Calibration-based PDF (Portable Document Format) document page number range matching method and system

Abstract

The invention discloses a calibration-based PDF (Portable Document Format) document page number range matching method and system. The method comprises the following steps: acquiring a PDF bidding document to be analyzed; converting the PDF bidding document into an editable bidding document with a reserved text content style structure, and performing page number calibration and directory calibration on the editable bidding document to obtain a standardized bidding document; obtaining a target review item for the bidding document, performing retrieval in the standardized bidding document based on at least one retrieval mode, and obtaining a page number range where a text associated with the target review item is located; and combining retrieval output contents obtained in the catalog and / or the text of the standardized bidding document in at least one retrieval mode to obtain an effective review page number range of the target review item. According to the method, effective review text content is positioned for expert review or AI simulation expert review, so that the review process can focus on related content to generate a review conclusion, and the review efficiency is improved.

Description

Technical Field

[0001] This invention relates to the field of PDF document processing technology, specifically to a calibration-based PDF document page number range matching method and system. Background Technology

[0002] In the field of bidding and tendering review, under the traditional model, expert reviewers need to manually browse through anywhere from 300 to 2000 pages of tender documents, searching for the corresponding content for each review item. This method relies entirely on manpower and is extremely inefficient. Currently, information retrieval and location methods for tender PDF documents are relatively limited. Although some general document processing tools can achieve some functions, they have obvious shortcomings. For example, when faced with problems such as inconsistent table of contents and page number correspondences or missing page numbers in the tender documents, they cannot accurately locate the page numbers of review items in the PDF document. This makes it difficult to accurately locate the relevant table of contents within a large document, and the return of a large amount of irrelevant information further increases the expert's screening burden. Summary of the Invention

[0003] To address the problems existing in the prior art, this invention provides a calibration-based PDF document page number range matching method and system, which can provide an effective review page number range for the review process and improve review efficiency. The technical solution is as follows: Firstly, a calibration-based PDF document page number range matching method is provided, comprising the following steps: Obtain the PDF tender document to be analyzed; Convert PDF bid documents into editable bid documents that retain the main text content style and structure, and perform page number and table of contents calibration on the editable bid documents to obtain standardized bid documents; The method involves obtaining target review items for a tender document, searching the standardized tender document using at least one search method, and obtaining the page range of the main text associated with the target review items. The search method has a corresponding input content generation method that uses the overall original content of the target review items as the data source to generate input content. The input content is then used to execute the corresponding search method to obtain the search output content under that search method. The input content represents key information about the target review items. The search output content includes the page range of the main text associated with the input content, i.e., the target review items. By combining the search output obtained from the table of contents and / or text of the standardized tender document under at least one search method, the effective review page range of the target review item is obtained.

[0004] In some implementations, the step of searching standardized tender documents based on at least one retrieval method to obtain the page range of the main text associated with the target review item includes: Step A: Search the standardized tender document catalog using at least one search method to obtain the page range of the main text associated with the target review item, denoted as the catalog search page range; And / or, Step B involves searching the main text of the standardized tender document using at least one search method to obtain the page range of the main text associated with the target review item, denoted as the main text search page range; wherein the search method used in Step A may be the same as or different from the search method used in Step B, and the number of search methods used in Step A may be different from or the same as the number of search methods used in Step B.

[0005] In some implementations, after searching the standardized tender document catalog based on at least one search method to obtain the catalog search page range, the method further includes: Step C: Based on at least one retrieval method, search the text within the range of page numbers in the directory retrieval to determine whether there is text in the text within the range of page numbers in the directory retrieval that is associated with the target review item; if so, proceed to step C1; otherwise, proceed to step C2; wherein, the retrieval method used in step A is different from or the same as the retrieval method used in step C. Step C1: Based on the page range of the main text that is actually associated with the target review item within the page range of the directory search, the precise page range is used as the precise page range. The page range of the directory search and the precise page range are used as the search output content. Step C2 determines that the page number range results from the directory search are invalid.

[0006] In some embodiments, the retrieval method includes a first retrieval method, which includes the following steps: Based on the overall original content of the target review item, the first search input content is generated using a large language model. The content generated by the large language model includes the keywords, constraints, and prompt words of the target review item. Based on the first search input content, semantic association analysis is performed at the selected search position in the standardized bidding document using a large language model to determine the first page range of the main text associated with the first search input content, i.e., associated with the target review item. The selected search position is the table of contents of the bidding document and / or the main text.

[0007] In some implementations, the first retrieval method includes the following steps: Based on the overall original content of the target review item, the first search input content is generated using a large language model. The content generated by the large language model includes the keywords, constraints, and prompt words of the target review item. Based on the first search input, at least one of the following first and second methods is executed to determine the first page range of the text associated with the target review item; wherein the first page range is determined based on the union of the results of the execution of the first and second methods; The first method: Based on the first search input content, semantic association analysis is performed on the table of contents of the standardized bidding document using a large language model to infer the table of contents titles associated with the first search input content, and the page number range of the main text associated with the target review item is determined based on the page number of the associated table of contents titles. The second method is to use a large language model to perform semantic matching analysis on the main body of the standardized bidding document based on the first search input content, and to determine the main body content that semantically matches the first search input content. Based on the page number of the semantically matched main body content, the range of page numbers of the main body content associated with the target review item is determined. In some embodiments, the retrieval method has a second retrieval method, which includes the following steps: Obtain and parse the bidding documents to extract the content arrangement requirements for the bid documents; A relational database of "review items - corresponding sections - content elements" is generated based on content arrangement requirements; For the target review item, the existence of a corresponding "review item-corresponding section-content element" data item is determined based on the associated database. If it exists, the text content of the data item is used as the second search input. The search is performed at the selected search position in the tender document to obtain the second page range of the main text associated with the target review item. The selected search position is the table of contents of the tender document and / or the main text.

[0008] In some implementations, after obtaining the second page range of the text associated with the target review item, the method further includes: verifying whether the text within the second page range is associated with the target review item, determining whether the text content associated with the review item in the tender document conforms to the content arrangement requirements of the tender document, and marking format mismatch information if it does not conform. The verification method includes: using a large language model to obtain the semantic understanding features of the overall original content of the target review item and the main text within the second page range, and determining whether the two are related.

[0009] In some embodiments, the retrieval method includes a third retrieval method, which comprises the following steps: The third search keyword is extracted based on the target review item. The third search keyword is used as the third search input content. The search is performed at the selected search position in the tender document to obtain the third page range of the main text associated with the target review item. The search includes text matching.

[0010] In some implementations, the step of combining the search outputs obtained from the table of contents and / or text of the standardized tender document under at least one search method to obtain the effective review page range of the target review item includes: determining the union of the search outputs obtained under at least one search method as the effective review page range of the target review item.

[0011] Secondly, a calibration-based PDF document page number range matching system is provided, including: The PDF document acquisition unit is used to acquire the PDF tender documents to be analyzed. The calibration unit is used to convert PDF bid documents into editable bid documents that retain the main text content style and structure, and to perform page number calibration and table of contents calibration on the editable bid documents to obtain standardized bid documents; A retrieval unit is used to obtain target review items for a tender document, and to search within a standardized tender document based on at least one retrieval method to obtain the page range of the main text associated with the target review item. The retrieval method has a corresponding method for generating retrieval input content, which uses the overall original content of the target review item as a data source to generate retrieval input content. The retrieval input content is then used to execute the corresponding retrieval method to obtain the retrieval output content under that retrieval method. The retrieval input content is used to characterize key information of the target review item. The retrieval output content includes the page range of the main text associated with the retrieval input content, i.e., the target review item. The page number range acquisition unit is used to combine the retrieval output content obtained from the table of contents and / or text of the standardized bidding document under at least one retrieval method to obtain the effective review page number range of the target review item.

[0012] This invention provides a calibration-based PDF document page number range matching method and system, which offers the following advantages: The PDF document page number range matching method of this invention, through the standardization process of bidding documents, is compatible with various PDF formats. By calibrating page numbers and table of contents in editable bidding documents, it effectively addresses the problem of missing or disorganized page numbers, ensuring that various types of bidding documents can be effectively parsed. Experts no longer need to search for information page by page; through the precise table of contents range provided by this method, they can directly locate relevant content, significantly improving the efficiency of the entire bidding process. The precise table of contents range provided by this invention provides a clear analytical boundary for expert review or AI-simulated expert review, enabling the review process to focus on relevant content to generate review conclusions, thereby improving review efficiency. Attached Figure Description

[0013] Figure 1 This is a flowchart illustrating the calibration-based PDF document page number range matching method in the embodiments of this application; Figure 2 This is a schematic diagram of the structure of the calibration-based PDF document page number range matching system in the embodiments of this application. Detailed Implementation

[0014] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0015] See Figure 1 This application provides a calibration-based PDF document page number range matching method, including the following steps: Step 1: Obtain the PDF tender document to be analyzed; Step 2: Convert the PDF bid document into an editable bid document that retains the main text content style and structure, and perform page number calibration and table of contents calibration on the editable bid document to obtain a standardized bid document; Step 3: Obtain the target review item for the tender document, and search the tender document using at least one search method to obtain the page range of the main text associated with the target review item; wherein, the search method has a search input content generation method corresponding to its own search method to generate search input content using the overall original content of the target review item as the data source, and use the search input content to execute the corresponding search method to obtain the search output content under the search method; the search input content is used to characterize the key information of the target review item; the search output content includes the page range of the main text associated with the search input content, i.e., the target review item; Step 4: Combine the search output obtained from the table of contents and / or text of the standardized tender document under at least one search method to obtain the effective review page range for the target review item.

[0016] In this embodiment, the page range of the tender document's main text for the target review item is determined based on different retrieval methods. This effectively locates the page range of the review item within the PDF tender document, providing effective review content for expert review or AI-simulated expert review. This allows the review process to focus on relevant content to generate review conclusions, improving review efficiency. In this embodiment, for example, there are three retrieval methods, denoted as retrieval method A, retrieval method B, and retrieval method C. Retrieval method A is used to search the entire tender document for the main text content associated with the review item, yielding page range A. Retrieval method B is used to search the entire tender document for the main text content associated with the review item, yielding page range B. Retrieval method C is used to search the entire tender document for the main text content associated with the review item, yielding page range C. Finally, the union of page ranges A, B, and C is used as the effective review page range for the target review item. Of course, the retrieval input content used when using retrieval methods A, B, and C must correspond to the corresponding retrieval method.

[0017] In this embodiment, converting PDF bid documents into editable bid documents can be achieved using existing PDF parsing and conversion tools. The parsed document retains the original PDF document's formulas, tables, images, and multi-level headings, maintaining their format and layout. For images, the parsed document can simultaneously display the extracted text content and the original image. The PDF parsing and conversion tool can convert to Markdown or JSON format, preserving the original format and layout, and can identify and obtain the coordinate position of each heading. This invention employs OCR technology, Markdown conversion, and JSON conversion to solve the problem of processing purely scanned PDFs. Furthermore, page number calibration and table of contents calibration effectively address the issue of missing or disordered page numbers, ensuring that various types of PDF bid documents can be effectively parsed. Page number calibration can be achieved through virtual page number generation combined with Markdown heading positioning techniques. It should be noted that the text representation associated with the target evaluation items in this application can be used to analyze the text content of the target evaluation item scores of bidding suppliers.

[0018] In one implementation, step 3 above, which involves searching the tender document using at least one retrieval method to obtain the page range of the main text associated with the target review item, includes: Step A: Search the table of contents of the tender document using at least one search method to obtain the page range of the main text associated with the target review item, denoted as the table of contents search page range; And / or, Step B involves searching the main text of the tender document using at least one search method to obtain the page range of the main text associated with the target review item, denoted as the main text search page range; wherein the search method used in step A may be the same as or different from the search method used in step B, and the number of search methods used in step A may be different from or the same as the number of search methods used in step B.

[0019] In this embodiment of the application, the retrieval process is divided into directory retrieval and text retrieval. It can be understood that the retrieval in the directory is actually to infer the correlation between the review item information and each title in the directory. When a title is found to be associated with the review item information (the correlation is greater than a preset threshold), the page number range of the text associated with the review item can be obtained based on the page number range between the page number of the title marked in the target and the page number of the next title at the same level as the title. That is, the page number range result of the directory retrieval. Searching within the main text involves performing a correlation analysis between the review item information and the content of specific areas within the main text of the tender document.

[0020] In this embodiment, in addition to searching the tender document based on different search methods, it also considers searching in the table of contents and / or in the main text. Therefore, searches can be performed in combination based on different types of search methods and search locations (table of contents, main text) to determine the search results, and the union of different search results is used as the effective review page range of the target review item. Using one search method to search the tender document yields the search output content, i.e., the page range of the main text associated with the target review item. This process may involve searching only the table of contents, only the main text, or simultaneously searching both the table of contents and the main text of the tender document, i.e., the full text of the tender document. Furthermore, other different search methods can be used to search the tender document again to obtain other search output content. Similarly, in the process of searching using these other different search methods, searches may be performed only on the table of contents, only on the main text, or simultaneously on both the table of contents and the main text of the tender document, i.e., the full text of the tender document. This can be illustrated illustratively. For example, if there are two search methods, denoted as Search Method A and Search Method B, then using Search Method A, we can search the table of contents to obtain page range A; using Search Method B, we can search the table of contents to obtain page range B; and combining the page ranges A and B, we obtain the effective review page range. Optionally, the union of the page ranges A and B can be used as the effective review page range. Alternatively, we can use Search Method A to search the table of contents to obtain page range A; using Search Method B, we can search the text to obtain page range B; and combining the page ranges A and B, we obtain the effective review page range. Optionally, the union of the page ranges A and B can be used as the effective review page range. Alternatively, a search can be performed in the text using search method A to obtain page number range A; a search can be performed in the text using search method B to obtain page number range B; and the effective review page number range can be obtained by combining the page number range results A and B. Optionally, the effective review page number range can be obtained by combining the page number range results A and B. Furthermore, a search can be performed in the table of contents using search method A to obtain page number range A1, in the text using search method A to obtain page number range A2, and in the text using search method B to obtain page number range B. The effective review page number range can be obtained by combining the page number range results A1, A2, and B. Optionally, the effective review page number range can be obtained by combining the page number range results A1, A2, and B, and so on.Of course, it should be noted that different search methods have different accuracy of search results. Between directory search and text search, text search takes longer but is more accurate and comprehensive, while directory search is more time-saving but may be incomplete or inaccurate. When performing a search based on different types of search methods and search locations (directory, text), it is necessary to comprehensively consider the balance between search time and accuracy and comprehensiveness to determine the appropriate combination.

[0021] The method of searching for a selected search location of a tender document provided in this application embodiment can have two specific methods. In specific implementation, both method one and method two can be used.

[0022] Method 1 includes the following steps: The first step is to match the table of contents of the tender document using a search method to determine the title associated with the target evaluation item in the table of contents. Then, based on the page number of the title, determine the page number of the main text in the tender document associated with the target evaluation item, and record it as the initial page number range. The second step is to match the text of the tender document based on a search method to determine the page range of the text that is associated with the target review item. This is denoted as the secondary page range. The search method used in the first step may be different or the same as the search method used in the second step. Based on the degree of overlap between the initial page number range and the secondary page number range obtained in the first step, the effective page number range is determined.

[0023] Method 2 includes the following steps: The first step is to match the table of contents of the tender document using a search method to determine the title associated with the target evaluation item in the table of contents. Then, based on the page number of the title, determine the page number of the main text in the tender document associated with the target evaluation item, and record it as the initial page number range. The second step involves matching the text within the initially defined page range using a search method to determine if any text related to the target review item exists. If it does, the page range containing the text associated with the target review item is determined and denoted as the secondary page range. This confirms the validity of the initially defined page range, and both the initial and secondary page ranges are provided as matching results. The initial page range represents a rough page range, while the secondary page range represents a more precise page range that has been further narrowed down. If the text does not exist, the initial page range is deemed invalid. The search method used in the first step may be different from or the same as the search method used in the second step.

[0024] In one implementation, after step A above, i.e., after searching the catalog of tender documents based on at least one search method to obtain the catalog search page range, the method further includes: Step C: Based on at least one retrieval method, search the text within the range of page numbers in the directory retrieval to determine whether there is text in the text within the range of page numbers in the directory retrieval that is associated with the target review item; if so, proceed to step C1; otherwise, proceed to step C2; wherein, the retrieval method used in step A is different from or the same as the retrieval method used in step C. Step C1: Based on the page range of the main text that is actually associated with the target review item within the page range of the directory search, the precise page range is used as the precise page range. The page range of the directory search and the precise page range are used as the search output content. Step C2 determines that the page number range results from the directory search are invalid.

[0025] This application provides a method for verifying the validity of directory search results. It should be noted that a search method has a corresponding search input content and a corresponding search output result. If the search methods used in steps A and C are the same, then the search input content used in the two steps is consistent. Otherwise, if the search methods used in steps A and C are different, then the search input content used in the two steps is inconsistent.

[0026] In one embodiment, the retrieval method in step 3 above includes a first retrieval method, which includes the following steps: Step 301: Based on the overall original content of the target review item, generate the first search input content using a large language model. The content generated by the large language model includes the keywords, constraints, and prompts of the target review item. Step 302: Based on the first search input content, use a large language model to perform semantic association analysis at the selected search position in the tender document to determine the first page range of the main text associated with the first search input content, i.e., associated with the target review item. The selected search position is the table of contents of the tender document and / or the main text.

[0027] Specifically, the large language model in this application embodiment can be implemented using conventional techniques, such as based on the Alibaba Qwen2.5 model. The semantic parsing process in the first retrieval method in this application embodiment includes the following steps: (1) Obtain the review item data from the collected historical tender document dataset to form a historical review item dataset; extract the entities of the review items based on the historical review item dataset, and classify the entities into primary key entities and secondary key entities. The primary key entities represent the keywords of the review items, and the secondary key entities represent the constraints of the review items. (2) Retrieve first supplementary information based on the primary key entity, the first supplementary information being used to characterize the general constraint information of the primary key entity; retrieve second supplementary information based on the secondary key entity, the second supplementary information being used to characterize the implicit constraint information contained in the secondary key entity; during the retrieval of the first supplementary information, the second supplementary information and the secondary key entity information are used as auxiliary supplementary information to assist the retrieval of the first supplementary information, and the retrieval process is based on the database of the review item dataset in the preset knowledge graph or the historical bidding document dataset; in this application, the retrieval of the first supplementary information is based on the second supplementary information and the secondary key entity as auxiliary, which on the one hand can make the model clearer about the specific application scenario of the review item, and can understand that the rules and constraints of the same review item are different in different application scenarios or business domains; on the other hand, it makes the first supplementary information clearer that the information that needs to be provided is inconsistent with the secondary key entity information and the secondary key entity can be used as a reference for the supplementary information. (3) Based on the main key entity, the first supplementary information, the secondary key entity, and the second supplementary information, semantic encoding and parsing are performed to form the retrieval input content.

[0028] Specifically, considering the large amount of data processing involved in the first retrieval method, especially for tender documents with hundreds or thousands of pages, the first retrieval method has a significant time difference between the table of contents and the main text, and can be used for different application scenarios. In this embodiment, the first retrieval method is specifically described as follows: The first retrieval method includes the following steps: Step 3011: Based on the overall original content of the target review item, generate the first search input content using a large language model. The content generated by the large language model includes the keywords, constraints, and prompt words of the target review item. Step 3012: Based on the first search input content, perform at least one of the following first method and second method to determine the first page range of the text associated with the target review item; wherein, the first page range is determined based on the union of the results of the first method and the second method. Step 3013, First Method: Based on the first retrieval input content, semantic association analysis is performed on the table of contents of the tender document using a large language model to infer the table of contents titles associated with the first retrieval input content. Based on the page numbers of the associated table of contents titles, the page number range of the main text associated with the target review item is determined. It should be noted that in this method, the page number range of the main text content under the associated table of contents title is determined based on the page number of the associated table of contents title, which is the first page number range. Specifically, it can be understood that based on the page number of the associated table of contents title, combined with the page number of the next sibling table of contents title associated with that associated table of contents title, the page number range of the main text content under that associated table of contents title can be determined, i.e., the first page number range. Step 3014, Second method: Based on the first search input content, use a large language model to perform semantic matching analysis on the main body of the tender document to determine the main body content that semantically matches the first search input content, and determine the page number range of the main body content associated with the target review item based on the page number of the semantically matched main body content. In this embodiment, a large language model is used to analyze the semantic compatibility between the overall content of the target review item and the main text of the tender document. Based on the overall content of the review item, keywords and constraints for retrieval are parsed. For example, if the overall content of a review item is "Bidder's performance: The bidder has completed ground centralized nitrogen production system design or mine design performance since January 1, 2022. 3 points are awarded for each completed design, up to a maximum of 15 points," then the keywords "ground centralized nitrogen production system design" and "mine design performance" can be parsed, along with the constraint "since January 1, 2022." Based on the parsing results of the overall content of the target review item using the large language model, combined with information such as the prompt word "Prompt," "question" data is automatically generated. Using the large language model, the "answer" is searched throughout the entire tender document to obtain the content location (page number range) of the entire tender document associated with the target review item. This embodiment achieves a deeper level of semantic adaptation.

[0029] In the embodiments of this application, the first retrieval method utilizes the concept of semantic adaptation. The execution time of the first method is less than that of the second method. In the first method, by understanding the overall semantics of the review items, the subordinate relationship and degree of association between each title and the review items are inferred from all titles in the document directory, and the title with the highest probability of containing the text associated with the review items is determined. This first method not only improves the effectiveness and comprehensiveness of page range determination from the perspective of semantic adaptation, but also reduces the amount of data processing for page range determination compared to the second method, that is, reduces the time required for page range determination.

[0030] Therefore, in practical applications, the choice between the first and second methods should be determined based on the application scenario's requirements and time constraints. Given sufficient time, the second method provides a more comprehensive and accurate range of page numbers for the main text associated with the review items. Specifically, the time required to analyze the entire main text of the current bid document using the second method can be assessed by comprehensively considering the total number of pages in the bid document, the number of pages in historical bid documents analyzed using the second method, and the time consumed.

[0031] Specifically, the first method further includes: recording the range of page numbers of the main text associated with the target review item, determined based on the page numbers of the associated table of contents titles, as the first candidate page number range; Based on the initial search input, semantic matching analysis is performed on the main text within the first candidate page range of the tender document to determine the main text content that semantically matches the initial search input. Based on the page numbers of the semantically matched main text content, the actual page range of main text associated with the target review item within the first candidate page range is determined. This step essentially verifies the results of the table of contents search page range. If no main text content semantically matches the initial search input exists in the table of contents search page range results, the table of contents search page range results are invalid; if such content exists, the table of contents search page range results are valid, and the page range of main text associated with the target review item can be further narrowed.

[0032] In one embodiment, the retrieval method in step 3 above has a second retrieval method, which includes the following steps: Step 311: Obtain and parse the bidding documents to extract the content arrangement requirements for the bid documents. Step 312: Generate a relational database of "review items - corresponding sections - content elements" based on the content arrangement requirements; Step 313: For the target review item, determine whether there is a corresponding "review item-corresponding section-content element" data item based on the associated database. If it exists, use the text content of the data item as the second search input content, and search at the selected search position in the tender document to obtain the second page range of the main text associated with the target review item. The selected search position is the table of contents of the tender document and / or the main text.

[0033] This application fully utilizes the information in the bidding documents to analyze the content arrangement requirements and format specifications for the tender documents. For the evaluation items disclosed in the bidding documents, it determines which section of the tender document the relevant content should be placed in, establishing a "evaluation item - corresponding section - content element" association database. It can be understood that the "section" name generally appears as a title. Based on this, the "corresponding section" name can be directly found in the title. Therefore, based on the page number corresponding to the title, the page number range of the main text under that title can be determined, i.e., the page number range of the main text associated with the evaluation item.

[0034] In this embodiment, a pre-built association database of "review item - corresponding section - content element" is used to optimize the conventional retrieval method. Based on the generation of retrieval input content based on review item information, the method of generating retrieval input content based on "corresponding section" and "content element" information is expanded. By utilizing the correspondence between "review item" and "corresponding section" and "content element", retrieval efficiency is improved and the retrieval input content is expanded.

[0035] It should be noted that: the text content based on the data item is used as the second search input, and searches are performed at selected search locations in the tender document, including: The search is performed directly based on the name of the "corresponding section" as the second search input, and the search is performed at the selected search location in the bidding document. The search includes text matching; the search is performed directly on the bidding document using the name of the "corresponding section", which saves search matching time.

[0036] Specifically, in one implementation, the step of constructing the associated database in the second retrieval method includes: Step 3101: Based on the target review items, extract the keywords of the review items and record them as the first keyword; Step 3102: Parse the tender document. In the content layout requirements section of the tender document, obtain the location of the keywords of the review items and the context of the keyword location. It should be noted that the context includes the keywords. The content layout requirements section includes all parts that may make requirements on the content layout of the tender document, such as the "response file format" and "instructions to respondents". Step 3103: Identify the context and determine the keywords in the context that match the structural titles in the tender document format file, denoted as the second keywords. Here, the structural titles represent the structural descriptive text content such as the title framework structure and content descriptions listed in the tender document format file provided in the tender document. Step 3104: Based on the distribution positions of the first keyword and the second keyword in the context, as well as the remaining text besides the first and second keywords, parse the semantics in the context to generate the positional relationship that the content corresponding to the first keyword and the content corresponding to the second keyword should satisfy in the entire tender document; in this step, parsing the context semantics to generate the positional relationship can be done using a preset neural network or deep learning trained on a sample set; Step 3105: Integrate all generated positional relationship expressions associated with the target review item to obtain a relational database of "first keyword - second keyword - content arrangement requirement keywords related to the first keyword", which is the corresponding relational database of "review item - corresponding section - content element". The content arrangement requirement keywords related to the first keyword include: key information representing the content arrangement requirements of the review item that is unrelated to the second keyword and its positional relationship with the second keyword, identified based on context parsing. For example, "performance proof materials should include the first page of the contract, the amount page, etc." This content arrangement requirement indicates that the key information of the performance proof materials includes the first page of the contract and the amount page.

[0037] It is understandable that the second keyword may not exist in step 3103 above. It is also understandable that the location of the review item may not be specified in some extracted contexts, but rather in terms of its own formatting requirements. For contexts where the second keyword is not identified, further contextual semantic analysis is still necessary and should be performed. Steps 3104 and 3105 should then be executed. Step 3104 involves analyzing the semantics of the context, not only analyzing positional relationships but also analyzing key information representing the content layout and formatting requirements that the first keyword should meet. This information is then integrated with all the analysis results in step 3105 to obtain all the content layout requirements for the review item, including positional relationships, layout information, and formatting requirements. Of course, for ambiguous formatting requirements in the tender document, industry practices can be used to supplement and clarify them, ensuring effective guidance for defining page ranges.

[0038] In one implementation, after step 314 above, that is, after obtaining the second page range of the text associated with the target review item, the method further includes: Step 315: Verify whether the text within the second page range is associated with the target review item, and determine whether the text content associated with the review item in the tender document meets the content arrangement requirements of the tender document. If it does not meet the requirements, mark the format mismatch information. The verification method includes: using a large language model to obtain the semantic understanding features of the overall original content of the target review item and the text within the second page range, and determine whether the two are associated. That is, based on the semantic features, analyze whether the two are semantically compatible and associated, thereby determining whether the text within the second page range can be used to review the score of the target review item.

[0039] This application provides a method for verifying whether a bid document conforms to the content layout and format requirements of the tender document. Utilizing a relational database retrieval method, if the bid document content follows the content layout requirements of the tender document, and if content associated with the "corresponding section" and "content element" is found in the tender document, it can be determined that the found second page range contains content associated with the target review item. If the main text within the second page range is associated with the target review item, it is determined that the main text content associated with the review item in the bid document conforms to the content layout requirements of the tender document; otherwise, if the main text within the second page range is not associated with the target review item, a format mismatch message is marked. This application fully utilizes the tender document's layout requirements for bid documents to improve page range matching efficiency.

[0040] In one embodiment, the retrieval method in step 3 above includes a third retrieval method, which includes the following steps: Step 321: Extract third search keywords based on the target review item, use the third search keywords as the third search input, and perform a search at the selected search location in the tender document to obtain the third page range of the text associated with the target review item. The search includes text matching. It can be understood that among the first, second, and third search methods provided in this application, the first search method uses semantic understanding for retrieval, and the overall accuracy of the search results is greater than that of the second and third search methods. The second search method fully utilizes the format requirements of the tender document. In one embodiment, a calibration-based PDF document page range matching method is provided, including the following steps: (101) Obtain the PDF tender document to be analyzed; (102) Convert the PDF bid document into an editable bid document with preserved text content style and structure, and perform page number calibration and table of contents calibration on the editable bid document to obtain a standardized bid document; (103) Using the first method of the first retrieval method described above, a search is performed in the catalog of the standardized tender documents to obtain the range of page numbers D1 of the main text associated with the review item; Using the second retrieval method described above, searches are performed in the table of contents and the main text of the standardized bidding document. Based on the table of contents search, the page range D2 of the main text associated with the review item is obtained, and based on the main text search, the page range D3 of the main text associated with the review item is obtained. Using the third retrieval method described above, searches were conducted in both the table of contents and the main text of the standardized bidding document to obtain the page range D4 of the main text associated with the review items. (104) Based on the combination of page ranges D1, D2, D3, and D4, the effective review page range of the target review item is obtained. Furthermore, the union of page ranges D1, D2, D3, and D4 can be used as the effective review page range of the target review item.

[0041] In one implementation, another calibration-based PDF document page number range matching method is provided, comprising the following steps: (111) Obtain the PDF tender document to be analyzed; (112) Convert the PDF bid document into an editable bid document with preserved text content style and structure, and perform page number calibration and table of contents calibration on the editable bid document to obtain a standardized bid document; (113) Using the first method of the first retrieval method described above, search the catalog of the standardized tender document to obtain the page range D1 of the main text associated with the review item; using the second method of the first retrieval method described above, search the main text of the standardized tender document to obtain the page range D2 of the main text associated with the review item. Using the second retrieval method described above, searches are performed in the table of contents and the main text of the standardized bidding document. Based on the table of contents search, the page range D3 of the main text associated with the review item is obtained, and based on the main text search, the page range D4 of the main text associated with the review item is obtained. Using the third retrieval method described above, searches were conducted in both the table of contents and the main text of the standardized bidding document to obtain the page range D5 of the main text associated with the review items. (114) Based on the combined page ranges D1, D2, D3, D4, and D5, the effective review page range for the target review item is obtained. Furthermore, the union of page ranges D1, D2, D3, D4, and D5 can be used as the effective review page range for the target review item. This embodiment utilizes multiple retrieval methods, integrating review item semantic parsing, tender document format requirements, and document page number directory calibration mechanisms. Compared to the simple keyword matching of existing technologies, the positioning error rate is significantly reduced. For example, for the "similar performance" review item, content that does not meet the time and type requirements can be accurately excluded, returning only the relevant directory range.

[0042] In one implementation, step 4 above, which involves combining the search outputs obtained from the table of contents and / or main text of the standardized tender document under at least one search method to obtain the effective review page range for the target review item, includes: determining the union of the search outputs obtained under at least one search method as the effective review page range for the target review item. Of course, in this embodiment, combining multiple search contents can be based on a mutual verification mechanism of the search outputs under multiple search methods, determining the page range simultaneously determined by multiple search methods as the final effective review page range; this can be understood as a voting mechanism, where the page range recognized by multiple search methods is considered more credible. However, in this embodiment, to avoid missing any page numbers that may contain content related to the review item, the union of multiple search outputs is taken. Based on this method, in subsequent steps, the main text of each page within the union of all page ranges can be semantically understood to verify whether it is a truly valid page range.

[0043] See Figure 2 This application provides a calibration-based PDF document page number range matching system, including: The PDF document acquisition unit is used to acquire the PDF tender documents to be analyzed. The calibration unit is used to convert PDF bid documents into editable bid documents that retain the main text content style and structure, and to perform page number calibration and table of contents calibration on the editable bid documents to obtain standardized bid documents; A retrieval unit is used to obtain target review items for a tender document, and to search within a standardized tender document based on at least one retrieval method to obtain the page range of the main text associated with the target review item. The retrieval method has a corresponding method for generating retrieval input content, which uses the overall original content of the target review item as a data source to generate retrieval input content. The retrieval input content is then used to execute the corresponding retrieval method to obtain the retrieval output content under that retrieval method. The retrieval input content is used to characterize key information of the target review item. The retrieval output content includes the page range of the main text associated with the retrieval input content, i.e., the target review item. The page number range acquisition unit is used to combine the retrieval output content obtained from the table of contents and / or text of the standardized bidding document under at least one retrieval method to obtain the effective review page number range of the target review item.

[0044] For specific limitations regarding the PDF document page number range matching system, please refer to the limitations of the PDF document page number range matching method mentioned above, which will not be repeated here. Each unit in the aforementioned PDF document page number range matching system can be implemented entirely or partially through software, hardware, or a combination thereof. These units can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each unit.

[0045] In some embodiments, the PDF document page number range matching system provided in this invention can be implemented using a combination of hardware and software. As an example, the PDF document page number range matching system provided in this invention can be directly embodied as a combination of software modules executed by a processor. The software modules can be located in a storage medium, which is located in a memory. The processor reads the executable instructions included in the software modules in the memory and combines them with necessary hardware (e.g., including the processor and other components connected to the bus) to complete the PDF document page number range matching method provided in this invention.

[0046] This invention is not limited to the specific embodiments described above. Any modifications made by those skilled in the art based on the above concept without creative effort are within the scope of protection of this invention.

Claims

1. A calibration-based PDF document page number range matching method, characterized in that, include: Obtain the PDF tender document to be analyzed; Convert PDF bid documents into editable bid documents that retain the main text content style and structure, and perform page number and table of contents calibration on the editable bid documents to obtain standardized bid documents; The method involves obtaining target review items for a tender document, searching the standardized tender document using at least one search method, and obtaining the page range of the main text associated with the target review items. The search method has a corresponding input content generation method that uses the overall original content of the target review items as the data source to generate input content. The input content is then used to execute the corresponding search method to obtain the search output content under that search method. The input content represents key information about the target review items. The search output content includes the page range of the main text associated with the input content, i.e., the target review items. By combining the search output obtained from the table of contents and / or text of the standardized tender document under at least one search method, the effective review page range of the target review item is obtained.

2. The calibration-based PDF document page number range matching method according to claim 1, characterized in that, The step of searching standardized bidding documents using at least one retrieval method to obtain the page range of the main text associated with the target review item includes: Step A: Search the standardized tender document catalog using at least one search method to obtain the page range of the main text associated with the target review item, denoted as the catalog search page range; And / or, Step B involves searching the main text of the standardized tender document using at least one search method to obtain the page range of the main text associated with the target review item, denoted as the main text search page range; wherein the search method used in Step A may be the same as or different from the search method used in Step B, and the number of search methods used in Step A may be different from or the same as the number of search methods used in Step B.

3. The calibration-based PDF document page number range matching method according to claim 2, characterized in that, After retrieving the standardized tender document catalog based on at least one retrieval method and obtaining the catalog's page range, the search also includes: Step C: Based on at least one retrieval method, search the text within the range of page numbers in the directory retrieval to determine whether there is text in the text within the range of page numbers in the directory retrieval that is associated with the target review item; if so, proceed to step C1; otherwise, proceed to step C2; wherein, the retrieval method used in step A is different from or the same as the retrieval method used in step C. Step C1: Based on the page range of the main text that is actually associated with the target review item within the page range of the directory search, the precise page range is used as the precise page range. The page range of the directory search and the precise page range are used as the search output content. Step C2 determines that the page number range results from the directory search are invalid.

4. The calibration-based PDF document page number range matching method according to claim 3, characterized in that, The search method includes a first search method, which includes the following steps: Based on the overall original content of the target review item, the first search input content is generated using a large language model. The content generated by the large language model includes the keywords, constraints, and prompt words of the target review item. Based on the first search input content, semantic association analysis is performed at the selected search position in the standardized bidding document using a large language model to determine the first page range of the main text associated with the first search input content, i.e., associated with the target review item. The selected search position is the table of contents of the bidding document and / or the main text.

5. The calibration-based PDF document page number range matching method according to claim 4, characterized in that, The first retrieval method includes the following steps: Based on the overall original content of the target review item, the first search input content is generated using a large language model. The content generated by the large language model includes the keywords, constraints, and prompt words of the target review item. Based on the first search input, at least one of the following first and second methods is executed to determine the first page range of the text associated with the target review item; wherein the first page range is determined based on the union of the results of the execution of the first and second methods; The first method: Based on the first search input content, semantic association analysis is performed on the table of contents of the standardized bidding document using a large language model to infer the table of contents titles associated with the first search input content, and the page number range of the main text associated with the target review item is determined based on the page number of the associated table of contents titles. The second method is to use a large language model to perform semantic matching analysis on the main body of the standardized bidding document based on the first search input content, and then determine the main body content that semantically matches the first search input content. Based on the page number of the semantically matched main body content, the range of page numbers of the main body content associated with the target review item is determined.

6. The calibration-based PDF document page number range matching method according to claim 3, characterized in that, The search method has a second search method, which includes the following steps: Obtain and parse the bidding documents to extract the content arrangement requirements for the bid documents; A relational database of "review items - corresponding sections - content elements" is generated based on content arrangement requirements; For the target review item, it is determined whether there is a corresponding "review item-corresponding section-content element" data item based on the associated database. If it exists, the text content of the data item is used as the second search input content, and a search is performed at the selected search position in the tender document to obtain the second page range of the main text associated with the target review item. The selected search position is the table of contents of the tender document and / or the main text.

7. The calibration-based PDF document page number range matching method according to claim 6, characterized in that, Following the second page range of the main text associated with the target review item, it also includes: Verify whether the text within the second page range is associated with the target review item, and determine whether the text content associated with the review item in the tender document conforms to the content arrangement requirements of the tender document. If it does not conform, mark the format mismatch information. The verification method includes: using a large language model to obtain the semantic understanding features of the overall original content of the target review item and the main text within the second page range, and determining whether the two are related.

8. The calibration-based PDF document page number range matching method according to claim 3, characterized in that, The search method includes a third search method, which includes the following steps: The third search keyword is extracted based on the target review item. The third search keyword is used as the third search input content. The search is performed at the selected search position in the tender document to obtain the third page range of the main text associated with the target review item. The search includes text matching.

9. A calibration-based PDF document page number range matching method according to claim 5, characterized in that, The step of combining the search output content obtained from the table of contents and / or text of the standardized tender document under at least one search method to obtain the effective review page range of the target review item includes: determining the union of the search output content obtained under at least one search method as the effective review page range of the target review item.

10. A calibration-based PDF document page number range matching system, characterized in that, include: The PDF document acquisition unit is used to acquire the PDF tender documents to be analyzed. The calibration unit is used to convert PDF bid documents into editable bid documents that retain the main text content style and structure, and to perform page number calibration and table of contents calibration on the editable bid documents to obtain standardized bid documents; A retrieval unit is used to obtain target review items for a tender document, and to search within a standardized tender document based on at least one retrieval method to obtain the page range of the main text associated with the target review item. The retrieval method has a corresponding method for generating retrieval input content, which uses the overall original content of the target review item as a data source to generate retrieval input content. The retrieval input content is then used to execute the corresponding retrieval method to obtain the retrieval output content under that retrieval method. The retrieval input content is used to characterize key information of the target review item. The retrieval output content includes the page range of the main text associated with the retrieval input content, i.e., the target review item. The page number range acquisition unit is used to combine the retrieval output content obtained from the table of contents and / or text of the standardized bidding document under at least one retrieval method to obtain the effective review page number range of the target review item.

Images