units based on the data content and metadata information. It can analyze whether the header or comments of the table document explicitly indicate the unit information of the numerical values, such as yuan, tons, percentages, etc., from the metadata information and data content obtained above.

[0055] Based at least on the data content, the system identifies meaningless characters in the table document and obtains character evaluation results. For example, the system checks whether the data content contains unnecessary special characters, spaces, or decorative symbols, etc.

[0056] The evaluation results of the two sub-dimensions, data unit evaluation and character evaluation, are comprehensively calculated to generate the final data clarity evaluation result. This dimension's evaluation result is crucial for ensuring data parsability and understandability, effectively improving data quality and usage efficiency. The data clarity evaluation result is the third key dimension of the AI ​​scoring engine, accounting for 20% of the weight.

[0057] In the embodiments provided in this application, a three-dimensional deep analysis of the table document can be performed, simultaneously understanding its physical structure (rows and columns, merging), data content (numerical values, text), and semantic metadata (table headers, annotations), and scoring can be performed based on the comprehensive analysis of the three, thereby achieving a comprehensive evaluation of the table document and improving the accuracy of the evaluation results.

[0058] In one embodiment, a multi-dimensional evaluation process is performed based on at least one of table structure information, data content, metadata information, and associated text to obtain a contextual relevance evaluation result, including: determining whether the associated text interprets the data content based at least on the data content and associated text, thereby obtaining a contextual relevance evaluation result.

[0059] The system analyzes data content and related text to determine whether the related text adequately explains and interprets the table content, thus obtaining a contextual relevance assessment result. Contextual relevance assessment is the fourth dimension of the AI ​​scoring engine, accounting for 15% of the weight. This dimension focuses on the degree of relevance between the table and the surrounding explanatory text, assessing whether these related texts can effectively explain the table's data sources, statistical methods, and key conclusions.

[0060] The evaluation results for table standardization are weighted at 40%, metadata integrity at 25%, data clarity at 20%, and contextual relevance at 15%. These weights can be used to weight the evaluation results across all dimensions, resulting in a comprehensive evaluation of the table document. This application's method not only provides insights into "good table-making habits" but also offers an automated tool and methodology for transforming massive amounts of human-readable spreadsheets into high-quality, structured knowledge assets that can be accurately understood and utilized by AI.

[0061] In one embodiment, the method provided by this application further includes: determining the content to be optimized in the table document based on the comprehensive evaluation results, and generating document optimization suggestions for the content to be optimized; generating a document evaluation report based on the comprehensive evaluation results and the document optimization suggestions, wherein the document evaluation report includes a multi-dimensional radar chart of the table document, a scorecard of the table document, document optimization suggestions, and location prompts for the content to be optimized in the table document.

[0062] The system will calculate the final comprehensive score based on the scores of the four evaluation dimensions mentioned above, according to the preset weights, and obtain the comprehensive evaluation result.

[0063] The comprehensive evaluation results can fully reflect the overall quality of the spreadsheet document. Simultaneously, the system will generate specific, actionable optimization suggestions for each deduction. For example, when the system detects that cells B2:D2 in worksheet 1 are merged cells, it will automatically generate corresponding optimization suggestions, such as: unmerge the cells in B2:D2 and fill the content "Second Quarter Sales" into cells B2, C2, and D2 respectively to ensure that each column has an independent header.

[0064] The document evaluation report is presented to users in a visual format. In addition to radar charts and scorecards, the report highlights specific problematic cells or areas in the electronic document preview and provides corresponding optimization suggestions, making it easy for users to understand at a glance.

[0065] As can be seen, the embodiments provided in this application can provide users with optimization suggestions and guidance on how to better improve the table structure, thereby enhancing its standardization, clarity and parsability.

[0066] In another embodiment, the tabular document includes at least one worksheet; the method further includes: determining a worksheet to be optimized from the at least one worksheet based on a comprehensive evaluation result; outputting optimization and refactoring instruction information based on the worksheet to be optimized; and in response to receiving an optimization and refactoring instruction for the optimization and refactoring instruction information, refactoring the worksheet to be optimized into a one-dimensional table, and generating an optimized worksheet based on the one-dimensional table to obtain an optimized tabular document.

[0067] This application also provides a one-click intelligent reconstruction solution. For common, regular, non-standard tables, such as typical multi-level header reports, the system will automatically convert the non-standard table into a flat, AI-friendly standard one-dimensional table and generate a new worksheet.

[0068] In another embodiment, the tabular document includes at least one worksheet; the method further includes generating a data dictionary for the tabular document, wherein the data dictionary describes each worksheet and the contents of each worksheet in natural language.

[0069] After analyzing a complex spreadsheet document containing multiple worksheets, the system can automatically generate a data dictionary for the file. The data dictionary can use natural language to describe the business meaning, data type, and value range of each worksheet and each column.

[0070] In another embodiment, the method provided by this application further includes: in response to receiving a question input by a user, retrieving a first answer from a form document based on the question, retrieving a second answer from an optimized form document based on the question, and outputting the first answer and the second answer.

[0071] The system also supports natural language question-and-answer simulation. In the scoring report, the system allows users to ask several questions about the table in natural language, such as which salesperson performed best. The system will display the first answer (which may be incorrect) that the AI ​​might give given the current table quality, and the second answer (which is correct) that might be given after optimizing the table document according to the suggestions.

[0072] This demonstrates that when users adjust the table document according to the optimization suggestions provided by the system, the system can almost always provide the correct answer in the second answer, reflecting the important role of optimization guidance in improving the quality, parsability, and comprehensibility of table documents.

[0073] Figure 3 This application provides a schematic diagram of an apparatus for evaluating and processing generated tabular documents.

[0074] like Figure 3 As shown, the evaluation processing apparatus 300 for retrieving generated tabular documents provided in this application includes: The parsing module 310 is used to parse the table document and obtain at least the table structure information, data content and metadata information of the table document.

[0075] The evaluation module 320 is used to perform multi-dimensional evaluation processing based on table structure information, data content and metadata information, and obtain at least two of the following evaluation results: table standardization evaluation result, metadata integrity evaluation result, data clarity evaluation result and context relevance evaluation result.

[0076] Processing module 330 is used to perform weighted processing on at least two of the table standardization assessment results, metadata integrity assessment results, data clarity assessment results, and context relevance assessment results to obtain a comprehensive assessment result of the table document.

[0077] In some embodiments, the evaluation module 320 is further configured to: determine merged cells in a table document based at least on table structure information, and obtain a cell merging evaluation result; determine the nesting level of the header structure in a table document based at least on table structure information, and obtain a header structure evaluation result; determine whether the data format of the same category of data in a table document is consistent based at least on data content, and obtain a data consistency evaluation result; and obtain a table standardization evaluation result based on the cell merging evaluation result, the header structure evaluation result, and the data consistency evaluation result.

[0078] In some embodiments, the apparatus 300 further includes a parsing module, which is configured to: parse the table document to obtain the associated text of the table document; the evaluation module 320 is further configured to: perform natural speech processing to recognize the clarity of the table header name based at least on metadata information, and obtain a table header name evaluation result; determine the meaning of the worksheet name of the table document based at least on metadata information, and obtain a worksheet naming evaluation result; determine whether there is annotation information for the target content in the metadata information or whether there is explanatory information for the target content in the associated text based at least on metadata information and associated text, and obtain an annotation description evaluation result; and obtain a data integrity evaluation result based on the table header name evaluation result, the worksheet naming evaluation result, and the annotation description evaluation result.

[0079] In some embodiments, the evaluation module 320 is further configured to: determine, at least based on the data content and metadata information, whether the header or annotation information in the metadata information of the table document contains unit information for numerical values, and obtain a data unit evaluation result; determine, at least based on the data content, meaningless characters in the table document, and obtain a character evaluation result; and obtain a data clarity evaluation result based on the data unit evaluation result and the character evaluation result.

[0080] In some embodiments, the evaluation module 320 is further configured to: determine whether the associated text interprets the data content based at least on the data content and the associated text, thereby obtaining a contextual relevance evaluation result.

[0081] In some embodiments, the apparatus 300 further includes a generation module, which is configured to: determine the content to be optimized in the table document based on the comprehensive evaluation results, and generate document optimization suggestions for the content to be optimized; and generate a document evaluation report based on the comprehensive evaluation results and the document optimization suggestions, wherein the document evaluation report includes a multi-dimensional radar chart of the table document, a scorecard of the table document, document optimization suggestions, and location information of the content to be optimized in the table document.

[0082] In some embodiments, the table document includes at least one worksheet; the apparatus 300 further includes a reconstruction module, which is configured to: determine a worksheet to be optimized from at least one worksheet based on a comprehensive evaluation result; output optimization reconstruction instruction information based on the worksheet to be optimized; and, in response to receiving an optimization reconstruction instruction for the optimization reconstruction instruction information, reconstruct the worksheet to be optimized into a one-dimensional table, and generate an optimized worksheet based on the one-dimensional table to obtain an optimized table document.

[0083] In some embodiments, the table document includes at least one worksheet; the apparatus 300 further includes a description module configured to: generate a data dictionary for the table document, wherein the data dictionary describes each worksheet and the contents of each worksheet in natural language.

[0084] In some embodiments, the apparatus 300 further includes a retrieval module, which is configured to: in response to receiving a question input by a user, retrieve a first answer from a form document based on the question, retrieve a second answer from an optimized form document based on the question, and output the first answer and the second answer.

[0085] It is understood that the specific functional implementation of the evaluation and processing device 300 for the retrieved form document can be referred to the evaluation and processing method for the retrieved form document described above, and will not be repeated here.

[0086] This application provides an electronic device, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps of the method described above.

[0087] This application provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the method in any of the above embodiments.

[0088] This application provides a computer program product that includes instructions that, when executed by a processor of a computer device, enable the computer device to perform the steps of the method described in any of the above embodiments.

[0089] Figure 4 A block diagram of an electronic device provided in an embodiment of this application.

[0090] This application provides an electronic device, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the method in any of the above embodiments.

[0091] like Figure 4 As shown, for ease of understanding, embodiments of this application illustrate a specific electronic device 400.

[0092] Electronic device 400 is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. Electronic device 400 may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0093] like Figure 4 As shown, device 400 includes a computing unit 401, which can perform various appropriate actions and processes based on a computer program stored in read-only memory (ROM) 402 or a computer program loaded from storage unit 408 into random access memory (RAM) 403. RAM 403 may also store various programs and data required for the operation of electronic device 400. The computing unit 401, ROM 402, and RAM 403 are interconnected via bus 404. Input / output (I / O) interface 405 is also connected to bus 404.

[0094] Multiple components in electronic device 400 are connected to I / O interface 405. These components include: input unit 406, such as a keyboard or mouse; output unit 407, such as various types of displays or speakers; storage unit 408, such as a disk or optical disk; and communication unit 409, such as a network interface card (NIC), modem, or wireless transceiver. Communication unit 409 allows electronic device 400 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0095] The computing unit 401 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 401 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 401 performs the various methods described above. For example, in some embodiments, any one or more of the methods described above can be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 408. In some embodiments, part or all of the computer program can be loaded and / or installed on the electronic device 400 via ROM 402 and / or communication unit 409. When the computer program is loaded into RAM 403 and executed by the computing unit 401, one or more steps of any one or more of the methods described above can be performed. Alternatively, in other embodiments, the computing unit 401 can be configured to perform any one or more of the methods described above by any other suitable means (e.g., by means of firmware).

[0096] It should be noted that the logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be specifically implemented in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this application, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which programs can be printed, because programs can be obtained electronically, for example, by optically scanning the paper or other media, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0097] It should be understood that various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0098] In the description of this application, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this application, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0099] In the description of this application, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", "clockwise", "counterclockwise", "axial", "radial", "circumferential", etc., indicating the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this application.

[0100] Furthermore, the terms "first," "second," etc., used in the embodiments of this application are for descriptive purposes only and should not be construed as indicating or implying relative importance, or implicitly specifying the number of technical features indicated in this embodiment. Therefore, features defined with terms such as "first" and "second" in the embodiments of this application can explicitly or implicitly indicate that the embodiment includes at least one of those features. In the description of this application, the word "multiple" means at least two or more, such as two, three, four, etc., unless otherwise explicitly and specifically defined in the embodiments.

[0101] In this application, unless otherwise explicitly specified or limited in the embodiments, the terms "installation," "connection," "joining," and "fixing" appearing in the embodiments should be interpreted broadly. For example, a connection can be a fixed connection, a detachable connection, or an integral part; it can also be a mechanical connection, an electrical connection, etc. Of course, it can also be a direct connection, or an indirect connection through an intermediate medium, or it can be the internal communication between two components, or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific implementation.

[0102] In this application, unless otherwise expressly specified and limited, "above" or "below" the second feature can mean that the first feature is in direct contact with the second feature, or that the first feature is in indirect contact with the second feature through an intermediate medium. Furthermore, "above," "on top of," and "over" the second feature can mean that the first feature is directly above or diagonally above the second feature, or simply that the first feature is at a higher horizontal level than the second feature. "Below," "below," and "under" the second feature can mean that the first feature is directly below or diagonally below the second feature, or simply that the first feature is at a lower horizontal level than the second feature.

Claims

1. A method for evaluating and processing retrieved tabular documents, characterized in that, The method includes: The table document is parsed to obtain at least the table structure information, data content, and metadata information of the table document; Based on the table structure information, the data content, and the metadata information, a multi-dimensional evaluation process is performed to obtain at least two of the following: table standardization evaluation result, metadata integrity evaluation result, data clarity evaluation result, and contextual relevance evaluation result of the table document. At least two of the table standardization assessment results, metadata integrity assessment results, data clarity assessment results, and context relevance assessment results are weighted to obtain a comprehensive assessment result for the table document.

2. The method according to claim 1, characterized in that, Based on at least one of the table structure information, the data content, and the metadata information, a multi-dimensional evaluation process is performed to obtain the table normalization evaluation result, including: Based at least on the table structure information, the merged cells in the table document are determined, and the cell merging evaluation result is obtained; Based at least on the table structure information, determine the nesting level of the table header structure in the table document to obtain the table header structure evaluation result; Based at least on the data content, determine whether the data format of the same category of data in the table document is consistent, and obtain the data consistency assessment result; Based on the cell merging evaluation results, the header structure evaluation results, and the data consistency evaluation results, the table standardization evaluation results are obtained.

3. The method according to claim 1, characterized in that, It also includes: parsing the table document to obtain the associated text of the table document; and performing multi-dimensional evaluation processing based on at least one of the table structure information, the data content, the metadata information, and the associated text to obtain the metadata integrity evaluation result, including: Based at least on the aforementioned metadata information, natural speech processing is used to identify the clarity of the table header name, and an evaluation result for the table header name is obtained. Based at least on the metadata information, determine the meaning of the worksheet name in the table document to obtain the worksheet naming evaluation result; Based at least on the metadata information and the associated text, determine whether the metadata information contains annotation information for the target content or whether the associated text contains explanatory information for the target content, and obtain the annotation and explanatory evaluation result; The data integrity assessment result is obtained based on the evaluation results of the header name, the worksheet naming, and the annotation description.

4. The method according to claim 1, characterized in that, Based on at least one of the table structure information, the data content, and the metadata information, a multi-dimensional evaluation process is performed to obtain the data clarity evaluation result, including: Based at least on the data content and the metadata information, determine whether the header of the table document or the annotation information in the metadata information contains unit information for numerical values, and obtain the data unit evaluation result; Based at least on the data content, identify meaningless characters in the table document and obtain character evaluation results; Based on the data unit evaluation results and the character evaluation results, the data clarity evaluation results are obtained.

5. The method according to claim 3, characterized in that, Based on at least one of the table structure information, the data content, the metadata information, and the associated text, a multi-dimensional evaluation process is performed to obtain the context relevance evaluation result, including: Based at least on the data content and the associated text, determine whether the associated text interprets the data content, and obtain the contextual relevance assessment result.

6. The method according to any one of claims 1-5, characterized in that, The method further includes: Based on the comprehensive evaluation results, the content in the table document that needs to be optimized is identified, and document optimization suggestions are generated for the content that needs to be optimized. Based on the comprehensive evaluation results and the document optimization suggestions, a document evaluation report is generated. The document evaluation report includes a multi-dimensional radar chart of the table document, a scorecard of the table document, the document optimization suggestions, and location information of the content to be optimized in the table document.

7. The method according to any one of claims 1-5, characterized in that, The table document includes at least one worksheet; the method further includes: Based on the comprehensive evaluation results, determine the worksheet to be optimized from the at least one worksheet; Based on the worksheet to be optimized, output optimization and reconstruction instructions; In response to receiving an optimization and restructuring instruction for the optimization and restructuring indication information, the worksheet to be optimized is restructured into a one-dimensional table, and an optimized worksheet is generated based on the one-dimensional table to obtain an optimized table document.

8. The method according to any one of claims 1-5, characterized in that, The table document includes at least one worksheet; the method further includes: A data dictionary is generated for the table document, wherein the data dictionary describes each worksheet and the contents of each worksheet in natural language.

9. The method according to claim 7, characterized in that, The method further includes: In response to receiving a question input by the user, a first answer is obtained by searching the table document based on the question, and a second answer is obtained by searching the optimized table document based on the question; Output the first answer and the second answer.

10. An apparatus for evaluating and processing generated tabular documents, characterized in that, The device includes: The parsing module is used to parse the table document and obtain at least the table structure information, data content and metadata information of the table document; The evaluation module is used to perform multi-dimensional evaluation processing based on the table structure information, the data content, and the metadata information to obtain at least two of the following evaluation results: table standardization evaluation result, metadata integrity evaluation result, data clarity evaluation result, and contextual relevance evaluation result of the table document. The processing module is used to perform weighted processing on at least two of the table standardization assessment results, the metadata integrity assessment results, the data clarity assessment results, and the context relevance assessment results to obtain a comprehensive assessment result of the table document.

Images