Financial professional report table data automatic updating method, device and equipment and storage medium

Abstract

The invention discloses an automatic updating method and device for financial professional report table data, equipment and a storage medium. The method comprises the following steps: calculating the similarity between a single old table and all new tables belonging to the same large chapter mapping pair; according to the calculated similarity value, a new table meeting the requirement is screened out according to the most similar matching rule, then whether the screened new table is a new table hit by other old tables or not is judged, if yes, replacement judgment is conducted on the new table according to the matching priority rule, and if not, replacement judgment is conducted on the new table; according to the method, all the old tables in the old file can be subjected to association matching, content replacement of the new and old tables is carried out in a unified mode, and the method has high generalization, can replace the content of the financial institution file table and has stability.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a method, device, equipment and storage medium for automatically updating financial professional report form data. Background Art

[0002] Updating financial institution documents means revising, updating or modifying the information in financial institution documents to reflect the latest information on operating conditions, financial status, risk factors, etc. This is a basic business practice to ensure that potential investors have access to the latest and accurate information to make informed investment decisions.

[0003] Updating financial institution files is often boring, repetitive, and time-consuming. There are many updated contents, but they are all patterned content replacements, replacing sentences and table contents of specified styles, and updating data, time, amount, proportion, etc. These sentences and table contents are similar in semantics, but there are differences in content, and this difference is uncertain. Therefore, there is a need for a system that can replace file content by judging text similarity through rules, automatically replace the content of financial institution files, replace the repetitive work of manual content replacement, and improve office efficiency.

[0004] The commonly used method is to create a one-to-one mapping relationship between the content of the file to be modified and the content of the template file, and update the content of the old file by updating the new template. However, this method has limitations. Every time a type of file is updated, a template file corresponding to the file of that type needs to be created, and the judgment logic and mapping relationship of the replacement need to be established. The content replaced by this method is very limited, the replacement range cannot be flexibly adjusted, and the template is not generalized.

[0005] Therefore, a new technology is needed that can use a generalized file template and a large model to calculate text similarity and replace file content. Summary of the invention

[0006] In view of this, the present invention provides a method, device, equipment and storage medium for automatically updating data in financial professional report forms. The present invention has strong generalization, can replace the content of financial institution file forms, and has stable capabilities.

[0007] A method for automatically updating data in a financial professional report table, specifically comprising the following steps:

[0008] S1, extract the table title, table chapter and table content in the new and old files respectively. The new file refers to the template file, and the old file refers to the file to be modified;

[0009] S2, normalize all extracted table titles, table chapters, and table contents;

[0010] S3, taking chapters as mapping scale, obtains the chapter mapping dictionary between the new and old files;

[0011] S4, calculating the similarity between a single old table and all remaining new tables that belong to the same large chapter mapping pair and that have been associated with the old table (and that have failed to replace the original associated old table) according to the chapter mapping dictionary;

[0012] S5, according to the calculated similarity value of the new and old tables, a new table that meets the matching conditions is screened according to the set most similar matching rule. If no new table that meets the most similar matching rule is found, it is considered that the current old table has failed to match and is not replaced. Step S4 is repeated to match the next old table. If a new table that meets the most similar matching rule is matched, step S6 is executed;

[0013] S6, establishing a preliminary replacement association between the new table that meets the most similar matching rule and the current old table, and then determining whether the new table that currently establishes the preliminary replacement association is a new table that has been hit by other old tables, if so, executing S7, otherwise the new table that currently establishes the preliminary replacement association is the new table that establishes the final replacement association with the current old table;

[0014] S7, according to the set matching priority rule, the new table currently establishing the preliminary replacement association is replaced and judged. If the replacement is successful, the new table currently establishing the preliminary replacement association is the new table that establishes the final replacement association with the current old table, and the replaced old table re-executes S4-S7. If the replacement fails, the current old table re-executes S4-S7;

[0015] S8, repeating steps S4-S7 to match the next old table until the associated matching of all old tables is completed;

[0016] S9, uniformly replace the contents of all old and new tables that implement replacement associations.

[0017] Preferably, the similarity between the new and old tables includes the title similarity S title , table content similarity S cell , and the table title and content similarity weighted score S title_cell .

[0018] Preferably, in step S4, the specific steps of calculating the similarity between a single old table and all remaining new tables belonging to the same large chapter mapping pair after removing all new tables that the old table has been associated with (and that failed to replace the original associated old table) are:

[0019] S41, obtaining the title subject keywords of the current old table, and matching the new table with the same title subject keywords from all new tables belonging to the same large chapter mapping pair according to the title subject keywords of the current old table. The old table of the old title that does not contain any subject keywords can only match the new table of the new title that does not contain any subject keywords. The matched new table is called a similar new table;

[0020] S42, respectively calculating the title similarity S between the current old table and each similar new table title Similarity S with table content cell ;

[0021] S43, the title similarity S between the current old table and each similar new table title Similarity S with table content cell Calculate the corresponding table title and content similarity weighted score S respectively title_cell ;

[0022] S44, calculate all the title similarity values ​​S title , table content similarity value S cell The weighted score S of the similarity between the table title and content title_cell Sort them in descending order according to the scores to get the title similarity sorting list {S title}、Table content similarity sorting list {S cell} and table title, content similarity weighted score sorting list {S title_cell}.

[0023] Preferably, in step S42, the title similarity S between the current old table and a single similar new table is calculated. title The specific steps are:

[0024] By comparing the set subject keyword table, extract the subject keywords of the current old table and similar new tables respectively;

[0025] Determine the title matching score S based on the similarity between the subject keywords of the current old table and the subject keywords of the similar new table. key ;

[0026] Obtain the table title semantic vectors of the current old table and similar new tables respectively;

[0027] Calculate the cosine similarity S of the new and old title vectors based on the semantic vectors of the table titles of the current old table and the similar new table vec , the cosine similarity S of the new and old title vectors vec As a reference value for title similarity;

[0028] According to the title matching score S keyThe cosine similarity S between the new and old title vectors vec Calculate the title similarity S between the current old table and the similar new table title .

[0029] Preferably, if the subject keywords of the current old table are completely consistent with the subject keywords of the similar new table, the title matching score S key =1;

[0030] If the subject keywords of the current old table are partially consistent with the subject keywords of the similar new table, the title matching score S key =0.5;

[0031] If the subject keywords of the current old table are inconsistent with the subject keywords of the similar new table, the title matching score S key =0.

[0032] Preferably, the cosine similarity S of the new and old title vectors is vec The calculation formula is:

[0033]

[0034] Among them, v old is the semantic vector of the title of the current old table, v new is the semantic vector of the title of the similar new table;

[0035] The title similarity S between the current old table and the similar new table title The calculation formula is:

[0036]

[0037] Among them, w k and w t All are set weight values.

[0038] Preferably, if the current old table has no title, the title similarity S between the current old table and the similar new table is calculated. title Set to 0.

[0039] Preferably, in step S42, the table content similarity S between the current old table and a single similar new table is calculated. cell The specific steps are:

[0040] Calculate the initial similarity S of the table contents of the two based on the normalized set of all cell contents of the current old table and the normalized set of all cell contents of the similar new table. cell_0 ;

[0041]

[0042] Among them, {O1,…,On} is the normalized set of all cell contents in the current old table, {N1,…,N n} is the normalized set of all cell contents in the similar new table, and LENGTH() takes the length of the set;

[0043] According to the relationship between the smallest chapter above the current old table and the smallest chapter above the similar new table, the initial similarity S of the table content is calculated. cell_0 Add the minimum chapter relationship weight to which the table belongs to obtain the table content similarity S between the current old table and the similar new table cell .

[0044] Preferably, if the minimum section texts of the current old table and the similar new table are not empty and the minimum section texts of the two tables are in a containment relationship, if S cell_0 ≥T contain_chapter , then S cell =S cell_0 +W contain_chapter ;

[0045] If the minimum chapter texts of the current old table and the similar new table are not empty and the minimum chapter texts of the two are equal, then S cell_0 =0, then S cell =W contain_chapter ;

[0046] If the minimum chapter texts of the current old table and the similar new table are not empty and the minimum chapter texts of the two are equal, then S cell_0 ≠0, then S cell =S cell_0 +W same_chapter ;

[0047] Among them, T contain_chapter is the set table content similarity threshold, W contain_chapter and W same_chapter These are the minimum chapter relationship weights of the set tables.

[0048] Preferably, if the difference between the minimum chapter text of the current old table and the similar new table is the specified keyword, S cell =-1.

[0049] Preferably, in step S5, the specific steps of screening and matching the new table that meets the conditions according to the calculated similarity value of the new and old tables according to the set most similar matching rule are:

[0050] S51, determine the title similarity ranking list {S title The S with the largest score among title_max Does it satisfy S? title_max >Ttitle_top If satisfied, then the similar new table corresponding to the title similarity value establishes a replacement association with the current old table, otherwise, execute step S52;

[0051] S52, determine the title similarity ranking list {S title The S with the largest score among title_max Whether it satisfies 0 title_max <T title_min If the condition is satisfied, the current old table is deemed to have failed to match and will not be replaced. Step S4 is repeated to match the next old table. Otherwise, step S53 is executed.

[0052] S53, determine the table title, content similarity weighted score sorting list {S title_cell The S with the largest score among title_cell_max Does it satisfy S? title_cel_maxl >T title_cell If the conditions are met, then the similar new table corresponding to the table title and content similarity weighted score is associated with the current old table for replacement, otherwise, step S54 is executed;

[0053] S54, determine the table content similarity ranking list {S cell The S with the largest score among cell_max Does it satisfy S? cell_max <T cell_min If the condition is satisfied, the current old table is deemed to have failed to match and will not be replaced. Step S4 is repeated to match the next old table. Otherwise, step S55 is executed.

[0054] S55, determine the table content similarity ranking list {S cell The S with the largest score among cell_max Does it satisfy S? cell_max >T cell If the condition is satisfied, then the similar new table corresponding to the table content similarity value is associated with the current old table for replacement, otherwise, step S56 is executed;

[0055] S56, determine the title similarity ranking list {S title The S with the largest score among title_max Does it satisfy S? title_max >T title If the conditions are met, the similar new table corresponding to the title similarity value is associated with the current old table for replacement. Otherwise, the current old table is deemed to have failed to match and is not replaced. Step S4 is repeated to match the next old table.

[0056] Among them, T title_top , T title_min and T title are the set title similarity thresholds, T title_cell ​is the weighted threshold for table title and content similarity, T cell_min and T cell These are the set table content similarity thresholds.

[0057] Preferably, the T title_min <T title <T title_top , T cell_min <T cell .

[0058] Preferably, the title similarity threshold T title_min =0.5, T title =0.829, T title_top =0.98, table content similarity threshold T cell_min =0.04, T cell =0.082, table title and content similarity weighted threshold T title_cell =0.7.

[0059] Preferably, in step S7, the specific steps of performing a replacement judgment on the new table currently establishing a preliminary replacement association according to the set matching priority rule, and determining the new table to establish a final replacement association with the current old table according to the replacement judgment result are:

[0060] S71, setting the current old table as the first old table, and setting the other old table corresponding to the new table hit by the other old tables as the second old table;

[0061] The final matching status of the first old table and the second old table when performing the most similar matching rule is obtained respectively, and the final matching similarity score S of the first old table is obtained. replace _new, the final matching similarity score S of the second oldest table replace _old;

[0062] S72, determine the priority status code C of the first old table according to the priority status code determination rule re _new and second oldest table priority status code C re _old;

[0063] S73, judging the matching priority of the first old table and the second old table according to the final matching similarity scores and priority status codes of the first old table and the second old table, if the matching priority of the first old table is higher than the matching priority of the second old table, using the new table currently establishing the preliminary replacement association as the new table establishing the final replacement association with the first old table, and canceling the replacement association between the new table and the second old table, and repeatedly executing step S4 until the second old table matches the new table or does not match the new table;

[0064] If the matching priority of the first old table is lower than the matching priority of the second old table, the replacement association between the first old table and the new table currently establishing the preliminary replacement association is cancelled, the replacement association between the second old table and the new table is kept unchanged, and step S4 is performed until the first old table matches the new new table or does not match the new new table;

[0065] If the first old table or the second old table does not match the new new table, it is considered a match failure and will not be replaced.

[0066] Preferably, the priority status code determination rule is:

[0067] If the final matching status of the old table when executing the most similar matching rule is high title similarity, which meets the requirements for establishing a replacement association, then C re =3;

[0068] If the final matching status of the old table when executing the most similar matching rule is that the weighted score of the table title and content similarity meets the requirements for establishing a replacement association, then C re =2;

[0069] If the final matching status of the old table when executing the most similar matching rule is that the table content similarity meets the requirements for establishing a replacement association, then C re =1;

[0070] If the final matching status of the old table when executing the most similar matching rule is that the table title similarity meets the requirements for establishing a replacement association, then C re =0;

[0071] If the final matching status of the old table when executing the most similar matching rule is matching failure, then C re =-1.

[0072] Preferably, the specific steps of determining the matching priority between the first old table and the second old table according to the final matching similarity scores and priority status codes of the first old table and the second old table in step S73 are:

[0073] Determine whether the final matching similarity score and priority status code of the first old table and the second old table meet condition C re _new>C re _old, or C re _new=C re _old and S replace _new>S replace _old, if the condition is met, it means that the matching priority of the first old table is higher than the matching priority of the second old table, otherwise the matching priority of the first old table is lower than the matching priority of the second old table.

[0074] Preferably, the final matching similarity scores of the first old table and the second old table are both similarity scores corresponding to the final matching states achieved when the most similar matching rule is executed.

[0075] A device for automatically updating financial professional report table data using the above method comprises:

[0076] A file processing module is used to extract table titles, table chapters and table contents from new and old files, and normalize all extracted table titles, table chapters and table contents;

[0077] A document chapter mapping module is used to obtain a chapter mapping dictionary between new and old files based on the chapter as the mapping scale;

[0078] The new-old table matching module is used to calculate the new-old table similarity between a single old table and all the remaining new tables that belong to the same large chapter mapping pair after removing all new tables that have been associated with the old table (and failed to replace the original associated old table) according to the chapter mapping dictionary, match the new table that meets the most similar matching rule according to the calculated new-old table similarity value, establish a preliminary replacement association between the new table that meets the most similar matching rule and the current old table, and determine whether the new table that currently establishes the preliminary replacement association is a new table that has been hit by other old tables. If so, perform a replacement judgment on the new table that currently establishes the preliminary replacement association according to the set matching priority rule;

[0079] The new and old table content replacement module is used to uniformly replace the contents of all new and old tables that implement replacement associations.

[0080] A computer device comprises: a processor and a memory, wherein the memory stores at least one program, and the at least one program is loaded and executed by the processor to implement the method described above.

[0081] A computer-readable storage medium stores at least one program, and the at least one program is loaded and executed by a processor to implement the method described above.

[0082] A computer program product, the computer program product includes at least one program segment, and the at least one program segment is stored in a computer-readable storage medium; a processor of a communication device reads the at least one program segment from the computer-readable storage medium, and the processor executes the at least one program segment, so that the communication device executes the method described above.

[0083] The beneficial effects of the present invention are:

[0084] The present invention calculates the similarity between a single old table and all new tables belonging to the same large chapter mapping pair, and selects new tables that meet the requirements according to the most similar matching rule based on the calculated similarity value, and then determines whether the selected new table is a new table that has been hit by other old tables. If so, the new table is replaced according to the matching priority rule to obtain a new table that establishes a final replacement association with the current old table. According to this method, all old tables in the old file can be associated and matched, and the content of the new and old tables can be uniformly replaced. This method has strong generalization, and can replace the content of the table of financial institution files with stable capabilities. BRIEF DESCRIPTION OF THE DRAWINGS

[0085] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0086] Figure 1 It is a flow chart of automatic updating of financial professional report table data of the present invention.

[0087] Figure 2 Is the matching subprocess for a single old table. DETAILED DESCRIPTION

[0088] In order to make the purpose, technical scheme and advantages of the present invention clearer, the present invention is described below by the specific embodiments shown in the accompanying drawings. However, it should be understood that these descriptions are only exemplary and are not intended to limit the scope of the present invention. In addition, in the following description, the description of well-known structures and technologies is omitted to avoid unnecessary confusion of the concept of the present invention.

[0089] The terms used in this disclosure are for the purpose of describing specific embodiments only and are not intended to limit the disclosure. The singular forms of "a", "said" and "the" used in this disclosure and the appended claims are also intended to include plural forms unless the context clearly indicates otherwise. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more associated listed items.

[0090] It should be understood that although the terms first and second may be used in the present disclosure to describe various information, these information should not be limited to these terms and cannot be understood as indicating or implying relative importance. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of the present disclosure, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0091] Updating financial institution files refers to revising, updating or modifying the information in financial institution files to reflect the latest information on operating conditions, financial conditions, risk factors, etc. The work of updating financial institution files mainly includes replacing sentences and table contents of specified styles, and updating data, time, amount, proportion, etc. The existing file content update method of hollowing out and filling has poor flexibility and cannot be applied to various types of files. When judging the similarity of tables, there may be equal content but differences in table attributes. For example, the contents of two tables are almost the same, but the titles and chapters of the two tables may be different. This situation often occurs in financial institution files, which brings obstacles to the determination of table similarity. The present invention provides a method for automatically updating financial professional report table data. The method has strong generalization and versatility, can be applied to various types of file updates, replaces the table content of financial institution files, and has stability and robustness.

[0092] The method for automatically updating financial professional report table data provided by the present invention comprises the following steps:

[0093] S1, extract the table title, table chapter and table content in the new and old files respectively. The new file refers to the template file, and the old file refers to the file to be modified;

[0094] S2, normalize all extracted table titles, table chapters, and table contents;

[0095] S3, taking chapters as mapping scale, obtains the chapter mapping dictionary between the new and old files;

[0096] S4, calculating the new-old table similarity between a single old table and all remaining new tables belonging to the same large chapter mapping pair after removing all new tables that have been associated with the old table (and failed to replace the original associated old table) according to the chapter mapping dictionary;

[0097] S5. Filter and match new tables that meet the conditions according to the calculated similarity values between the new and old tables and the set most similar matching rule. If no new table that meets the most similar matching rule is found, it is considered that the current old table fails to match and will not be replaced. Repeat step S4 to match the next old table. If a new table that meets the most similar matching rule is matched, execute step S6.

[0098] S6. Establish a preliminary replacement association between the new table that meets the most similar matching rule and the current old table. Then, determine whether the new table with the current preliminary replacement association has been hit by other old tables. If so, execute S7. Otherwise, the new table with the current preliminary replacement association is the new table that establishes the final replacement association with the current old table.

[0099] S7. Perform a replacement judgment on the new table with the current preliminary replacement association according to the set matching priority rule. If the replacement is successful, the new table with the current preliminary replacement association is the new table that establishes the final replacement association with the current old table, and the replaced old table repeats steps S4 - S7. If the replacement fails, the current old table repeats steps S4 - S7.

[0100] S8. Repeat steps S4 - S7 to match the next old table until all old tables are associated and matched.

[0101] S9. Uniformly replace the content of all new and old tables that have achieved replacement associations.

[0102] To better understand the technical solution of the present invention, the present invention will be described in detail below with reference to the accompanying drawings.

[0103] The automatic update method for financial professional report form data provided by the present invention specifically includes the following steps:

[0104] S1. Parse the new and old files respectively, and extract the table titles, table chapters to which the tables belong, and table contents in the new and old files.

[0105] For the table title: Extract the first-line text content located above the table, without punctuation at the end (except for "Unit: 10,000 yuan", "Unit: 100 million yuan", or "Unit: yuan", etc.), and without a numerical serial number.

[0106] For the table chapter to which the table belongs: Extract the text content corresponding to each level that starts with a numerical serial number and has different numerical serial numbers and no punctuation at the end when going forward from the start of the table. The numerical serial number format is usually: I., 1, (1), ①. The chapter is generally a multi-level structure.

[0107] For the table content: Extract all the text content in all cells of the table.

[0108] S2, normalize all extracted table titles, table chapters and table contents, such as modifying all time expressions to "within the reporting period" and removing unnecessary punctuation, spaces and illegal characters from all texts.

[0109] Specifically, the normalization process includes the following steps:

[0110] Normalize the table titles: remove the initial expressions such as "Table:", "Table", "Table 1-1", "1," etc.; remove the ending word "Table".

[0111] Normalize the chapters to which the table belongs: remove numerical serial numbers beginning with one, 1, (1), ①, etc.; remove similar expressions beginning with "Chapter 1", "Section 1", etc., and retain the text of the latter chapters. For example, in "Chapter 6: Main Financial Status of the Issuer", remove the text "Chapter 1" and the space behind it, and retain the text "Main Financial Status of the Issuer".

[0112] Normalize the table contents: All contents in the table except those that need to be hidden include: keyword texts such as "item", "total", "amount", "proportion", "ratio", "category", "subtotal", and "serial number".

[0113] Use the normalized text of the table title, table chapter, and table content to match the new table in the new file according to the rules of the subsequent steps.

[0114] S3, with chapters as the mapping scale, uses the GPT-4 large multimodal model to obtain the chapter mapping dictionary between new and old files.

[0115] The chapter mapping dictionary refers to the chapter mapping relationship composed of the highest-level chapters to which the tables in the new and old files belong.

[0116] The specific steps for obtaining the chapter mapping dictionary between the new and old files are: input the maximum level chapter of the old file and the multi-level chapter information of the new file into the GPT-4 large multimodal model respectively, and the GPT-4 large multimodal model determines the maximum level chapter of the new file that is most similar to the maximum level chapter of each old file based on semantics.

[0117] For example, if the old Form A in the old document belongs to the chapters "Chapter 1. Issuer's main financial situation", "II. Issuer's financial statements for the latest three years and one period", "(I) Consolidated statements for the latest three years and one period",

[0118] The new Form A in the new document belongs to the following chapters: "Chapter VI: Issuer's Main Financial Status", "II. Issuer's Financial Statements for the Last Three Years and One Period", and "(II) Parent Company Statements for the Last Three Years and One Period".

[0119] The chapter mapping relationship between the old form A and the new form a is "Issuer's main financial status" corresponds to "Issuer's main financial status" (normalized large chapter text mapping). Chapter "Chapter 1 Issuer's main financial status" corresponds to "Chapter 6 Issuer's main financial status", which is the largest chapter corresponding to the old and new forms.

[0120] S4, according to the chapter mapping dictionary, calculating the new-old table similarity between a single old table and all remaining new tables belonging to the same large chapter after removing all new tables that have been associated with the old table (and failed to replace the original associated old table).

[0121] Specifically, S41, obtain the title body keyword of the current old table, and match the new table with the same title body keyword from all the remaining new tables that belong to the same large chapter mapping pair after removing all new tables that the old table has been associated with (and failed to replace the original associated old table). For the old table without the title body keyword, only the new table without the title body keyword can be matched. The matched new table is called a similar new table;

[0122] The title subject keyword is used to limit the matching range of new and old tables (some tables may not have titles) that both have titles. For example, an old table with an old title containing subject keywords can only match a new title containing subject keywords, and an old table with an old title that does not contain any subject keywords can only match a new table with a new title that does not contain any subject keywords. For example, an old table with a title containing the title subject keyword "parent company" can only match a new table with a title containing the subject keyword "parent company".

[0123] Assuming that the old Form A in the old document belongs to the major chapter "Chapter 1 The Issuer's Main Financial Status", and the title "The Issuer's Parent Company's Asset and Liability Structure Table from the end of 2020 to 2022 and March 2023" contains the subject keyword "parent company", according to the title subject keyword "parent company", a new Form a with the same title "The Issuer's Parent Company's Asset and Liability Structure Table from the end of 2021 to 2023 and March 2024" containing the subject keyword "parent company" is matched from the "Chapter 6 The Issuer's Main Financial Status" chapter of the new document.

[0124] Assuming that the old Form A in the old document belongs to the major chapter "Chapter 1 The Issuer's Main Financial Status" and the title "Detailed Composition Table of the Issuer's Period Expenses from the End of 2020 to the End of 2022 and the End of March 2023" does not contain any subject keywords, the new document's "Chapter 6 The Issuer's Main Financial Status" chapter matches the new Form a with the title "Detailed Composition Table of the Issuer's Period Expenses from the End of 2021 to the End of 2023 and the End of March 2024" which also does not contain any subject keywords.

[0125] Assume that an old table A in an old file has three new tables a, b, and c matching it. The three matched new tables a, b, and c are called similar new tables.

[0126] S42, respectively calculating the title similarity S between the current old table and each similar new table title :

[0127] (1) Extract the subject keywords from the titles of the current old forms and similar new forms by comparing them with the set subject keyword table. For example, the title “Profitability Structure Table of the Issuer from the End of 2021 to 2023 and the End of March 2024” contains the subject keyword “profitability”;

[0128] Zero or more subject keywords can appear in a table title;

[0129] (2) Determine the title matching score S based on the similarity between the subject keywords of the current old table and the subject keywords of the similar new table. key (also called subject keyword hit weight score), that is, subject keywords are extracted from both new and old titles at the same time, and the title matching score is increased according to the subject keyword hit situation;

[0130] If the subject keywords of the current old table are exactly the same as the subject keywords of the similar new table, the title matching score S key =1, for example, the subject keywords of the current old table and the similar new table are both "cash flow", "profitability", and "operating efficiency", then the subject keywords of the two tables are exactly the same, S key =1;

[0131] If the subject keywords of the current old table are partially consistent with the subject keywords of the similar new table, the title matching score S key = 0.5. If the subject keyword of the current old table is "cash flow", and the subject keywords of the similar new table are "cash flow" and "profitability", the subject keywords of the two tables are partially consistent, S key =0.5;

[0132] If the subject keywords of the current old table are inconsistent with the subject keywords of the similar new table, the title matching score Skey = 0, if the subject keyword of the current old table is "cash flow" and the subject keyword of the similar new table is "profitability", the subject keywords of the two are inconsistent, S key =0.

[0133] (3) Obtain the table title semantic vectors of the current old table and similar new tables respectively through the BGE semantic vector model;

[0134] V new = {v new_1 ,…,v new_n}

[0135] V old = {v old_1 ,…,v old_n}

[0136] V new The set of semantic vectors of table titles representing similar new tables, V old A collection of semantic vectors representing table titles of old tables.

[0137] (4) Calculate the cosine similarity S of the new and old title vectors based on the semantic vectors of the table titles of the current old table and the similar new table. vec , the cosine similarity S of the new and old title vectors vec As the title similarity reference value, the title similarity matrix is ​​obtained;

[0138] Cosine similarity S of new and old title vectors vec The calculation formula is:

[0139]

[0140] Among them, v old is the semantic vector of the title of the current old table, v new is the semantic vector of the title of the similar new table.

[0141] (5) According to the title matching score S key The cosine similarity S between the new and old title vectors vec Calculate the title similarity S between the current old table and the similar new table title .

[0142] The title similarity S between the current old table and the similar new table title The calculation formula is:

[0143]

[0144] Among them, w k and w t All are set weight values.

[0145] w k and w t The specific value of is set according to experience. In this embodiment, the best setting is w k =0.4, w t =0.6.

[0146] If the old table has no title, then the title similarity S between it and all similar new tables is calculated. title Set all to 0.

[0147] For example, according to the above calculation method, the title similarity values ​​S between the old table A and the similar new tables a, b, and c can be calculated respectively. title_Aa , S title_Ab , S title_Ac .

[0148] Calculate the table content similarity S between the current old table and each similar new table cell :

[0149] (1) Calculate the initial similarity S of the table contents of the two tables based on the normalized set of all cell contents of the current old table and the normalized set of all cell contents of the similar new table. cell_0 ;

[0150]

[0151] Among them, {O1,…,O n} is the normalized set of all cell contents in the current old table, {N1,…,N n} is the normalized set of all cell contents in the similar new table, and LENGTH() takes the length of the set;

[0152] (2) According to the relationship between the smallest chapter above the current old table and the smallest chapter above the similar new table, the initial similarity S of the table content is calculated. cell_0 Add the minimum chapter relationship weight to which the table belongs to obtain the table content similarity S between the current old table and the similar new table cell ,Right now:

[0153] If the text of the smallest section of the current old table and the similar new table is not empty and the text of the smallest section of the two is an inclusive relationship (for example, the text of the smallest section of the current old table is "Profitability", and the text of the smallest section of the similar new table is "Profitability Analysis", then the text of the smallest section of the two is not empty and is an inclusive relationship), if S cell_0 ≥T contain_chapter , then S cell =S cell_0 +W contain_chapter ;

[0154] If the text of the smallest section of the current old table and the similar new table is not empty and the text of the smallest section of the two is equal (for example, the text of the smallest section of the current old table is "Profitability", and the text of the smallest section of the similar new table is also "Profitability", then the text of the smallest section of the two is not empty and is equal), if S cell_0 =0, then S cell =W contain_chapter ;

[0155] If the minimum chapter texts of the current old table and the similar new table are not empty and the minimum chapter texts of the two are equal, then S cell_0 ≠0, then S cell =S cell_0 +W same_chapter ;

[0156] Among them, T contain_chapter is the set table content similarity threshold, W contain_chapter and W same_chapter The minimum chapter relationship weight value of the table is set, T contain_chapter , W contain_chapter and W same_chapter The specific value of is set according to experience. In this embodiment, the best setting is T contain_chapter =0.019,W contain_chapter =0.31,W same_chapter =0.35.

[0157] Preferably, if the difference between the minimum chapter text of the current old table and the similar new table is the specified keyword, directly set S cell =-1.

[0158] For example, according to the above calculation method, the title similarity values ​​S between the old table A and the similar new tables a, b, and c can be calculated respectively. cell_Aa , S cell_Ab , S cell_Ac .

[0159] S43, the title similarity S between the current old table and each similar new table title Similarity S with table content cell Calculate the corresponding table title and content similarity weighted score S respectively title_cell ;

[0160] Table title, content similarity weighted score S title_cell The calculation formula is:

[0161]

[0162] w tt and w tc is the weight value, w tt and wt c The specific value of is set according to experience. In this embodiment, the best setting is w tt =0.6, w tc =0.4.

[0163] For example, the weighted scores S of the table title and content similarity between the old table A and similar new tables a, b, and c can be calculated respectively. title_cell_Aa , S title_cell_Ab , S title_cell_Ac .

[0164] S44, calculate all the title similarity values ​​S title , table content similarity value S cell The weighted score S of the similarity between the table title and content title_cell Sort them in descending order according to the scores to get the title similarity sorting list {S title}、Table content similarity sorting list {S cell} and table title, content similarity weighted score sorting list {S title_cell}.

[0165] For example, the title similarity value S calculated in step S42 above title_Aa , S title_Ab , S title_Ac (Assume S title_Aa >S title_Ab >S title_Ac ) Sort by score from large to small to get a title similarity sorting list {S title};

[0166] The title similarity value S calculated in step S42 above cell_Aa , S cell_Ab , S cell_Ac (Assume S cell_Ab >S cell_Aa >S cell_Ac ) Sort by scores from large to small to get a table content similarity sorting list {S cell};

[0167] The table title and content similarity weighted score S calculated in step S43 above is title_cell_Aa , S title_cell_Ab , S title_cell_Ac (Assume S title_cell_Ac >S title_cell_Aa >S title_cell_Ab) Sort by scores from large to small to get a weighted score sorting list of table titles and content similarity {S title_cell}.

[0168] S5, based on the calculated similarity values ​​of the new and old tables, filter and match new tables that meet the conditions according to the set most similar matching rules. If no new table that meets the most similar matching rules is found, it is deemed that the current old table has failed to match and is not replaced. Repeat step S4 to match the next old table. If a new table that meets the most similar matching rules is matched, execute step S6.

[0169] The specific steps of filtering and matching new tables that meet the conditions according to the set most similar matching rules based on the calculated similarity values ​​of the new and old tables are as follows:

[0170] S51, determine the title similarity ranking list {S title The S with the largest score among title_max Does it satisfy S? title_max >T title_top If it is satisfied, then the similar new table corresponding to the title similarity value establishes a preliminary replacement association with the current old table (for example, {S title} title_Aa The score is the largest, judging S title_Aa Does it satisfy S? title_Aa >T title_top , if satisfied, then S title_Aa The corresponding new table a establishes a preliminary replacement association with the current old table), otherwise, execute step S52;

[0171] S52, determine the title similarity ranking list {S title The S with the largest score among title_max Whether it satisfies 0 title_max <T title_min (For example, judging S title_Aa Whether it satisfies 0 title_Aa <T title_min ), if satisfied, the current old table is deemed to have failed to match, and will not be replaced, and step S4 will be repeated to match the next old table, otherwise, step S53 will be executed;

[0172] S53, determine the table title, content similarity weighted score sorting list {S title_cell The S with the largest score among title_cell_max Does it satisfy S? title_cell_max >T title_cell If the similarity is satisfied, the new table corresponding to the table title and content similarity weighted score is initially associated with the current old table (for example, {S title_cell} title_cell_Ac The score is the largest, judging S title_cell_Ac ​​Does it satisfy S? title_cell_Ac >T title_cell , if satisfied, then this step will S title_cell_Ac The corresponding new table c establishes a preliminary replacement association with the current old table), otherwise, execute step S54;

[0173] S54, determine the table content similarity ranking list {S cell The S with the largest score among cell_max Does it satisfy S? cell_max <T cell_min (For example {S cell} cell_Ab The score is the largest, judging S cell_Ab Does it satisfy S? cell_Ab <T cell_min ), if satisfied, the current old table is deemed to have failed to match, and will not be replaced, and step S4 will be repeated to match the next old table, otherwise, step S55 will be executed;

[0174] S55, determine the table content similarity ranking list {S cell The S with the largest score among cell_max Does it satisfy S? cell_max >T cell If it is satisfied, then the similar new table corresponding to the table content similarity value establishes a preliminary replacement association with the current old table (for example, {S cell} cell_Ab The score is the largest, judging S cell_Ab Does it satisfy S? cell_Ab >T cell , if satisfied, then this step will S cell_Ab The corresponding new table b establishes a preliminary replacement association with the current old table), otherwise, execute step S56;

[0175] S56, determine the title similarity ranking list {S title The S with the largest score among title_max Does it satisfy S? title_max >T title If it is satisfied, then the similar new table corresponding to the title similarity value establishes a preliminary replacement association with the current old table (for example, {S title} title_Aa The score is the largest, judging S title_Aa Does it satisfy S? title_Aa >T title , if satisfied, then this step will S title_Aa The corresponding new table a establishes a preliminary replacement association with the current old table), otherwise the current old table is deemed to have failed to match and is not replaced, and step S4 is repeated to match the next old table;

[0176] Among them, T title_top, T title_min and T title are the set title similarity thresholds, T title_cell is the weighted threshold for table title and content similarity, T cell_min and T cell are all set table content similarity thresholds, and the T title_min <T title <T title_top , T cell_min <T cell .

[0177] T title_top , T title_min 、、T title_cell , T cell_min and T cell The specific value of is set based on experience. In this embodiment, T title_to =0.98, T title_cell =0.7, T cell =0.082, T title =0.829, T cell_min =0.04, T title_min =0.5.

[0178] Preferably, if the minimum text difference between the new and old tables is a specified keyword, no replacement will be made.

[0179] Preferably, if the specified keyword pair appears in the new and old table titles respectively, they will not be replaced.

[0180] Still taking the above example, when the old table A is matched with similar new tables a, b, and c according to the set most similar matching rules, if the program is terminated when step S51 is executed, a preliminary replacement association will be established between the new table a that meets the criteria of this step and the current old table; if the criteria of S51 and S52 are not met, the program will not be terminated until step S53 is executed, and a preliminary replacement association will be established between the new table c that meets the criteria of S53 and the current old table; if the criteria of S51-S55 are not met, the program will not be terminated until the last step S56 is executed, and a preliminary replacement association will be established between the new table a that meets the criteria of S56 and the current old table.

[0181] S6, establishing a preliminary replacement association between the new table that meets the most similar matching rule and the current old table (still taking the above example as an example, assuming that the old table A and the similar new tables a, b, c in step S5 are executed according to the set most similar matching rule until the program is terminated at S53, then the new table c is established with the current old table).

[0182] Then, it is determined whether the new table (new table c) currently establishing a preliminary replacement association is a new table that has been hit by other old tables (i.e., whether there is an old table B that has established a preliminary replacement association with the new table c). If so, S7 is executed; otherwise, the new table currently establishing a preliminary replacement association is the new table that has established a final replacement association with the current old table.

[0183] S7, performing a replacement judgment on the new table that is currently establishing a preliminary replacement association according to the set matching priority rule, and determining a new table that is currently establishing a final replacement association with the current old table according to the replacement judgment result.

[0184] The specific steps for performing replacement judgment on the new table currently establishing a preliminary replacement association according to the set matching priority rules are as follows:

[0185] S71, let the current old table (old table A) be the first old table, and let the other old table (old table B) corresponding to the new table hit by the other old table be the second old table;

[0186] The final matching status of the first old table (old table A) and the second old table (old table B) when performing the most similar matching rule is obtained respectively, and the final matching similarity score S of the first old table is obtained. replace _new, the final matching similarity score S of the second oldest table replace _old.

[0187] The final matching similarity scores of the first old table and the second old table are both similarity scores corresponding to the final matching states when the most similar matching rule is executed. Taking the above example, since the old table A and the similar new tables a, b, and c in step S5 are executed according to the set most similar matching rule until S53, S53 is the final matching state of the old table A. The weighted scores S53 of the table titles and contents similarity that meet the criteria of S53 are title_cell_Ac As the final matching similarity score S of the first old table replace _new, that is, S replace _new=S title_cell_Ac ; Similarly, the final matching similarity score S of the old table B can be obtained replace _old.

[0188] S72, determine the priority status code C of the first old table according to the priority status code determination rule re _new and second oldest table priority status code C re _old.

[0189] The priority status code determination rule is:

[0190] If the final matching status of the old table when executing the most similar matching rule is high title similarity, which meets the requirements for establishing a preliminary replacement association, then C re =3;

[0191] If the final matching status of the old table when executing the most similar matching rule is that the weighted score of the table title and content similarity meets the requirements for establishing a preliminary replacement association, then C re =2;

[0192] If the final matching status of the old table when executing the most similar matching rule is that the table content similarity meets the requirements for establishing a preliminary replacement association, then C re =1;

[0193] If the final matching status of the old table when executing the most similar matching rule is that the table title similarity meets the requirements for establishing a preliminary replacement association, then C re =0;

[0194] If the final matching status of the old table when executing the most similar matching rule is matching failure, then C re =-1.

[0195] Still taking the above example, since the final matching status of the old table A when matching with the similar new tables a, b, and c according to the set most similar matching rule is that the weighted scores of the table title and content similarity meet the requirements for establishing a preliminary replacement association, the priority status code of the old table A is C re _new=2; Similarly, the priority status code C of the old table B can be obtained re _old.

[0196] S73, determining the matching priority between the first old table and the second old table according to the final matching similarity scores and priority status codes of the first old table and the second old table;

[0197] Specifically, it is determined whether the final matching similarity scores and priority status codes of the first old table and the second old table meet the condition C. re _new>C re _old, or C re _new=C re _old and S replace _new>S replace _old, if this condition is met, it means that the matching priority of the first old table is higher than the matching priority of the second old table, otherwise the matching priority of the first old table is lower than the matching priority of the second old table;

[0198] If the matching priority of the first old table is higher than the matching priority of the second old table, the new table currently establishing the preliminary replacement association is used as the new table establishing the final replacement association with the first old table, and the replacement association between the new table and the second old table is cancelled, and the second old table repeats step S4 until the second old table matches the new new table or does not match the new new table (the second old table repeats step S4 to re-search for a new new table that matches it in a new round of matching, and the obtained title similarity ranking list {S title}、Table content similarity sorting list {S cell}, and the table title, content similarity weighted score sorting list {S title_cell {S} are all similarity score sets of all remaining new tables after removing all new tables that have been associated with the second oldest table (and failed to replace the original associated old table). That is, new tables that have been associated with the second oldest table but failed to be associated need to be eliminated in this round of matching, and their similarity scores are not placed in the sorting list. title}, {S cell}, {S title_cell}middle);

[0199] If the matching priority of the first old table is lower than the matching priority of the second old table, the replacement association between the first old table and the new table currently establishing the preliminary replacement association is cancelled, and the replacement association between the second old table and the new table remains unchanged. Meanwhile, the first old table executes step S4 until the first old table matches the new new table or does not match the new new table (similarly, the first old table repeats step S4 to re-search for a new new table that matches it in a new round of matching. The title similarity ranking list {S title}、Table content similarity sorting list {S cell}, and the table title, content similarity weighted score sorting list {S title_cell {S} are all similarity score sets of all remaining new tables after removing all new tables that have been associated with the first old table (and failed to replace the original associated old table), that is, new tables that have been associated with the first old table but failed to be associated need to be eliminated in this round of matching, and their similarity scores are not placed in the sorting list {S title}, {S cell}, {S title_cell}middle);

[0200] If the first old table or the second old table does not match the new new table, it is considered a match failure and will not be replaced.

[0201] Still taking the above example, assuming that it is determined that the matching priority of the old table A is higher than the matching priority of the old table B (that is, the similarity between the old table A and the new table C is higher than that between the old table A and the old table B), then the final replacement association is established between the new table C and the old table A, and the preliminary replacement association between the new table C and the old table B is cancelled, and then the old table B repeats step 4 to find a new table that matches it again. If the new table that matches the old table B is not found by repeating step 4, it is deemed that the old table B fails to match, and the old table B is not replaced;

[0202] On the contrary, assuming that it is determined that the matching priority of old table B is higher than the matching priority of old table A (that is, old table B is more similar to new table c than old table A), then the initial replacement association established between new table c and old table A is canceled, and the replacement association between old table B and new table c remains unchanged. Then old table A executes step S4 to find a new matching new table again. If step 4 is repeated and no new new table matching old table A is found, it is deemed that old table A fails to match and old table A is not replaced.

[0203] S8, repeat steps S4-S7 for the next old table until the association matching of all old tables is completed;

[0204] S9, uniformly replace the contents of all new and old tables that have achieved replacement association, that is, all new tables that have established replacement associations are updated and replaced with their corresponding old tables (that is, the entire table is replaced, including the table title, table content, table cell content, etc. of the entire new table). The old tables that failed to match and did not establish replacement associations do not need to be processed and replaced.

[0205] The present invention also provides a device for automatically updating financial professional report table data using the above-mentioned method, including a file processing module, a document chapter mapping module, a new and old table matching module and a new and old table content replacement module.

[0206] The file processing module is used to extract table titles, table chapters and table contents from new and old files, and normalize all extracted table titles, table chapters and table contents;

[0207] The document chapter mapping module is used to obtain a chapter mapping dictionary between new and old files based on the chapter as the mapping scale;

[0208] The new-old table matching module is used to calculate the new-old table similarity between a single old table and all new tables belonging to the same large chapter according to the chapter mapping dictionary, match a new table that meets the most similar matching rule according to the calculated new-old table similarity value, establish a preliminary replacement association between the new table that meets the most similar matching rule and the current old table, and then determine whether the new table that currently establishes the preliminary replacement association is a new table that has been hit by other old tables. If so, perform a replacement judgment on the new table that currently establishes the preliminary replacement association according to the set matching priority rule;

[0209] The new and old table content replacement module is used to uniformly replace the contents of all new and old tables that implement replacement associations.

[0210] The present invention also provides a computer device, which includes: a processor and a memory, wherein at least one program is stored in the memory, and the at least one program is loaded and executed by the processor to implement the above-mentioned method.

[0211] The present invention also provides a computer-readable storage medium, wherein at least one program is stored in the computer-readable storage medium, and the at least one program is loaded and executed by a processor to implement the above-mentioned method.

[0212] The present invention also provides a computer program product, which includes at least one program segment, and the at least one program segment is stored in a computer-readable storage medium; a processor of a communication device reads the at least one program segment from the computer-readable storage medium, and the processor executes the at least one program segment, so that the communication device executes the method described above.

[0213] It should be clear that the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

Claims

1. A method for automatically updating data in a financial professional report table, characterized in that: The specific steps include: S1, extract the table title, table chapter and table content in the new and old files respectively. The new file refers to the template file, and the old file refers to the file to be modified; S2, normalize all extracted table titles, table chapters, and table contents; S3, taking chapters as mapping scale, obtains the chapter mapping dictionary between the new and old files; S4, calculating, based on the chapter mapping dictionary, the new-old table similarity between a single old table and all remaining new tables that belong to the same large chapter mapping pair after removing all new tables that the old table has been associated with but failed to be associated with; S5, according to the calculated similarity value of the new and old tables, a new table that meets the matching conditions is screened according to the set most similar matching rule. If no new table that meets the most similar matching rule is found, it is considered that the current old table has failed to match and is not replaced. Step S4 is repeated to match the next old table. If a new table that meets the most similar matching rule is matched, step S6 is executed; S6, establishing a preliminary replacement association between the new table that meets the most similar matching rule and the current old table, and then determining whether the new table that currently establishes the preliminary replacement association is a new table that has been finally replaced by other old tables. If so, executing S7; otherwise, the new table that currently establishes the preliminary replacement association is the new table that has been finally replaced by the current old table. S7, performing a replacement judgment on the new table that is currently establishing a preliminary replacement association according to the set matching priority rule, and determining a new table that is currently establishing a final replacement association with the current old table according to the replacement judgment result; S8, repeating steps S4-S7 to match the next old table until the associated matching of all old tables is completed; S9, uniformly replace the contents of all old and new tables that implement replacement associations.

2. The method for automatically updating financial professional report table data according to claim 1, characterized in that: The similarity between the new and old tables includes the title similarity S title , table content similarity S cell , and the table title and content similarity weighted score S title_cell .

3. The method for automatically updating financial professional report table data according to claim 2, characterized in that: The specific steps of calculating the similarity between a single old table and all remaining new tables belonging to the same large chapter mapping pair after removing all new tables that the old table has been associated with but failed to be associated with are as follows: S41, obtaining the title subject keywords of the current old table, matching the new table with the same title subject keywords from all new tables belonging to the same large chapter mapping pair according to the title subject keywords of the current old table, the old table with the old title that does not contain any subject keywords can only match the new table with the new title that does not contain any subject keywords, and the matched new table is called a similar new table; S42, respectively calculating the title similarity S between the current old table and each similar new table title Similarity S with table content cell ; S43, the title similarity S between the current old table and each similar new table title Similarity S with table content cell Calculate the corresponding table title and content similarity weighted score S respectively title_cell ; S44, calculate all the title similarity values ​​S title , table content similarity value S cell The weighted score S of the similarity between the table title and content title_cell Sort them in descending order according to the scores to get the title similarity sorting list {S title }、Table content similarity sorting list {S cell } and table title, content similarity weighted score sorting list {S title_cell }.

4. The method for automatically updating financial professional report table data according to claim 3, characterized in that: In step S42, the title similarity S between the current old table and a single similar new table is calculated. title The specific steps are: By comparing the set subject keyword table, extract the subject keywords of the current old table and similar new tables respectively; Determine the title matching score S based on the similarity between the subject keywords of the current old table and the subject keywords of the similar new table. k ey; Obtain the table title semantic vectors of the current old table and similar new tables respectively; Calculate the cosine similarity S of the new and old title vectors based on the semantic vectors of the table titles of the current old table and the similar new table v e c , the cosine similarity S of the new and old title vectors v e c As a reference value for title similarity; According to the title matching score S key The cosine similarity S between the new and old title vectors vec Calculate the title similarity S between the current old table and the similar new table title .

5. The method for automatically updating financial professional report table data according to claim 4, characterized in that: If the subject keywords of the current old table are exactly the same as the subject keywords of the similar new table, the title matching score S k e y =1; If the subject keywords of the current old table are partially consistent with the subject keywords of the similar new table, the title matching score S k e y =0.5; If the subject keywords of the current old table are inconsistent with the subject keywords of the similar new table, the title matching score S k e y =0.

6. The method for automatically updating financial professional report table data according to claim 4, characterized in that: The cosine similarity S of the new and old title vectors v e c The calculation formula is: Among them, v old is the semantic vector of the title of the current old table, v new is the semantic vector of the title of the similar new table; The title similarity S between the current old table and the similar new table title The calculation formula is: Among them, w k and w t All are set weight values.

7. The method for automatically updating financial professional report table data according to claim 3 or 4, characterized in that: If the current old table has no title, then the title similarity S between the current old table and the similar new table is calculated. title Set to 0.

8. The method for automatically updating financial professional report table data according to claim 3, characterized in that: In step S42, the table content similarity S between the current old table and a single similar new table is calculated. cell The specific steps are: Calculate the initial similarity S of the table contents of the two based on the normalized set of all cell contents of the current old table and the normalized set of all cell contents of the similar new table. cell_0 ; Among them, {O1,…,O n } is the normalized set of all cell contents in the current old table, {N1,…,N n } is the normalized set of all cell contents in the similar new table, and LENGTH() takes the length of the set; According to the relationship between the smallest chapter above the current old table and the smallest chapter above the similar new table, the initial similarity S of the table content is calculated. cell_0 Add the minimum chapter relationship weight to which the table belongs to obtain the table content similarity S between the current old table and the similar new table cell .

9. The method for automatically updating financial professional report table data according to claim 8, characterized in that: If the minimum chapter texts of the current old table and the similar new table are not empty and the minimum chapter texts of the two are inclusive, then S cell_0 ≥T contain_chapter , then S cell =S cell_0 +W contain_chapter ; If the minimum chapter texts of the current old table and the similar new table are not empty and the minimum chapter texts of the two are equal, then S cell_0 =0, then S cell =W contain_chapter ; If the minimum chapter texts of the current old table and the similar new table are not empty and the minimum chapter texts of the two are equal, then S cell_0 ≠0, then S cell =S cell_0 +W s ame_chapte r ; Among them, T contain_chapter is the set table content similarity threshold, W contain_chapter and W same_chapter These are the minimum chapter relationship weights of the set tables.

10. The method for automatically updating financial professional report table data according to claim 8 or 9, characterized in that: If the difference between the current old table and the similar new table in the smallest chapter text is the specified keyword, S cell =-1.

11. The method for automatically updating financial professional report table data according to claim 3, characterized in that: In step S5, the specific steps of screening and matching the new table that meets the conditions according to the calculated similarity value of the new and old tables according to the set most similar matching rule are: S51, determine the title similarity ranking list {S title The S with the largest score among title_max Does it satisfy S? title_max >T title_top If satisfied, then the similar new table corresponding to the title similarity value establishes a replacement association with the current old table, otherwise, execute step S52; S52, determine the title similarity ranking list {S title The S with the largest score among title_max Whether it satisfies 0 title_max <T title_min If the condition is satisfied, the current old table is deemed to have failed to match and will not be replaced. Step S4 is repeated to match the next old table. Otherwise, step S53 is executed.​ S53, determine the table title, content similarity weighted score sorting list {S title_cell The S with the largest score among title_cell_max Does it satisfy S? title_cell_max >T title_cell If the conditions are met, then the similar new table corresponding to the table title and content similarity weighted score is associated with the current old table for replacement, otherwise, step S54 is executed; S54, determine the table content similarity ranking list {S cell The S with the largest score among cell_max Does it satisfy S? cell_max <T cell_min If the condition is satisfied, the current old table is deemed to have failed to match and will not be replaced. Step S4 is repeated to match the next old table. Otherwise, step S55 is executed. S55, determine the table content similarity ranking list {S cell The S with the largest score among cell_max Does it satisfy S? cell_max >T cell If satisfied, then the similar new table corresponding to the table content similarity value establishes a replacement association with the current old table, otherwise, execute step S56; S56, determine the title similarity ranking list {S title The S with the largest score among title_max Does it satisfy S? title_max >T title If the conditions are met, the similar new table corresponding to the title similarity value is associated with the current old table for replacement. Otherwise, the current old table is deemed to have failed to match and is not replaced. Step S4 is repeated to match the next old table. Among them, T title_top , T title_min and T title are the set title similarity thresholds, T title_cell is the weighted threshold for table title and content similarity, T cell_min and T cell These are the set table content similarity thresholds.

12. The method for automatically updating financial professional report table data according to claim 11, characterized in that: The T title_min <T title <T title_top , T cell_min <T cell .

13. The method for automatically updating financial professional report table data according to claim 11, characterized in that: The title similarity threshold T title_min =0.5, T title =0.829, T title_t o p =0.98, table content similarity threshold T cell_min =0.04, T cell =0.082, table title and content similarity weighted threshold T title_cell =0.

7.

14. The method for automatically updating financial professional report table data according to claim 1, characterized in that: In step S7, the specific steps of performing a replacement judgment on the new table that is currently establishing a preliminary replacement association according to the set matching priority rule, and determining the new table that is currently establishing a final replacement association with the current old table according to the replacement judgment result are: S71, setting the current old table as the first old table, and setting the other old table corresponding to the new table hit by the other old tables as the second old table; The final matching status of the first old table and the second old table when performing the most similar matching rule is obtained respectively, and the final matching similarity score S of the first old table is obtained. replace _new, the final matching similarity score S of the second oldest table replace _old; S72, determine the priority status code C of the first old table according to the priority status code determination rule re _new and second oldest table priority status code C re _old; S73, judging the matching priority of the first old table and the second old table according to the final matching similarity scores and priority status codes of the first old table and the second old table, if the matching priority of the first old table is higher than the matching priority of the second old table, using the new table currently establishing a preliminary replacement association as the new table establishing a final replacement association with the first old table, and canceling the replacement association between the new table and the second old table, and repeatedly executing step S5 until the second old table matches the new table or does not match the new table; If the matching priority of the first old table is lower than the matching priority of the second old table, the replacement association between the first old table and the new table currently establishing the preliminary replacement association is cancelled, the replacement association between the second old table and the new table is kept unchanged, and step S5 is performed until the first old table matches the new table or does not match the new table; If the first old table or the second old table fails to match the new table, the first old table or the second old table that failed to match is deemed to have failed to match and will not be replaced.

15. The method for automatically updating financial professional report table data according to claim 14, characterized in that: The priority status code determination rule is: If the final matching status of the old table when executing the most similar matching rule is high title similarity, which meets the requirements for establishing a replacement association, then C re =3; If the final matching status of the old table when executing the most similar matching rule is that the weighted score of the table title and content similarity meets the requirements for establishing a replacement association, then C re =2; If the final matching status of the old table when executing the most similar matching rule is that the table content similarity meets the requirements for establishing a replacement association, then C re =1; If the final matching status of the old table when executing the most similar matching rule is that the table title similarity meets the requirements for establishing a replacement association, then C re =0; If the final matching status of the old table when executing the most similar matching rule is matching failure, then C re =-1.

16. The method for automatically updating financial professional report table data according to claim 14, characterized in that: The specific steps of determining the matching priority between the first old table and the second old table according to the final matching similarity scores and priority status codes of the first old table and the second old table in step S73 are: Determine whether the final matching similarity score and priority status code of the first old table and the second old table meet condition C re _new>C re _old, or C re _new=C re _old and S replace _new>S replace _old, if the condition is met, it means that the matching priority of the first old table is higher than the matching priority of the second old table, otherwise the matching priority of the first old table is lower than the matching priority of the second old table.

17. The method for automatically updating financial professional report table data according to claim 14, characterized in that: The final matching similarity scores of the first old table and the second old table are both similarity scores corresponding to the final matching states achieved when the most similar matching rule is executed.

18. A device for automatically updating financial professional report table data using the method according to any one of claims 1 to 17, characterized in that: include: A file processing module is used to extract table titles, table chapters and table contents from new and old files, and normalize all extracted table titles, table chapters and table contents; A document chapter mapping module is used to obtain a chapter mapping dictionary between new and old files based on the chapter as the mapping scale; The new-old table matching module is used to calculate the new-old table similarity between a single old table and all new tables belonging to the same large chapter mapping pair according to the chapter mapping dictionary, match the new table that meets the most similar matching rule according to the calculated new-old table similarity value, establish a preliminary replacement association between the new table that meets the most similar matching rule and the current old table, and then determine whether the new table that currently establishes the preliminary replacement association is a new table that has been hit by other old tables. If so, perform a replacement judgment on the new table that currently establishes the preliminary replacement association according to the set matching priority rule; The new and old table content replacement module is used to uniformly replace the contents of all new and old tables that implement replacement associations.

19. A computer device, characterized in that: The computer device comprises: a processor and a memory, wherein at least one program is stored in the memory, and the at least one program is loaded and executed by the processor to implement the method according to any one of claims 1 to 17.

20. A computer-readable storage medium, characterized in that: The readable storage medium stores at least one program, and the at least one program is loaded and executed by a processor to implement the method according to any one of claims 1 to 17.

21. A computer program product, characterized in that The computer program product includes at least one program segment, and the at least one program segment is stored in a computer-readable storage medium; the processor of the communication device reads the at least one program segment from the computer-readable storage medium, and the processor executes the at least one program segment, so that the communication device executes the method as described in any one of claims 1 to 17.

Images